Index, A to Z.
Every page, guide, example and public function. Type to filter; without JavaScript, the whole list is here.
Nothing matches. Try a function name like grad, or a task like join.
a
- Hugin.ablineabline ~slope ~intercept () draws, for each row, the line y = slope × x + intercept across the x domain of each panel, cut at the domain's edges in data, before a scale clamps, so a clamped scale draws the same line: a straight segment if the panel's x and y scales are both linear, and otherwise a curve through points one point of the page apart along x.val · hugin
- Ymir.Transform.aboutabout c maps directions to angular offsets about c, in radians, x east and y north: the inverse of celestial Arc at c with LONPOLE 180°.val · ymir
- Jera.Tol.absabs a is v ~rel:0.val · jera
- Nx.absabs x is x's absolute value.val · nx
- Nx_kernel.Prog.E.absval · nx
- Tolk.Bigint.absabs n is the absolute value of n.val · tolk
- Hugin.Theme.accentaccent th is the accent of th.val · hugin
- Hugin_kit.Scheme.accentaccent is Brewer's Accent, eight colours.val · hugin
- Rig_remote_proxy.Wire.acceptaccept fd ~key ~admit runs the accepting end's handshake on the connected socket fd, and is the dialing end's process and this end's, as the dialing end names it.val · rig
- Nx_kernel.acceptsaccepts r dt is true iff r reduces terms of dt: Sum and Prod every dtype but booleans, Max, Min and Arg every dtype, Logsumexp and Moments floats.val · nx
- Nx_kernel.Prog.accepts1accepts1 k x y is true iff k takes an operand of x to a result of y.val · nx
- Nx_kernel.Prog.accepts2accepts2 k x is true iff k takes two operands of x.val · nx
- Nx_kernel.Prog.accepts3accepts3 k c x is true iff k takes a first operand of c and two of x.val · nx
- Rig.accessThe type for accesses to memory: what a buffer admits, what a host access waits for, and how fixed memory is used (Submission.make).type · rig
- Rig.Buffer.accessaccess b is the access b admits, which its views and borrows admit too.val · rig
- Rig_mlx5.Region.accessThe type for what peers may do: to a region (register), and through a queue pair (Qp.connect).type · rig
- Rig_remote_proxy.Wire.accountThe type for an agent's account of one of its devices.type · rig
- Nx_kernel.accumulatesaccumulates dt is true iff a contraction computes in dt: every dtype but booleans and floats narrower than float32.val · nx
- Kaun.Metric.accuracyaccuracy predictions labels is the fraction of examples whose predicted class equals their label.val · kaun
- Nx.acosacos x is the arccosine, in [0, π]: on complex numbers its real part, with asin's branch cuts.val · nx
- Nx_kernel.Prog.E.acosval · nx
- Nx.acoshacosh x is the inverse hyperbolic cosine: 0.val · nx
- Nx_kernel.Prog.E.acoshval · nx
- Rig_nv_abi.Method.acquireacquire addr v stops the channel until the semaphore w at addr is at least v in circular order: until w - v, computed modulo 2{^64}, is below 2{^63}.val · rig
- Talon.Expr.acrossacross k sel f is the outputs f n (Col.v k n) for each name n that sel selects, in order: across Kind.float Sel.(prefix "wk") (fun n x -> n := over (rank x)).val · talon
- Fehu_envs.Cartpole.acttype · fehu
- Fehu_envs.Grid_world.acttype · fehu
- Fehu_envs.Mountain_car.acttype · fehu
- Fehu_envs.Random_walk.acttype · fehu
- Fehu.Env.action_spaceaction_space env is the space of valid actions.val · fehu
- Fehu.Vec_env.action_spaceaction_space t is the shared action space.val · fehu
- Tolk.Search.actionsactions () is the optimisations the search tries under the current settings, in order:val · tolk
- Quill_markdown.Edit.active_blockactive_block t ~cursor is the block containing byte offset cursor, or None if cursor is outside all blocks.val · quill
- Vega.adafactor_initadafactor_init p params is the initial state for optimizing params: zero statistics and step = 0.val · vega
- Vega.adafactor_ptreeadafactor_ptree p is Nx.Ptree.nest (module Adafactor_state) p.val · vega
- Vega.adafactor_stateThe state for adafactor_step.type · vega
- Vega.Adafactor_stateThe structure of Adafactor states.module · vega
- Vega.adafactor_stepadafactor_step p ~lr st ~params ~grads is (params', st') after one Adafactor step (Shazeer and Stern, 2018): an adaptive step that keeps no first moment and, for a factored leaf of m rows and n columns, estimates the second moment from m + n numbers rather than m * n.val · vega
- Vega.adagrad_initadagrad_init p params is the initial state for optimizing params: an all-zero sum and step = 0.val · vega
- Vega.adagrad_ptreeadagrad_ptree p is Nx.Ptree.nest (module Adagrad_state) p.val · vega
- Vega.adagrad_stateThe state for adagrad_step: the sum of all squared gradients so far, with the shape of the parameters, and the number of completed steps as a scalar tensor.type · vega
- Vega.Adagrad_stateThe structure of Adagrad states.module · vega
- Vega.adagrad_stepadagrad_step p ~lr st ~params ~grads is (params', st') after one Adagrad step (Duchi, Hazan and Singer, 2011).val · vega
- Vega.adam_initadam_init p params is the initial state for optimizing params: all-zero moments and step = 0.val · vega
- Vega.adam_ptreeadam_ptree p is Nx.Ptree.nest (module Adam_state) p, the structure of an Adam state over parameters of structure p.val · vega
- Vega.adam_statetype · vega
- Vega.Adam_stateThe structure of Adam states.module · vega
- Vega.adam_stepadam_step p ~lr st ~params ~grads is (params', st') after one Adam step (Kingma and Ba, 2015).val · vega
- Vega.adamw_initadamw_init is adam_init: AdamW shares Adam's state.val · vega
- Vega.adamw_stepadamw_step p ~lr st ~params ~grads is like adam_step with decoupled weight decay (Loshchilov and Hutter, 2019): with d Adam's bias-corrected direction, the parameter update becomesval · vega
- Vega.adan_initadan_init p params is the initial state for optimizing params: every part zero and step = 0.val · vega
- Vega.adan_ptreeadan_ptree p is Nx.Ptree.nest (module Adan_state) p.val · vega
- Vega.adan_stateThe state for adan_step.type · vega
- Vega.Adan_stateThe structure of Adan states.module · vega
- Vega.adan_stepadan_step p ~lr st ~params ~grads is (params', st') after one Adan step (Xie et al., 2022), an adaptive Nesterov momentum.val · vega
- Jera.Piecewise.adaptadapt s ~degree ~tol ~budget f a b matches f on [a, b] to tol by series of degree.val · jera
- Jera.Quad.adaptiveadaptive r ~tol ~budget f range is the integral of f over each element of range.val · jera
- Fehu.Buffer.addadd buf tr appends tr, overwriting the oldest transition when at capacity.val · fehu
- Hugin_kit.Time.addadd ~tz_offset_s i n t is t moved by n strides of i, towards the future iff n is positive.val · hugin
- Nx.addadd a b is the elementwise sum.val · nx
- Nx.Ptree.Path.addadd seg p is p then seg, in constant time, sharing p.val · nx
- Nx_kernel.Prog.E.addval · nx
- Nx_wide.addadd a b is a + b within 3u² / (1 - 4u) relative.val · nx
- Rune.Total.addadd t v adds v to the innermost open collect of t, and does nothing if none is open.val · rune
- Talon.Expr.Temporal.addadd a d is a advanced by d, of a's type: a datetime, a duration or a clock, d being a duration of a's unit or a coarser one, or a date, d being whole days.val · talon
- Talon.Record.addadd k name v r is r with a last field name of kind k holding v, or null if v is None.val · talon
- Tolk.Bigint.addadd n0 n1 is n0 + n1.val · tolk
- Tolk.Ops.addval · tolk
- Tolk.Ops.Elementwise.addval · tolk
- Tolk.Ops.Upat.addval · tolk
- Tolk.Renderer.Estimates.addadd e0 e1 is the estimate of both e0's and e1's work: each count is the sum of theirs, exactly: an Int where it fits one, else a constant node.val · tolk
- Ymir.Observation.addadd a b is a's data plus b's, in a's unit, on a's grid.val · ymir
- Ymir.Time.addadd t span is t moved by span, a time, their shapes broadcast.val · ymir
- Ymir_units.Quantity.addadd a b is map2 Nx.add a b.val · ymir
- Ymir_fits.Fits.Header.add_commentaryadd_commentary k s h appends s as k records of 72 bytes each.val · ymir
- Hugin_vg_vector.Vector.add_exactadd_exact b v adds the shortest decimal that reads back as v, in %g notation, and 0 for zero, whatever its sign.val · hugin
- Hugin_vg_vector.Vector.add_fixedadd_fixed b d v adds v rounded to d decimals, without trailing zeros or point, and 0 for a number that rounds to zero, whatever its sign.val · hugin
- Quill.Doc.add_outputadd_output id o d appends o to cell id with Cell.append_output, except that a display with id Some s replaces in place the display with id Some s of find_display s d, in whichever cell it is.val · quill
- Quill_top.add_packagesadd_packages pkgs resolves each findlib package name and adds its directory to the toplevel load path, marking each as already linked into the executable.val · quill
- Nx.add_sadd_s x c is add x c.val · nx
- Munin.Store.add_tagsadd_tags s id tags adds tags to the run id.val · munin
- Brot.add_tokensadd_tokens t tokens is t with tokens registered as added tokens, exactly as if they had been passed as added_tokens at construction: matched atomically in the input, numbered from the end of the vocabulary, and skipped by decode when their special is set.val · brot
- Brot.added_tokenThe type for added token configurations.type · brot
- Brot.added_tokenadded_token content is an added token configuration for content.val · brot
- Brot.added_tokensadded_tokens t is t's added tokens: those given as added_tokens at construction, plus the bos_token, eos_token and pad_token role markers.val · brot
- Brot.Post_processor.added_tokensadded_tokens t ~is_pair is the number of special tokens t adds.val · brot
- Rig_elf.addendThe type for a relocation's addend.type · rig
- Tolk.Dtype.addr_spaceThe type for address spaces: the memory a pointer addresses.type · tolk
- Tolk.Dtype.addr_space_of_stringaddr_space_of_string s is the address space named s: GLOBAL, LOCAL, REG or ALU, the name pp_addr_space formats after AddrSpace..val · tolk
- Rig.Submission.addressaddress s b is the address of b's first byte as the work of s's device addresses it, where b's bytes lie within a buffer of s's fixed memory.val · rig
- Rig_bin.AddressMachines and addresses, written one way: a name or an address, an IPv6 address in brackets, as URIs write it: h100-b, 10.0.0.2, [fd00::2], [::1]:7000.module · rig
- Rig_mlx5.Qp.addressThe type for the addresses of ports.type · rig
- Rig_pci.Memory.addressaddress r is the GPU address of r's first byte.val · rig
- Rig_pci.Space.addressaddress x is the first address of x's range.val · rig
- Rig_pci.Window.addressaddress w is the address of w's first byte on its machine.val · rig
- Rig_pci.Function.addressingaddressing f is how f reaches system memory.val · rig
- Rig_pci.Machine.addressingThe type for how a function reaches system memory.type · rig
- Tolk.Ops.addrspaceaddrspace u is the address space u lives in.val · tolk
- Sowilo.adjust_brightnessadjust_brightness factor img scales pixel values by factor and clips to [0, 1].val · sowilo
- Sowilo.adjust_contrastadjust_contrast factor img adjusts contrast around the per-channel mean.val · sowilo
- Sowilo.adjust_gammaadjust_gamma gamma img applies gamma correction: img ** gamma.val · sowilo
- Sowilo.adjust_hueadjust_hue delta img rotates hue by delta.val · sowilo
- Sowilo.adjust_saturationadjust_saturation factor img adjusts color saturation via HSV.val · sowilo
- Tolk.Stage.admitsadmits ren sink is true iff operands stages every operand of each tensor core of sink on ren, tiles permitting (depth): the question the heuristic asks before it gives a core the staged tile (Heuristic.hand_coded_optimizations).val · tolk
- Hugin_font.Font.advanceadvance f g is the horizontal distance the pen moves after drawing g (hmtx).val · hugin
- Kaun.Cache_index.advanceadvance index is the index of the next token of every lane.val · kaun
- Hugin_gg.AffineAffine maps of the plane.module · hugin
- Norn.Bij.affineaffine ~loc ~scale maps u to loc + scale u, onto the reals.val · norn
- Norn.Bij.affine_trilaffine_tril ~loc ~scale_tril maps a vector u to loc + L u, L the lower triangle of scale_tril, onto the reals.val · norn
- Tolk.Ops.afterafter u deps is u ordered after deps, or u if deps is empty.val · tolk
- Tolk.Ops.Upat.afterval · tolk
- Tolk.Ops.after_disjointafter_disjoint u stores is after u stores with a promise from its maker: stores, the stores of one write of u's storage, write pairwise disjoint elements, none reads an element another writes, and every kernel of them may run at once.val · tolk
- Ymir.Cosmology.ageage c z is the time since the big bang at z.val · ymir
- Rig_bin.Agentrig agent: a machine's half of a job, and the agent it starts.module · rig
- Rig_bin.Agent.agentagent ~firmware host port is the agent listening at host and port.val · rig
- Rig_bin.Help.agentagent is rig agent's page.val · rig
- Rig_remote.agentThe type for agents of this machine.type · rig
- Rig_remote_proxy.Wire.agentThe type for the agents of a job, as a join names them.type · rig
- Talon.Expr.aggThe shape of expressions with one value per frame.type · talon
- Talon.Query.aggregateaggregate ~by os q is one row per group of q's rows that have the same keys in the columns by, by key identity (null is one key, and NaN is one), in order of first appearance.val · talon
- Ymir.Grid.agreeagree a b is where a and b are the same cells in the same world: false if their kinds, shapes, stages, codes, frames or units differ, and otherwise whether every parameter holds the same numbers, NaN equal to NaN, one boolean per batch element.val · ymir
- Munin.Store.aliasesaliases s name is name's aliases and the numbers of the versions they name, sorted by alias.val · munin
- Rig.Buffer.alignedaligned b ~at n is true iff the address at bytes past b's first byte, in the addresses of b's device's work, is a multiple of n, for n a power of two.val · rig
- Brot.Normalizer.alignmentThe type for maps from the bytes of a normalized text back to the bytes of the text it was normalized from.type · brot
- Rig.alignmentalignment d is the power of two that the first byte of every non-empty buffer Buffer.create makes on d starts on, whatever its kind and size, as d's driver states it from its allocator (Rig_edge.facts): 16 for host, 64 for a memory_device, and 1 for an io device.val · rig
- Hugin_kit.Dash.allall is solid, dashed, dotted and dash_dot, in this order: the patterns of the categories of a band scale that sets none.val · hugin
- Nx.allall ~axes x is true where every element is non-zero, as cast into bool reads it: a NaN is non-zero, and a complex number is where a part is.val · nx
- Nx_array.Dtype.allall is every dtype, in code order.val · nx
- Talon.Join.allall matches every pair of rows.val · talon
- Talon.Sel.allall selects every column, in schema order.val · talon
- Tolk.Dtype.allall is every data type a program stores: floats @ ints @ [Bool].val · tolk
- Tolk.Op.Set.allall is every operation.val · tolk
- Quill_project.all_notebooksall_notebooks toc flattens toc into an ordered list of all notebooks, including placeholders.val · quill
- Tolk.Helpers.all_sameall_same equal l is true iff every element of l is equal to the first.val · tolk
- Rig_amd.allocalloc d m n is n bytes of d's memory m, or None if d has none.val · rig
- Rig_cuda.allocalloc d m n is n bytes of d's memory m, or None if d has none.val · rig
- Rig_edge.Driver.allocalloc d m n is n bytes of d's memory m, or None if d has none.val · rig
- Rig_edge.Io.allocalloc d n is n new bytes of d's memory, or None if d has no room.val · rig
- Rig_metal.allocalloc d m n is n bytes of d's memory m, or None if d has none.val · rig
- Rig_nv.allocalloc d m n is n bytes of d's memory m, or None if d has none.val · rig
- Rig_pci.Memory.allocalloc m k n is n new bytes of kind k.val · rig
- Rig_pci.Page_table.allocalloc t h n is n bytes, rounded up to 4 KiB, of new physical memory from the pool palloc takes from by default, mapped at new virtual addresses of space t, with h's hold on them (Space.alloc).val · rig
- Rig_pci.Space.allocalloc s h n is h's hold on n free addresses of s, aligned to a, the larger of align (defaults to 4096) and the largest power of two not above n, so that large ranges map with large pages.val · rig
- Rig_remote_proxy.allocalloc d m n is n bytes of d's memory m, or None if d has none.val · rig
- Tolk.Shape.allocalloc ~slot shape dt is call-local storage of shape, in slot (default Ops.unique_num).val · tolk
- Tolk.Support_memory.Tlsf_allocator.allocalloc ~align a n is the first address of a new block of max block_size n addresses, a multiple of align (default 1) after base.val · tolk
- Rig_pci.Function.alloc_dmaalloc_dma f n is Some (w, runs), with w on n new, zeroed bytes of f's machine, rounded up to Machine.page, and runs the addresses at which f reaches them.val · rig
- Tolk.Shape.alloc_likealloc_like ~slot u is alloc of u's greatest shape and type.val · tolk
- Rig.allocatedallocated d is the bytes that d holds now of the memory that counts in its budget (Reclamation).val · rig
- Rig_elf.allocatedallocated ~machine s is true iff s, a section of an object for machine (e_machine), is code or data a loader's memory holds.val · rig
- Tolk.Shape.allocated_startallocated_start is (16, 0), the start of storage that scheduling allocates: on a 16-byte boundary.val · tolk
- Tolk.Setting.allow_tf32allow_tf32 lets float32 matrix multiplications use TF32 tensor cores on NVIDIA devices, from ALLOW_TF32.val · tolk
- Hugin_gg.Color.alphaalpha c is the opacity of c, from 0.val · hugin
- Ymir_units.Codata.alpha_particle_massalpha_particle_mass r is mα, the mass of the helium-4 nucleus, in kg.val · ymir
- Tolk.Op.Set.alualu is the arithmetic and logic operations: unary, binary and ternary.val · tolk
- Tolk.Ops.alualu x op rest is the operation op on x :: rest as given.val · tolk
- Tolk.Ops.Elementwise.alualu x op rest is the operation op on x :: rest as given.val · tolk
- Tolk.Ops.Upat.alualu x op rest is the operation op on x :: rest as given.val · tolk
- Rig_bin.Images.amdamd is rig.amd.pci's list.val · rig
- Tolk.Tc.amdamd arch is the tensor cores of the AMD architecture arch: amd_cdna3 for gfx942, amd_cdna4 for gfx950, amd_rdna4 for gfx1200 and gfx1201, and amd_rdna3 for any other.val · tolk
- Tolk.Tc.amd_cdna3amd_cdna3 is CDNA 3's tensor cores, of K 32 for the 8-bit floats without infinities and 16 for the 16-bit floats.val · tolk
- Tolk.Tc.amd_cdna4amd_cdna4 is CDNA 4's tensor cores, of K 128 and 32 for the 8-bit floats, 32 and 16 for the 16-bit floats.val · tolk
- Tolk.Tc.amd_rdna3amd_rdna3 is RDNA 3's 16 x 16 x 16 tensor cores.val · tolk
- Tolk.Tc.amd_rdna4amd_rdna4 is RDNA 4's 16 x 16 x 16 tensor cores, of the 16-bit floats, of Dtype.fp8_ocp into Dtype.t.Float32 and of Dtype.t.Int8 into Dtype.t.Int32.val · tolk
- Tolk.Helpers.amdgpu_disassembleamdgpu_disassemble lib prints the instructions of the AMD GPU code object lib, as find_llvm_objdump's llvm-objdump -d prints them, without the padding instructions that end it (s_nop 0 and s_code_end).val · tolk
- Ymir_units.Unit.ampereampere is the symbol A.val · ymir
- Ymir.Healpix.analysisanalysis h ~lmax ~iterations f is the coefficients [...; lmax + 1; lmax + 1] of each map of f, [...; pixels h] in h's scheme, by the quadrature over pixel centres refined iterations times.val · ymir
- Jera.System.andersonanderson ~memory is for a fixed point x = g x stated as the residual f x = g x − x.val · jera
- Nx.angleangle dt z is z's argument in radians, atan2 of its imaginary and real parts, computed as magnitude is.val · nx
- Ymir.Cosmology.angular_diameter_distanceangular_diameter_distance c z is DM(z)/(1 + z).val · ymir
- Ymir.Region.annulusannulus p ~inner ~outer is the ring between the discs of radii inner and outer about the origin of p's plane.val · ymir
- Tolk.Helpers.ansilenansilen s is the number of characters of ansistrip s, counting UTF-8 encoded characters once.val · tolk
- Tolk.Helpers.ansipadansipad s w is s followed by the spaces that bring its ansilen to w, none if it has that length already.val · tolk
- Tolk.Helpers.ansistripansistrip s is s without its ANSI escape sequences that erase a line or set graphic attributes.val · tolk
- Nx_array.answerThe type for what a kernel answers, and the door.type · nx
- Rig_remote_proxy.Link.answeranswer l r a sends a as the answer to r.val · rig
- Talon.Source.answerThe type for a source's answers about a conjunct.type · talon
- Nx.anyany ~axes x is true where some element is non-zero, as all.val · nx
- Nx_array.anyThe type for arrays whose dtype is chosen at run time.type · nx
- Nx_array.Dtype.anyThe type for dtypes chosen at run time.type · nx
- Nx_kernel.Prog.E.anyThe type for expressions of any dtype.type · nx
- Nx_kernel.Prog.E.anyany e is e.val · nx
- Talon.Type.anytype · talon
- Tolk.Ops.Upat.anyany ps matches what any of ps matches.val · tolk
- Nx_kernel.any_floatThe type for a float dtype, whatever its storage.type · nx
- Hugin_gg.Path.appendappend q p is the subpaths of p followed by those of q.val · hugin
- Norn.Draws.appendappend u d d' is the draws of d followed, in every chain, by those of d'.val · norn
- Talon.Query.appendappend rest q is q's rows, then rest's, written q |> append rest.val · talon
- Tolk.Ops.Pattern_matcher.appendappend m0 m1 tries m0's rules, then m1's, each declining as its matcher says.val · tolk
- Quill.Cell.append_outputappend_output o c appends o to the outputs of c, except that a display with id Some d replaces in place a display of c with id Some d, if there is one.val · quill
- Tolk.Postrange.Scheduler.applied_optsapplied_opts k is the optimisations applied to k, in order.val · tolk
- Brot.Normalizer.applyapply n s is s normalized by n.val · brot
- Kaun.Attention.applyapply ~head_dim p x is multi-head self-attention over x, with:val · kaun
- Kaun.Batch_norm.applyapply p stats ~training x is (y, stats') where y normalizes x per feature and applies the affine transform: y = gamma * (x - mean) / sqrt (var + eps) + beta.val · kaun
- Kaun.Conv.applyapply p x cross-correlates p's filters with x and adds the bias (no shift when p.b is None).val · kaun
- Kaun.Dropout.applyapply ~rate ~training ?key x is x with dropout applied.val · kaun
- Kaun.Embedding.applyapply p ids gathers the table row of each id: the result has ids's shape with a trailing axis of size dim appended, and its (i, ..., :) slice is row ids.(i, ...) of p.table.val · kaun
- Kaun.Layer_norm.applyapply p x normalizes each vector along x's last axis and rescales it:val · kaun
- Kaun.Linear.applyapply p x is x @ p.w + p.b (the shift is omitted when p.b is None).val · kaun
- Kaun.Rms_norm.applyapply p x scales each vector along x's last axis to unit root mean square and rescales it:val · kaun
- Kaun.Rope.applyapply t ~pos x rotates x, of shape [| batch; heads; seq; head_dim |], by the positions pos, of shape [| batch; seq |] (or [| 1; seq |], shared by the batch).val · kaun
- Nx_quant.applyapply w x is w applied to x as a linear map from k inputs to n outputs, summed at acc, float64 for a float64 x and float32 otherwise, which holds each value of w exactly, and rounded once to x's dtype:val · nx
- Ymir.Transform.applyapply t x maps x through each stage.val · ymir
- Brot.Normalizer.apply_alignedapply_aligned n s is apply n s with its alignment on s.val · brot
- Nx_quant.apply_expertsapply_experts ~ids w x is each row of x through the expert its id names: at each position p of ids, the product apply defines of w's matrix ids.{p} with p's row of x, or zeros where ids.{p} is not in [0, e), whatever that row holds.val · nx
- Tolk.Indexing.apply_movement_opapply_movement_op in_shape m idxs is the index, into a source of shape in_shape, of the element that the movement m of that source places at the index idxs: one node per axis of the source, from one per axis of the result.val · tolk
- Tolk.Postrange.Scheduler.apply_optapply_opt k opt applies opt to k, records it among k's applied optimisations if append_opt (default true), and is Ok the axes it made: a split's [r'; s] (shift_to), a pad's grown axis, a tensor core's three axes N, M and K, and [] for a swap.val · tolk
- Tolk.Postrange.apply_optsapply_opts ~beam ~hand_coded ast ren is the kernel ast optimised for ren (Scheduler.get_optimized_ast), named by its argument unless that is "test".val · tolk
- Quill.Session.apply_outputapply_output id output s adds output to cell id in the document with Doc.add_output, so that a display with an id replaces the earlier display with that id.val · quill
- Nx.arangearange dt start stop step is the 1-d host value start, start + step, … before stop, empty where stop is not past start in step's direction.val · nx
- Tolk.Shape.arangearange ~start ~step ~dtype stop is the vector of the integers from start (default 0) up to stop, excluded, by step (default 1), down to it for a negative step, of type dtype (default the first of Dtype.default_int, Dtype.t.Int32, Dtype.t.Int64 and Dtype.t.Uint64 that holds them).val · tolk
- Nx.arange_atarange_at p dt start stop step is arange dt start stop step at p.val · nx
- Nx.arange_farange_f dt start stop step is the 1-d host value of the ⌈(stop - start) / step⌉ values start + i step, each computed at float64 with one rounding, then stored in dt; empty where that count is not positive.val · nx
- Nx.arange_f_atarange_f_at p dt start stop step is arange_f dt start stop step at p.val · nx
- Hugin_gg.Path.arcarc c r ~start ~sweep p is p with the circular arc of centre c and radius r from angle start through sweep radians, positive sweeps turning from the positive x axis towards the positive y axis.val · hugin
- Rig.archarch d names the code d's processor runs.val · rig
- Tolk.Host_config.architectureval · tolk
- Nx_io.ArchiveArchives of entries.module · nx
- Ymir_units.Unit.arcminutearcminute is π/10800 rad.val · ymir
- Ymir_units.Unit.arcsecondarcsecond is π/648000 rad.val · ymir
- Hugin.areaarea ~y () fills, for each series of its rows (Mark.series), as line splits them, the region between the curve through the points of y and the curve through those of its baseline (Curve.area), with fill, the accent by default, at opacity.val · hugin
- Hugin_gg_kit.Pgon2.areaarea p is the signed area of the rings of p, the sum of their Ring2.areas.val · hugin
- Hugin_gg_kit.Ring2.areaarea r is the signed area of r, computed with rounding: positive if r is positively oriented and negative if negatively, unless the area is within rounding error of zero, and 0.val · hugin
- Hugin_kit.Curve.areaarea c ~x0 ~y0 xs ys is the region between curve c through the points (xs.(i), ys.(i)) and curve c through the baseline points (x0.(i), y0.(i)), each drawn forwards as path draws it.val · hugin
- Hugin_vg_vector.Vector.areaarea m cut q sink gives sink the subpaths of q mapped through m into a frame, each closed, as filling and clipping see them, as Hugin_gg.Path.crop crops them to cut.val · hugin
- Rig_program.areaThe type for the areas of a rail's end (Rails).type · rig
- Rig_remote_abi.areaThe type for the host memory of an end.type · rig
- Ymir.Healpix.areaarea h is each pixel's solid angle, π / (3 nside²) steradians, rounded once, on the host.val · ymir
- Ymir.Observation.areaarea o is the measure o's data are per, when not the cells' own.val · ymir
- Hugin_kit.Scale.areasareas s is the symbol areas of s, if set.val · hugin
- Tolk.Ops.argThe type for node arguments.type · tolk
- Tolk.Ops.argarg u is u's argument.val · tolk
- Talon.Expr.arg_maxarg_max a is the position of a's first greatest value, like arg_min.val · talon
- Talon.Expr.arg_minarg_min a is the zero-based position in the frame of a's first least value, as int64.val · talon
- Nx.argmaxargmax ~axis x is the position along axis of the first element greatest as max orders them: the first NaN where there is one, and 1 for [-0.; 0.].val · nx
- Nx.argminargmin ~axis x is the position of the first element least as min orders them: the first NaN where there is one, and 1 for [0.; -0.].val · nx
- Nx.argsortargsort ~descending ~axis x is snd (sort ~descending ~axis x).val · nx
- Tolk.Helpers.argsortargsort l is the positions of the elements of l, ordered by increasing element; equal elements keep their order.val · tolk
- Nx.argwhereargwhere t is the coordinates of t's elements other than zero, a row each in C order: of shape [k; ndim t] for k such elements, whose row i is the coordinates in t's shape of the ith entry of positions (flatten (not_equal t (zeros_like t))).val · nx
- Rig_mlx5.Cq.armarm cq asks cq's NIC to raise one Completion event at the next completion of cq after those poll consumed.val · rig
- Rig_mlx5_abi.Doorbell.armarm ~sequence ~count ~cq is the two 32-bit words of the store that arms completion queue number cq at consumer count count, after which the NIC raises an event at the next completion.val · rig
- Rig_mlx5_abi.Doorbell.armedarmed is the byte offset in a completion queue's record of its arm word, the first word of arm.val · rig
- Ymir.Grid.aroundaround x ~shape g is the shape block of g's base centred on the cell holding each point x, located through the inverse transform, with the extra cell of an even size on the high side.val · ymir
- Ymir.Observation.aroundaround x ~shape o is o on Grid.around x ~shape (grid o), sliced as window slices.val · ymir
- Nx.Repr.arrayarray x is Some a iff x lies on one device, a its array, and None for a value on several devices.val · nx
- Array OperationsThis guide covers reshaping, broadcasting, joining, slicing, and the views that let Nx rearrange arrays without copying them.guide · nx
- Nx.array_equalarray_equal a b is the 0-d true iff a and b have one shape and equal holds at every index: false where an element is a NaN, and true for two values without elements.val · nx
- Munin.ArtifactArtifact versions.module · munin
- Munin.Store.artifactartifact s name sel is the version of name that sel selects: its latest (Artifacts), the version numbered n, or the version the alias names.val · munin
- ArtifactsAn artifact is a name, such as model, with versions numbered from 1.guide · munin
- Munin.Store.artifactsartifacts s is the versions of s not deleted, by name, then by number.val · munin
- Tolk.Shape.as_shapeas_shape u is the shape the node u denotes: its constant, its stack's sources, or u itself.val · tolk
- Talon.Order.ascasc name orders the column name ascending, nulls last.val · talon
- Hugin_font.Font.ascentascent f is the distance above the baseline that lines of f reserve: the typographic ascender of the OS/2 table if OS/2 sets USE_TYPO_METRICS, and the ascender of the hhea table otherwise.val · hugin
- Nx.asinasin x is the arcsine, in [-π/2, π/2]: on complex numbers its real part, with branch cuts on the real axis beyond -1 and 1.val · nx
- Nx_kernel.Prog.E.asinval · nx
- Nx.asinhasinh x is the inverse hyperbolic sine, odd, x itself near 0.: within 3 ulps at float32 and float64.val · nx
- Nx_kernel.Prog.E.asinhval · nx
- Nx_cpu.assembleassemble ~fill ~dst pieces stores into dst, at each index, the element at that index of the last piece whose region holds it, or fill where none does.val · nx
- Nx_kernel.S.assembleassemble ~fill ~dst pieces stores into dst, at each index, the element at that index of the last piece whose region holds it, or fill where none does.val · nx
- Nx_kernel.Shape.assembleassemble y regions is the shape of each piece of an assembly of shape y: its region's, the elements regions.(i) keep of each axis of y, as Nx_array.Move.t.Slice keeps them.val · nx
- Tolk.Setting.assert_compileassert_compile makes compiling a source whose binary no disk cache table holds raise (Renderer.Compiler.compile_cached), from ASSERT_COMPILE.val · tolk
- Tolk.Op.Set.associativeassociative is the binary operations that rewrites may regroup, each satisfying f (f x y) z = f x (f y z): Add, Mul, And, Or, Max and Xor.val · tolk
- Nx.associative_scanassociative_scan s f x is the inclusive scan of f along axis over the structure x: its element i combines x's elements 0 to i by f, each earlier run as f's first operand, an element being the structure of every array's slice at one index of its axis.val · nx
- Tolk.Ops.assumingassuming u subs is u with each node of subs replaced by its pair, as substitute does outside calls, for a fact that holds only where u is read: it leaves each range it does not replace whole, bound included, since a loop is one loop wherever it is read.val · tolk
- Tolk.Postrange.Scheduler.astast k is k's kernel.val · tolk
- Hugin_kit.Symbol.asteriskasterisk is three strokes of equal length crossing at their middles, a sixth of a turn apart, one parallel to the y axis.val · hugin
- Ymir_units.Unit.astronomical_unitastronomical_unit is 149597870700 m.val · ymir
- Nx.atanatan x is the arctangent, in [-π/2, π/2]: on complex numbers its real part, with branch cuts on the imaginary axis beyond -i and i.val · nx
- Nx_kernel.Prog.E.atanval · nx
- Nx.atan2atan2 y x is the angle of the point (x, y), in [-π, π]: -π for y = -0.val · nx
- Nx_kernel.Prog.E.atan2val · nx
- Nx.atanhatanh x is the inverse hyperbolic tangent: inf at 1., -inf at -1., NaN beyond them, odd, x itself near 0.; within 3 ulps at float32 and float64.val · nx
- Nx_kernel.Prog.E.atanhval · nx
- Rig_amd_pci.attachattach i gives GPU i of this machine back to amdgpu as Rig_pci.Driver.attach does, resetting it first as reset does if it is detached.val · rig
- Rig_bnxt.attachattach i gives NIC i of this machine back to its kernel driver as Rig_pci.Driver.attach does.val · rig
- Rig_nv_pci.attachattach i gives GPU i of this machine back to its kernel driver, resetting it first as reset does if no kernel driver has it (Rig_pci.Driver.attach).val · rig
- Rig_pci.Driver.attachattach d m i gives function i of m back to its kernel driver.val · rig
- Kaun.Attention.attendattend q k v is grouped-query attention of q, of shape [| batch; heads; n; d |], over k and v, of shape [| batch; kv_heads; m; d |] and [| batch; kv_heads; m; dv |]: the result has shape [| batch; heads; n; dv |].val · kaun
- Kaun.AttentionMulti-head self-attention, and the pure Attention.scaled_dot_product_attention core.module · kaun
- Nx.attentionattention ~acc ~kv q is (o, lse) for q [lead…; hq; tq; dk], hk dividing hq (zero heads divide zero), query head h reading key head h / (hq / hk).val · nx
- Nx_cpu.attentionattention p ~scale ~o ~lse ?lengths q kv loads stores the attention of q's rows over kv (Attention), p the pair's program over its loads loads, into o and lse, whose dtype is the accumulator.val · nx
- Nx_kernel.S.attentionattention p ~scale ~o ~lse ?lengths q kv loads stores the attention of q's rows over kv (Attention), p the pair's program over its loads loads, into o and lse, whose dtype is the accumulator.val · nx
- Nx_kernel.Shape.attentionattention q kv ~lengths loads is the shapes of attention's o and lse, [lead…; hq; tq; dv] and [lead…; hq; tq], for q [lead…; hq; tq; dk].val · nx
- Brot.Encoding.attention_maskattention_mask enc is 1 for real tokens and 0 for padding tokens.val · brot
- Nx_kernel.attention_refusalattention_refusal p ~q ~rank ~o ~acc ~loads:v loads is as map_refusal for S.attention's program p over loads, read by v, a query of dtype q and rank rank, o of dtype o and lse of dtype acc.val · nx
- Ymir_units.Unit.attoatto u is 10-18 u.val · ymir
- Quill.Cell.attrsThe type for cell display attributes.type · quill
- Quill.Cell.attrsattrs c is the display attributes of cell c.val · quill
- Kaun.Metric.auc_rocauc_roc scores labels is the area under the ROC curve of a binary classifier: the probability that a uniformly drawn positive example outscores a uniformly drawn negative one, tied pairs counting half — equivalently, trapezoidal integration of the ROC curve.val · kaun
- Kaun.Metric.averageThe type for combining per-class scores.type · kaun
- Kaun.Pool.avg_pool2davg_pool2d ~kernel_size x is like max_pool2d except each window is reduced to its mean.val · kaun
- Ymir_units.Unit.avogadroavogadro is NA, 6.02214076·1023 mol-1.val · ymir
- Ymir.Transform.axesaxes p ~origin maps x to x.(p.(k)) + origin in component k.val · ymir
- Tolk.Postrange.Scheduler.axes_ofaxes_of ~reduce k types is the positions of k's axes of a role in types, or of any role if types is [], that are reduce axes if reduce is true and are not if it is false.val · tolk
- Hugin.axisaxis ~side ~grid ~show ~title name is the axis of the position or facet scale name in each panel of the figures it is layered with, in place of the one the coordinate system generates (Guides).val · hugin
- Rig_amd_abi.Code_object.axisThe type for the axes of a grid.type · rig
- Rig_nv_abi.Qmd.axisThe type for the axes of a launch's grid and blocks.type · rig
- Rune.axisThe type for names of maps.type · rune
- Rune.axisaxis () is a fresh name, distinct from every other.val · rune
- Tolk.Opt.axisaxis o is the axis o acts on.val · tolk
- Hugin.axis'axis' ~side ~grid ~show ~title name is axis with the rich text title: axis ~title:s name is axis' ~title:(Text.v s) name, and axis' ~title:Text.(concat [ v "x"; sup (v "2") ]) "x" titles x with x².val · hugin
- Tolk.Tc.axis_coordsaxis_coords tc is every bit of the tile, the N bits, then the M bits, then the K bits, each from bit 0.val · tolk
- Tolk.Ops.axis_idaxis_id r is the identity of the range r.val · tolk
- Tolk.Ops.axis_typeaxis_type r is the role of the range r.val · tolk
- Tolk.Ops.Axis_typeAxis types.module · tolk
- Tolk.Postrange.Scheduler.axis_typesaxis_types k is the role of each of k's axes.val · tolk
- Nx.Ptree.axpyaxpy s k x y is x with each array a replaced by k' * a + b, b y's array at its path and k' k cast to a's format, the product and the sum each rounded to it, a complex a part by part; and by b where a is not in the vector.val · nx
b
- Hugin_gg.Color.bb c is the encoded blue component of c.val · hugin
- Tolk.Ops.backedgeIn a kernel, backedge u ~loop ~cond runs u, then repeats loop while cond holds.val · tolk
- Tolk.Ops.Upat.backedgeval · tolk
- Tolk.Ops.backward_slicebackward_slice ~calls u is the nodes u reaches, without u, in toposort ~calls's order.val · tolk
- Tolk.Ops.backward_slice_with_selfbackward_slice_with_self ~calls u is u followed by backward_slice ~calls u.val · tolk
- Hugin_kit.Scale.bandband () is a categorical scale (Categorical scales) with:val · hugin
- Jera.Linear.bandedbanded ~width is for an a whose matrix has entries only within width of its diagonal, such as a finite difference in one dimension.val · jera
- Hugin_kit.Scale.bandwidthbandwidth s is the width of the bands of s in normalised units, (1 - p) / (n + p) for n categories and padding p, and 0.val · hugin
- Rig_nv_abi.Cubin.bankThe type for the constant banks a kernel reads.type · rig
- Rig_nv_abi.Launch.banksbanks l is the constant banks a launch addresses: the kernel's (Cubin.kernel), with a bank 0 of 352 bytes at offset 0 first if it has none.val · rig
- Rig_pci.Function.barbar f i is the bus address of f's BAR i, as its BAR register holds it and as another function on the bus reaches it, with its size in bytes.val · rig
- Rig_disk.barrierbarrier b returns once the bytes written to b's file before it, by copies or through a borrow, are ordered before every later change to the file system (Copies).val · rig
- Rig_pci.Window.barrierbarrier () orders every access before it before every access after it, as devices and other processors see them.val · rig
- Tolk.Ops.barrierbarrier u rest is the Op.t.Barrier of u and rest.val · tolk
- Ymir.Earth.barycentricbarycentric t is (p, v) as heliocentric but from the barycentre of the solar system.val · ymir
- Rig_pci.Page_table.basebase t is the first virtual address the tables translate.val · rig
- Rig_pci.Space.basebase s is s's first address.val · rig
- Tolk.Ops.basebase u is u without its movements.val · tolk
- Tolk.Tc.base_upcast_axesbase_upcast_axes tc is the bits of C's elements, then the K bits, most significant first.val · tolk
- Hugin_kit.Curve.basisbasis is the uniform cubic B-spline whose control points are the points, the first and the last counted three times: it starts at the first point and ends at the last, has continuous curvature, and in between passes near the points rather than through them.val · hugin
- Kaun.Cache_index.batchbatch index is the number of lanes.val · kaun
- Tolk.Renderer.Compiler.batchbatch ~cachekey ~disassemble compile is v's compiler, whose compile is given several sources at once and returns their binaries in their order.val · tolk
- Batch ProcessingReal-world usage requires encoding multiple texts into uniform-length sequences for model input.guide · brot
- Kaun.Batch_normBatch normalization, with running statistics as explicit state.module · kaun
- Kaun.Data.batchesbatches ~batch_size x is the sequence of minibatches of x: consecutive slices of batch_size examples along axis 0, in order.val · kaun
- Talon.batchesbatches t is t's batches, each a table of one batch, in order.val · talon
- Kaun.Data.batches2batches2 ~batch_size (x, y) is like batches for a dataset of paired examples — inputs x and targets y with equal axis-0 sizes.val · kaun
- Tolk.Setting.beambeam is the width of the beam search that picks kernel optimizations, from BEAM.val · tolk
- Tolk.Setting.beam_debugbeam_debug is the verbosity of the beam search's diagnostics, from BEAM_DEBUG: from 1, the kernel searched, the candidates whose timing failed and the result; from 2, every candidate timed.val · tolk
- Tolk.Setting.beam_estimatebeam_estimate lets the beam search time a program that launches more than 65536 workgroups launching fewer, and scale its time up, from BEAM_ESTIMATE.val · tolk
- Tolk.Setting.beam_local_maxbeam_local_max is the most warp and local threads a candidate of the beam search may have, from BEAM_LOCAL_MAX.val · tolk
- Tolk.Setting.beam_log_surpass_maxbeam_log_surpass_max prints the candidates the beam search drops for exceeding beam_upcast_max, beam_local_max or beam_uops_max, from BEAM_LOG_SURPASS_MAX.val · tolk
- Tolk.Setting.beam_min_progressbeam_min_progress is the microseconds a round of the beam search must gain for the search to go on, from BEAM_MIN_PROGRESS.val · tolk
- Tolk.Setting.beam_padtobeam_padto adds pads of axes to the candidates of the beam search (Search.actions), from BEAM_PADTO.val · tolk
- Tolk.Search.beam_searchbeam_search ~link ~time ~allow_test_size amt k is k, or a copy of it, with the optimisations that a beam search of width amt finds fastest.val · tolk
- Tolk.Setting.beam_strict_modebeam_strict_mode makes the beam search raise what the compilation of a candidate raises, but Failure, instead of dropping the candidate, from BEAM_STRICT_MODE.val · tolk
- Tolk.Setting.beam_uops_maxbeam_uops_max is the number of instructions from which the beam search drops a candidate, none if it is 0 or less, from BEAM_UOPS_MAX.val · tolk
- Tolk.Setting.beam_upcast_maxbeam_upcast_max is the most upcast and unrolled lanes a candidate of the beam search may have (Search.get_kernel_actions), from BEAM_UPCAST_MAX.val · tolk
- Ymir_units.Unit.becquerelbecquerel is s-1, which is hertz.val · ymir
- Brot.Pre_tokenizer.behaviorDelimiter handling behavior for splitting operations.type · brot
- Norn.Dist.bernoullibernoulli ~logits is true with probability sigmoid logits.val · norn
- Nx.Rng.bernoullibernoulli p is true with probability p rounded up to a multiple of 2{^-53}, elementwise: 53 random bits, read as an integer, below p 2{^53}.val · nx
- Brot.Normalizer.bertbert () is a BERT normalizer.val · brot
- Brot.Post_processor.bertbert ~sep ~cls is a BERT-style post-processor.val · brot
- Brot.Pre_tokenizer.bertbert applies BERT-style pre-tokenization.val · brot
- Nx.besidebeside x is Placement.uncut (placement x), whole on each device x lies on, and no sum.val · nx
- Jera.Solution.bestbest s is the estimate in every lane.val · jera
- Munin.Metric.bestbest goal st is the finite sample that best meets goal: the least of st's finite samples under `Minimize and the greatest under `Maximize, the earliest among equals.val · munin
- Norn.Dist.betabeta ~a ~b has density proportional to x^(a - 1) (1 - x)^(b - 1) on (0, 1).val · norn
- Nx.Rng.betabeta a b is beta draws on (0, 1) of concentrations a and b, elementwise over their broadcast shape: G(a) / (G(a) + G(b)) for two independent gamma draws, formed from their logarithms so that two draws that underflow keep their ratio.val · nx
- Nx.betaincbetainc a b x is the regularised incomplete beta function I_x(a, b) = (1/B(a, b)) ∫₀ˣ t{^a-1} (1 - t){^b-1} dt, the distribution function at x of the beta distribution of shapes a and b, for a, b ≥ 0 and x in [0., 1.].val · nx
- Nx.betainccbetaincc a b x is 1 - I_x(a, b) (betainc), the upper tail, to its own relative precision where I rounds to 1.: betaincc 1.val · nx
- Nx.betainccinvbetainccinv a b q is the x with betaincc a b x = q, as betaincinv from the upper tail: 1.val · nx
- Nx.betaincinvbetaincinv a b p is the x with betainc a b x = p, the beta distribution's quantile, for a, b ≥ 0 and p in [0., 1.].val · nx
- Jera.Minimize.bfgsbfgs ~linear is BFGS, for a smooth objective of up to some thousands of unknowns.val · jera
- Nx.bfloat16val · nx
- Nx.bfloat16_elttype · nx
- Nx_array.Dtype.bfloat16_eltbfloat16: binary32's sign, 8 exponent bits and the top 7 fraction bits.type · nx
- Nx.bfloat16_ttype · nx
- Nx_array.bigarraybigarray a is a bigarray of a's shape over a's own bytes.val · nx
- Tolk.BigintArbitrary-precision integers.module · tolk
- Norn.BijBijectors: maps from unconstrained coordinates onto a support.module · norn
- Nx_kernel.Prog.binaryThe type for kinds of two operands of one dtype, of that dtype.type · nx
- Nx_kernel.Prog.E.binarybinary k x y is Op2 (Binary k, x, y).val · nx
- Talon.BinaryByte strings.module · talon
- Talon.Col.binarybinary name is v Kind.binary name.val · talon
- Talon.Kind.binarybinary reads binary columns.val · talon
- Talon.Type.binarybinary is Binary.val · talon
- Tolk.Op.Set.binarybinary is the operations of two operands: Add, Mul, Truncdiv, Max, Truncmod, Cmplt, Cmpne, Cmpeq, Xor, Shl, Shr, Or, And, Threefry, Sub, Fdiv, Pow, Div and Mod.val · tolk
- Hugin_float_text.Float_text.binary32binary32 is { precision = 24; emin = -126 }.val · hugin
- Munin_float_text.Float_text.binary32binary32 is { precision = 24; emin = -126 }.val · munin
- Norn_float_text.Float_text.binary32binary32 is { precision = 24; emin = -126 }.val · norn
- Hugin_float_text.Float_text.binary64binary64 is { precision = 53; emin = -1022 }.val · hugin
- Munin_float_text.Float_text.binary64binary64 is { precision = 53; emin = -1022 }.val · munin
- Norn_float_text.Float_text.binary64binary64 is { precision = 53; emin = -1022 }.val · norn
- Hugin.bindbind k f is the figure f v, where v is the value of k in the view resolve is given (View.get).val · hugin
- Hugin.Mark.bindbind ~imply ~guide ~stack r c binds the channel c to the role r, where:val · hugin
- Tolk.Shape.bindbind v x is v bound to x.val · tolk
- Hugin.Mark.bindingThe type for channels bound to roles.type · hugin
- Tolk.Setting.bindingThe type for settings bound to values.type · tolk
- Norn.Dist.binomialbinomial ~trials ~logits is the number of successes in trials independent trials, each a success with probability p = sigmoid logits: the count x in {0, ..., trials} with probability C(trials, x) p^x (1 - p)^(trials - x).val · norn
- Nx.Rng.binomialbinomial n p is the number of successes in n trials that each succeed with probability p, elementwise over their broadcast shape: 0 at p = 0, n at p = 1.val · nx
- Hugin_kit.Stats.binsThe type for histograms on the device set 'd.type · hugin
- Nx.bitval · nx
- Tolk.Tc.bitThe type for the bits of tile coordinates: M i is bit i of the row, N i bit i of the column, K i bit i of the reduction index.type · tolk
- Nx.bit_elttype · nx
- Nx_array.Dtype.bit_eltBooleans, one to a bit.type · nx
- Nx.bit_ttype · nx
- Nx.bitcastbitcast dt x is x's bits read as elements of dt, NaN payloads included: x itself where dt is x's dtype.val · nx
- Nx_array.bitcastbitcast dt a is a's bits read as elements of dt, over a's buffer.val · nx
- Nx_kernel.Prog.E.bitcastval · nx
- Tolk.Dtype.bitcastbitcast d0 d1 v is the d1 value stored by the bits that store v as a d0 value.val · tolk
- Tolk.Ops.bitcastbitcast x dt reinterprets x's bits as dt, or is x if it is of type dt.val · tolk
- Tolk.Ops.Upat.bitcastbitcast ~dtype p matches a bitcast of p, to dtype if given.val · tolk
- Nx.Rng.bitsbits shape is uniformly random 32-bit words: word j in C order is word j mod 2 of the generator's block j / 2.val · nx
- Nx_array.Dtype.bitsbits dt is the width of one element of dt in bits: 1 for Bit, 4 for the 4-bit formats, 8 for Bool, up to 128 for Complex128.val · nx
- Nx_kernel.Prog.bitsbits dt x is the bits of x, an element of dt, in the host's byte order: Nx_array.Dtype.bytes dt 1 bytes, a sub-byte element in the low bits of one byte.val · nx
- Nx.Rng.bits_atbits_at p shape is bits shape at p.val · nx
- Tolk.Dtype.bitsizebitsize dt is the number of bits of a dt value: 0 for Void, 1 for Bool, and 800 for the weak data types, wider than any value a program stores.val · tolk
- Nx.bitwise_andbitwise_and a b is the bitwise and of integers, the logical and of booleans.val · nx
- Tolk.Ops.bitwise_andval · tolk
- Tolk.Ops.Elementwise.bitwise_andval · tolk
- Tolk.Ops.Upat.bitwise_andval · tolk
- Nx.bitwise_notbitwise_not x is every bit of x flipped, the logical not of booleans.val · nx
- Tolk.Ops.bitwise_notbitwise_not x is logical_not x for booleans, and x xor all ones otherwise.val · tolk
- Tolk.Ops.Elementwise.bitwise_notbitwise_not x is logical_not x for booleans, and x xor all ones otherwise.val · tolk
- Tolk.Ops.Upat.bitwise_notbitwise_not x is logical_not x for booleans, and x xor all ones otherwise.val · tolk
- Nx.bitwise_orbitwise_or a b is the bitwise or, as bitwise_and.val · nx
- Tolk.Ops.bitwise_orval · tolk
- Tolk.Ops.Elementwise.bitwise_orval · tolk
- Tolk.Ops.Upat.bitwise_orval · tolk
- Nx.bitwise_xorbitwise_xor a b is the bitwise exclusive or, as bitwise_and.val · nx
- Tolk.Ops.bitwise_xorval · tolk
- Tolk.Ops.Elementwise.bitwise_xorval · tolk
- Tolk.Ops.Upat.bitwise_xorval · tolk
- Hugin_gg.Color.blackblack is v 0.val · hugin
- Rig.Buffer.blit_from_stringblit_from_string s i b j n copies the n bytes of s from i into b from its byte j.val · rig
- Rig_pci.Window.blit_stringblit_string s soff w off n copies the n bytes of s from byte soff to byte off of w.val · rig
- Rig.Buffer.blit_to_bytesblit_to_bytes b i s j n copies the n bytes of b from its byte i into s from j.val · rig
- Quill_markdown.Edit.blockA top-level block extracted from a markdown source string.type · quill
- Rig.Submission.blockThe type for a launch's or a fill's block in a run of one submission (Run).type · rig
- Rig.Submission.blockblock s i is the block of s's part i in a run (Run).val · rig
- Rig_edge_primitives.blockblock () runs what a poll point runs (the handlers of the signals pending, finalisers, Gc.Memprof callbacks, a switch to another thread), then blocks again what unblock gave back (rig_block).val · rig
- Quill_markdown.Edit.block_kindtype · quill
- Quill_markdown.Edit.block_sourceblock_source t block extracts the raw source substring for block.val · quill
- Quill_markdown.Edit.block_to_htmlblock_to_html t block renders block to an HTML fragment.val · quill
- Quill_markdown.Edit.blocksblocks t is the list of top-level blocks in document order.val · quill
- Hugin_gg.Color.blueblue is v 0.val · hugin
- Hugin_kit.Scheme.bluesblues is Brewer's Blues.val · hugin
- Tolk.Ops.bodybody c is the body the call c calls.val · tolk
- Ymir_units.Codata.bohr_magnetonbohr_magneton r is μB in J T-1.val · ymir
- Ymir_units.Codata.bohr_radiusbohr_radius r is a0 in m.val · ymir
- Hugin_font.Font.boldbold is Inter Bold, weight 700.val · hugin
- Hugin_text.Text.boldbold t is t with every character bold.val · hugin
- Ymir_units.Unit.boltzmannboltzmann is k, 1.380649·10-23 kg m2 s-2 K-1.val · ymir
- Nx.boolval · nx
- Talon.Col.boolbool name is v Kind.bool name.val · talon
- Talon.Expr.boolbool b is the boolean b.val · talon
- Talon.Kind.boolbool reads bool columns.val · talon
- Talon.Type.boolbool is Bool.val · talon
- Tolk.Ops.boolbool ~dtype b is const ~dtype (`Bool b).val · tolk
- Tolk.Ops.Elementwise.O.boolbool b is the literal b.val · tolk
- Tolk.Ops.O.boolbool b is the literal b.val · tolk
- Tolk.Ops.Upat.boolbool b is the literal b.val · tolk
- Tolk.Ops.Upat.O.boolbool b is the literal b.val · tolk
- Tolk.Setting.boolbool ~reach key default is a setting whose initial value is true iff the variable key holds a nonzero integer, and default if key is unset.val · tolk
- Ymir_fits.Fits.Value.boolbool is the logical T or F.val · ymir
- Nx.bool_elttype · nx
- Nx_array.Dtype.bool_eltBooleans, one to a byte: true iff the byte is not zero.type · nx
- Nx.bool_ttype · nx
- Rig_pci.Page_table.bootedbooted t ends booting.val · rig
- Nx_array.borrowborrow d a is Some a', a's elements on d without a copy: a's dtype and layout over a borrow of its buffer (Rig.Buffer.borrow), which shares a's memory, so writes through either show in the other.val · nx
- Rig.Buffer.borrowborrow d b is a buffer on d over b's bytes, without a copy, or None if d cannot map b's memory (Rig.reaches) or its driver never maps it, such as pages mapped read-only.val · rig
- Brot.bos_tokenbos_token t is t's beginning-of-sequence token, if any.val · brot
- Tolk.Ops.Bottom_up_gateRaised by a rule before the sources (Before_sources, Around_sources's before) under Fixed_point to keep the node it last produced and leave its sources unvisited.exception · tolk
- Tolk.Ops.bound_loadsbound_loads r is true iff the range r has a bound that, its ordering after effects (Op.t.After) aside, reads memory, as j < x[i]: how many trips the loop runs is what memory holds.val · tolk
- Fehu.Space.boundary_valuesboundary_values s is a list of representative edge-case values for s.val · fehu
- Fehu.Space.Box.boundsbounds s is (low, high) copies of the bound vectors.val · fehu
- Hugin_font.Font.boundsbounds f is the box enclosing every glyph of f as recorded in its head table.val · hugin
- Hugin_font.Run.boundsbounds r is the smallest box containing the ink boxes (Font.ink) of the glyphs of r, each scaled by size r and moved to its glyph's origin, or None if no glyph has ink.val · hugin
- Hugin_gg.Path.boundsbounds p is the smallest box containing what fold visits in p, each curve bounded exactly rather than by its control points, or None if fold visits nothing.val · hugin
- Hugin_gg_kit.Pgon2.boundsbounds p is the smallest box containing the points of the rings of p, or None if they have no point.val · hugin
- Hugin_gg_kit.Ring2.boundsbounds r is the smallest box containing the points of r, or None if r has no point.val · hugin
- Hugin_vg.Picture.boundsbounds p is a box containing every point p paints, or None if p has no extent: no leaf of p has a path that visits a point, a glyph with ink or an image.val · hugin
- Hugin_vg_instances.Instances.boundsbounds p is Picture.bounds p, or everywhere if the bounds lie beyond the range of floats, so that a picture that far is taken to show everywhere.val · hugin
- Norn.Dist.boundsbounds d is the least and greatest value each element of d's values may take, of d's shape: -inf and inf where unbounded, the support's interval for an element of a vector.val · norn
- Fehu.Space.Boxmodule · fehu
- Hugin_text.Text.Layout.boxbox l is the smallest box holding each line's box, which spans the line's width and its ascent above and descent below its baseline.val · hugin
- Hugin_vg_vector.Vector.boxbox ctx p is a box of the frame holding what p paints there, grown by one point and cut by ctx.cut, or None if p paints nothing there.val · hugin
- Jera.Quad.BoxBoxes of integration, one per lane.module · jera
- Rig.Buffer.boxThe type for boxes: rows rows of width bytes.type · rig
- Rig_nv_abi.Method.boxbox ~dst ~src ~dst_pitch ~src_pitch ~rows n copies rows rows of n bytes on the copy engine, row i from src + i * src_pitch to dst + i * dst_pitch, in one multi-line launch.val · rig
- Hugin_gg.Box2Axis-aligned boxes.module · hugin
- Sowilo.box_blurbox_blur ~ksize img applies a ksize * ksize averaging filter.val · sowilo
- Brot.bpebpe () is a BPE (Byte Pair Encoding) tokenizer.val · brot
- Brot.Decoder.bpebpe ~suffix () is a decoder for BPE-encoded tokens.val · brot
- Jera.Minimize.bracketbracket ~tol f ~lo ~hi is a minimum of f in each [lo, hi], elementwise: every element is its own problem, the ends broadcast together and come in either order, and f must compute each element of its result from the same element of its argument alone.val · jera
- Jera.Root.bracketbracket ~tol f ~lo ~hi is a zero of f between lo and hi in each element, given in either order, which broadcast together.val · jera
- Hugin_kit.Scheme.brbgbrbg is Brewer's BrBG, brown to blue-green.val · hugin
- Jera.Piecewise.breaksbreaks p is p's breaks, of shape [pieces + 1].val · jera
- Hugin.Mark.broadcastbroadcast ~shape bindings is the shape of the mark that v ~shape bindings makes: shape, by default [||], and the shapes of the channels broadcast together (Data and rows).val · hugin
- Tolk.Ops.broadcastbroadcast u n is the Op.t.Stack of n copies of u, or u if n is 1.val · tolk
- Tolk.Ops.Upat.broadcastbroadcast p matches a stack whose sources each match p.val · tolk
- Nx.broadcast_arraysbroadcast_arrays xs is each of xs stretched to broadcast_shapes of their shapes, in order.val · nx
- Tolk.Shape.broadcast_axesbroadcast_axes src out is the axes of out that broadcasting src to out adds or expands.val · tolk
- Tolk.Ops.broadcast_shapebroadcast_shape shapes is the shape shapes broadcast to: aligned to the right, each axis is the size of its sources that is not 1, or 1 if all are.val · tolk
- Nx.broadcast_shapesbroadcast_shapes ss is the shape that ss broadcast to, [||] for [].val · nx
- Nx.broadcast_tobroadcast_to s x is x stretched to shape s: an axis of extent 1, or one x lacks, repeats x's elements along it.val · nx
- Tolk.Op.Set.broadcastablebroadcastable is binary and ternary: the operations whose operands are brought to a common shape and type.val · tolk
- brotBrot tokenizes text into token IDs for language models and reverses the process.library · brot
- BrotTokenization for OCaml.module · brot
- Brot vs. HuggingFace Tokenizers -- a practical comparisonThis guide explains how Brot relates to Python's HuggingFace Tokenizers, focusing on:guide · brot
- Jera.Sde.BrownianBrownian paths with their space–time Lévy area.module · jera
- Jera.System.broydenbroyden is Broyden's method, for an f with no usable derivative.val · jera
- Jera.Ode.bs3bs3 is Bogacki and Shampine's (1989) method of order 3 with an embedded order 2: three evaluations per step, its last stage the next step's first.val · jera
- Rig.budgetbudget d is d's budget (Reclamation).val · rig
- Rig_edge.Io.budgetbudget d is the bytes d should hold at most.val · rig
- Tolk.Heuristic_host.budgetedbudgeted k is true iff k renders for the CPU device of a renderer that knows its vector registers.val · tolk
- Tolk.Ops.buf_uopbuf_uop u is the storage node u accesses.val · tolk
- Fehu.BufferReplay buffer for off-policy experience storage.module · fehu
- Nx_array.bufferbuffer a is a's buffer.val · nx
- Rig.BufferBuffers of device memory.module · rig
- Rig_mlx5_abi.bufferThe type for memory a process and a NIC share, such as a ring.type · rig
- Tolk.Ops.bufferizebufferize ~opts u ranges is the Op.t.Stage of u over ranges, into a buffer placed by opts if given.val · tolk
- Tolk.Ops.bufferize_optsThe type for the arguments of an Op.t.Stage that makes a buffer.type · tolk
- Tolk.Postrange.Scheduler.bufsbufs k is k's indexed accesses (Op.t.Index), in the reverse of Ops.backward_slice's order.val · tolk
- Hugin_kit.Scheme.bugnbugn is Brewer's BuGn, blue to green.val · hugin
- Quill_book.BuildProject build pipeline.module · quill
- Quill_book.Build.buildbuild ~create_kernel project builds the project.val · quill
- Quill_book.Build.build_filebuild_file ~create_kernel path builds a single notebook file to a self-contained HTML page.val · quill
- Hugin_kit.Scheme.bupubupu is Brewer's BuPu, blue to purple.val · hugin
- Rig.busbus d is the bus address of d's PCI function on its machine, such as "0000:03:00.0", as its driver states it (Rig_edge.facts), or None for a device that has none: host, a memory device, an io device, a Metal device and a device of another machine.val · rig
- Rig_bnxt.busbus nic is the bus address of nic's PCI function, such as "0000:41:00.0".val · rig
- Rig_mlx5.busbus nic is the bus address of nic's PCI function, such as "0000:17:00.0".val · rig
- Rig_pci.Function.busbus f is f's bus address on its machine.val · rig
- Rig_pci.Memory.busbus m r ~client is where the function at the bus address client of m's machine reaches r's memory by DMA: its bus addresses, as (address, bytes) pairs in order from r's first byte, each a multiple of 4 KiB.val · rig
- Rig_amd_amdgpu.busesbuses () is the bus addresses of the machine's AMD GPUs (Rig_amd.is_gpu), in bus order: by domain, bus, device and function, each as a number.val · rig
- Rig_amd_pci.busesbuses ~machine () is the bus addresses of the AMD GPUs of machine (defaults to Rig_pci.Machine.this) (Rig_amd.is_gpu), in bus order (Rig_pci.Machine.functions): GPU i is the ith, whichever driver holds it.val · rig
- Rig_bnxt.busesbuses ~machine () is the bus addresses of the NICs of machine (defaults to Rig_pci.Machine.this), in bus order.val · rig
- Rig_cuda.busesbuses () is the PCI bus addresses of the GPUs CUDA sees, in gpu's order, as Linux writes them under /sys/bus/pci, such as "0000:01:00.0", or count's Error.val · rig
- Rig_nv_nvidia.busesbuses () is the bus addresses of this machine's NVIDIA GPUs (Rig_nv.is_gpu), in bus order: by domain, bus, device and function, each as a number.val · rig
- Rig_nv_pci.busesbuses ~machine () is the bus addresses of the NVIDIA GPUs of machine (defaults to Rig_pci.Machine.this), in bus order (Rig_pci.Machine.functions): GPU i is the ith.val · rig
- Rig_pci.Driver.busesbuses d m is the bus addresses of d's functions on m, in bus order.val · rig
- Brot.Decoder.byte_fallbackbyte_fallback is a decoder for byte fallback tokens.val · brot
- Brot.Decoder.byte_levelbyte_level is a collapsing decoder that reads GPT-2 style byte-to-Unicode tokens back as the text their bytes spell.val · brot
- Brot.Normalizer.byte_levelbyte_level is GPT-2 style byte-level encoding: each byte of the text is mapped to a printable Unicode character using the GPT-2 byte-to-unicode table.val · brot
- Brot.Post_processor.byte_levelbyte_level () is a byte-level post-processor that adjusts character offsets for byte-level encoding.val · brot
- Brot.Pre_tokenizer.byte_levelbyte_level () is a byte-level pre-tokenizer.val · brot
- Brot.Pre_tokenizer.byte_level_decodebyte_level_decode s is the bytes the byte-level alphabet spells s from: one byte per character.val · brot
- Hugin_font.Font.bytesbytes f is an OpenType file holding exactly the face of f.val · hugin
- Nx_array.Dtype.bytesbytes dt n is the number of bytes n elements of dt fill, ⌈n · bits dt / 8⌉.val · nx
- Rig_remote_proxy.Link.bytesbytes l ~device ~value b sends b's bytes as those of the next copy into the controller's memory of the work of device at value.val · rig
c
- Tolk.CShared libraries of the system.module · tolk
- Nx_cpu.c2cc2c d ~axes ~dst x stores the transform d of x into dst.val · nx
- Nx_kernel.S.c2cc2c d ~axes ~dst x stores the transform d of x into dst.val · nx
- Nx_kernel.Shape.c2cc2c ~axes x is x, where axes are strictly increasing axes of x, not empty.val · nx
- Nx_cpu.c2rc2r ~axes ~dst x reads n/2 + 1 bins X along the last of axes, n dst's extent there.val · nx
- Nx_kernel.S.c2rc2r ~axes ~dst x reads n/2 + 1 bins X along the last of axes, n dst's extent there.val · nx
- Nx_kernel.Shape.c2rc2r ~axes ~n x is x but n points along the last of axes, with axes as c2c takes them, where x has n/2 + 1 bins there and n is not negative.val · nx
- Kaun.Attention.CacheKey-value caches.module · kaun
- Rig.cache_bytescache_bytes d is Some n with n the bytes of d's last-level cache, the cache farthest from its compute units through which its work's memory accesses pass, as d's driver states it at open (Rig_edge.facts), and None where its driver does not know it.val · rig
- Tolk.Helpers.cache_dircache_dir is the directory of the files tolk caches: tolk in XDG_CACHE_HOME if that variable is set, and otherwise in ~/Library/Caches on macOS and ~/.cache elsewhere.val · tolk
- Kaun.Cache_indexWhere a call's tokens sit in a cache: each token's position and the slots of each sequence.module · kaun
- Kaun_hf.cache_pathcache_path ~file repo_id is the local path where download_file caches file from repo_id: cache_dir/repo/revision/file, where repo is repo_id with "/" replaced by "-".val · kaun
- Kaun.Attention.cachedcached ~head_dim p cache index x is causal self-attention of the tokens x, of shape [| batch; seq; embed_dim |], over their sequences, and cache with their keys and values written.val · kaun
- Tolk.Graph.cachedcached ~table ~key ~valid make is the graph that the Helpers.Diskcache holds for key in table, and true, if its entry reads as a graph that valid accepts; otherwise it is make (), which replaces the entry, and false.val · tolk
- Tolk.Helpers.cachedbcachedb is the directory of the Diskcache, from the variable CACHEDB.val · tolk
- Tolk.Renderer.Compiler.cachekeycachekey c is the name of c's toolchain and options, if c has one: the table of its binaries, whether or not they are kept.val · tolk
- Tolk.Setting.cachelevelcachelevel enables the Helpers.Diskcache when positive, from CACHELEVEL.val · tolk
- Tolk.Setting.cachesizecachesize is the most bytes the files of the Helpers.Diskcache hold, from CACHESIZE; 0 sets no bound.val · tolk
- Ymir_units.Unit.caesium_frequencycaesium_frequency is ΔνCs, the caesium 133 hyperfine transition frequency, 9192631770 s-1.val · ymir
- Tolk.Compiler_worker.callcall w library request has a worker of w serving library answer request: one that is idle, or else one it starts, as PROGRAM LIBRARY TEMP PARENT, PARENT this process's pid.val · tolk
- Tolk.Ops.callcall ~inline body args calls body on args; a call of code in the target's language is written into its reader if inline (default false).val · tolk
- Tolk.Ops.call_infoThe type for the arguments of Op.t.Call.type · tolk
- Tolk.Realize.call_valuescall_values call u is u, a launch size, a split's iterations or an estimate of call's program, with each scalar parameter replaced by what call gives it: a nameless parameter, call's argument in its slot; a variable, the variable call binds to its name, or else the variable, at slot -1, that a run binds by name.val · tolk
- Tolk.Shape.call_with_outputcall_with_output value args is the one output of call_with_outputs [value] args.val · tolk
- Tolk.Shape.call_with_outputscall_with_outputs values args calls a body that computes values from args, and is the outputs, each ordered after the call.val · tolk
- Rune.callbackcallback ~lanes s ~like f x is f x, run on x's values once per execution of the call (Executions).val · rune
- CallbacksA callback runs OCaml on the values of a traced or transformed function (the Callbacks section of rune.mli).guide · rune
- Tolk.Ops.callsThe type for how a walk of a graph treats a call (Op.t.Call), whose first source is its body.type · tolk
- Tolk.Dtype.can_lossless_castcan_lossless_cast d0 d1 is true iff a cast from d0 to d1 is known to keep every value, that is iff d0 is d1 or Bool, or:val · tolk
- Quill.Session.can_redocan_redo s is true if a redo operation is available.val · quill
- Quill.Session.can_undocan_undo s is true if an undo operation is available.val · quill
- Ymir_units.Unit.candelacandela is the symbol cd.val · ymir
- Sowilo.cannycanny ~low ~high img applies the Canny edge detector.val · sowilo
- Hugin_gg.Stroke.capThe type for line caps.type · hugin
- Hugin_gg.Stroke.capcap s is the line cap of s.val · hugin
- Hugin_font.Font.cap_heightcap_height f is the height of flat capital letters above the baseline: the OS/2 value if f records a positive one, else the top of the ink of f's glyph for H if it has ink, else ascent, as CSS assumes.val · hugin
- Rig.capabilitycapability d k is the record that d's driver declares under k (Rig_edge.capability), filled when d opened, or None if it declares none under k.val · rig
- Rig_amd.capabilitycapability g is the record g's facts carry as capability (facts).val · rig
- Rig_amd_abi.CapabilityWhat compiled code needs from an AMD device.module · rig
- Rig_cuda.capabilitycapability g is g's record for compiled code, which g's facts declare (Rig_cuda_abi.declare): the functions of the CUDA library g was opened with, and the maker of graphs in g's context.val · rig
- Rig_edge.capabilityThe type for what compiled code needs from a device, a record that the driver's ABI library declares under its key.type · rig
- Rig_metal.capabilitycapability d is d's record for compiled code, which d's facts declare (Rig_metal_abi.declare).val · rig
- Rig_nv.capabilitycapability g is g's GPU as compiled code sees it, which g's facts declare (Rig_nv_abi.declare).val · rig
- Rig_remote_proxy.capabilitycapability d is the record device was given, which its facts declare (Rig_remote_abi.declare).val · rig
- Fehu.Buffer.capacitycapacity buf is the maximum number of transitions buf can hold.val · fehu
- Tolk.Ops.Nodes.cardinalcardinal s is the number of nodes of s.val · tolk
- Nx_kernel.carriercarrier dt is the dtype kernels reduce and scan terms of dt in, and the dtype of a scan's carry, the state a scan continues from (S.scan): float32 for a float narrower than it, int8 and uint8 for int4 and uint4, bool for bit, else dt.val · nx
- Nx_kernel.Shape.carrycarry ~axis y is the shape of a carry of a scan along axis of a loop of shape y: y with extent 1 along axis, with axis as scan takes it.val · nx
- Hugin.Coord.cartesiancartesian ~aspect () maps normalised positions affinely onto the box of the panel, x from its left edge to its right and y from its bottom edge to its top, and draws x axes below the panel and y axes to its left.val · hugin
- Fehu_envs.CartpoleClassic cart-pole balancing (CartPole-v1).module · fehu
- Nx.Ptree.Walk.casecase c tag reports the variant case tag at c's path.val · nx
- Nx.castcast dt x is x's elements stored in dt; x itself where dt is x's dtype.val · nx
- Nx.Ptree.castcast (module U) dt x is x with each array at a position of U's parameter cast to dt, as Nx.cast.val · nx
- Nx_kernel.Prog.E.castval · nx
- Talon.Expr.castcast ty a converts a's values to ty:val · talon
- Tolk.Ops.castcast x dt is x converted to dt, or x if it is of type dt.val · tolk
- Tolk.Ops.Elementwise.castcast x dt is x converted to dt, or x if it is of type dt.val · tolk
- Tolk.Ops.Upat.castcast x dt is x converted to dt, or x if it is of type dt.val · tolk
- Hugin.catcat ~scale ~valid ~title ~labels codes is the categories of the integer codes, one datum per element, with scale, valid and title as for num, and:val · hugin
- Tolk.Shape.catcat ~axis u rest concatenates u :: rest along axis (default 0).val · tolk
- Catalogue Cut and Cross-MatchA sky survey catalogue holds one row per detected source.guide · talon
- Norn.Dist.categoricalcategorical ~logits is the index k with probability softmax logits along the last axis.val · norn
- Nx.Rng.categoricalcategorical logits is a position along axis of each row of logits, drawn with probability proportional to the exponential of its logit: axis is the last by default, a negative one counted from the end.val · nx
- Talon.Type.categoricalcategorical d is Categorical with a copy of the dictionary d, which may be empty.val · talon
- Hugin_kit.Scale.categoriesThe type for the categories of band scales.type · hugin
- Talon.Kit.categorizecategorize cs q is q with each text column of cs cast, in place, to the categorical type whose dictionary is the column's distinct non-null values in byte order, the order of Order.asc on text, which does not depend on the order or the batches of q's rows.val · talon
- Hugin_kit.Curve.catmull_romcatmull_rom is the centripetal Catmull–Rom spline through the points: it passes through every point, and each segment is shaped by the points before and after it, the parameter advancing by the square root of the distance between consecutive points, which keeps segments free of cusps and self-intersections.val · hugin
- Norn.Dist.cauchycauchy ~loc ~scale has density 1 / (π scale (1 + z²)).val · norn
- Nx.causalcausal p holds where p.k <= p.q + (p.len - p.tq): the last query row sees every key before the row's length, and each row before it one key fewer.val · nx
- Rig_bin.Proc.causecause st is "exited with status N", or "killed by SIGNAME", SIGNAME a POSIX signal's name or "signal N".val · rig
- Tolk.Setting.cccc is the program Compiler_cpu.clang runs, from CC, read when the compiler is made.val · tolk
- Tolk.Setting.ccacheccache keeps compiled programs and binaries in the Helpers.Diskcache, from CCACHE.val · tolk
- Tolk.Ops.ccastccast u dt is u as a dt constant if u is a constant, and cast u dt otherwise.val · tolk
- Tolk.Ops.cconstcconst dt c is the cast to dt of the literal c, even where const would fold the cast away.val · tolk
- Tolk.Bigint.cdivcdiv a b is a / b rounded toward positive infinity.val · tolk
- Hugin_kit.Time.ceilceil ~tz_offset_s i t is the earliest boundary of i at tz_offset_s that is not before t.val · hugin
- Nx.ceilceil x rounds toward positive infinity.val · nx
- Nx_kernel.Prog.E.ceilval · nx
- Tolk.Helpers.ceildivceildiv num amt is num / amt rounded up.val · tolk
- Ymir.Transform.celestialcelestial code f ~pv ~native ~crval ~lonpole ~latpole is the projection code with parameters pv, [...; m], followed by the rotation taking the native point native (φ₀, θ₀) to crval (lon, lat) in f, with the celestial pole at native longitude lonpole; latpole picks between two poles where FITS defines two.val · ymir
- Quill.CellNotebook cells.module · quill
- Ymir.Grid.cellcell is the unscoped symbol counting cells.val · ymir
- Quill.Session.cell_statustype · quill
- Quill.Session.cell_statuscell_status id s is the execution status of cell id in s.val · quill
- Hugin.Mark.cellscells draws a panel's rows as one image whose pixels are their cells, when the projection is affine, x and y read band scales without padding, the mark binds no x2, y2 or stroke, and no two rows share a cell; a cell no row covers paints nothing.val · hugin
- Quill.Doc.cellscells d is the ordered list of cells in d.val · quill
- Sowilo.center_cropcenter_crop ~height ~width img crops a centered rectangle.val · sowilo
- Ymir_units.Unit.centicenti u is 10-2 u.val · ymir
- Ymir.Healpix.centrecentre h p is the centre of each pixel of p, unit vectors within 2e-15 radians of the exact centres, and a NaN row for -1.val · ymir
- Ymir.Grid.centrescentres g is each cell's centre in the world, batch @ shape g points.val · ymir
- Tolk.Linearizer.cfg_contextThe type for the order a kernel's loops run in.type · tolk
- Tolk.Linearizer.cfg_contextcfg_context sink is the order the loops of sink run in.val · tolk
- Jera.Linear.cgcg ~rel ~budget ~precondition is conjugate gradients from u = 0, preconditioned by precondition, for a symmetric positive-definite a such as a Hessian at a minimum or a damped Gauss–Newton matrix: an application of a and of precondition per iteration, and O(n) other work on three vectors.val · jera
- Tolk.Tc.chainThe type for how an accumulator becomes an operand.type · tolk
- Tolk.Tc.chainchain x y op is how x's accumulator is y's operand op with every element in the lane that holds it, if it can be.val · tolk
- Talon.Schema.changeThe type for the differences between two schemas s0 and s1.type · talon
- Hugin.channelThe type for channels for roles of range 'r: data whose values, of the domain type 'd, a scale normalises and the role maps into 'r, or a constant of 'r.type · hugin
- Fehu.Render.Pixel.channelschannels fmt is the number of channels for fmt.val · fehu
- Quill_book.Render.chapter_htmlchapter_html doc renders doc to an HTML content fragment.val · quill
- Brot.Pre_tokenizer.char_delimiterchar_delimiter c splits on the character c, removing it from the output.val · brot
- Brot.charschars () is a byte-level tokenizer: each byte of the input is one token whose ID is the byte's value.val · brot
- Jera.Grid.chebyshevchebyshev ~degree ~pieces f ~lo ~hi interpolates f on the box from lo to hi, both of shape [d], split into pieces equal pieces along each axis, at the tensor product of each piece's degree + 1 Chebyshev points of the second kind.val · jera
- Jera.Piecewise.chebyshevchebyshev s ~degree ~pieces f a b interpolates f on pieces equal pieces of [a, b] at each piece's degree + 1 Chebyshev points of the second kind, ends included (degree = 0 takes the midpoint).val · jera
- Munin.Store.checkcheck s is each path of s that breaks the format, with what is wrong, sorted by path:val · munin
- Norn.Dist.checkcheck ?unless what d is d whose parameters are checked now, and whose eliminators then check none: a caller that computes parameters names itself in the refusal.val · norn
- Nx.checkcheck s ok data fail raises fail i d if an element of ok is false, i the index of the first in C order and d the structure data with each array replaced by its element at i, a 0-d value whole on each device of that array's placement (scalar_like).val · nx
- Tolk.Uniform.checkcheck sink raises Invalid_argument, naming the kernel (or a called body, which has no name), the operation and the loop, if a barrier, a matrix multiply-accumulate, or a call whose body holds one, runs in a serial loop whose trips may differ between the threads of a workgroup: a loop whose index varying calls varying.val · tolk
- Rune.check_gradscheck_grads p f params compares the gradient g of the real scalar f at params with central differences of step eps (default 1e-4) along two fixed directions v, ones and sin params + 1.1 on the float and complex arrays: it is Ok () if Re Σ g v, the derivative g predicts, is within tol (default 1e-2) of each difference, relative to its magnitude where above 1, and Error msg naming the direction and both derivatives otherwise.val · rune
- Tolk.Setting.check_oobcheck_oob makes specification checks prove that memory accesses stay within their buffers, from CHECK_OOB.val · tolk
- Rig_remote_abi.check_transferscheck_transfers ~send ~receive is Ok () iff a rail can carry send and receive: they are not both empty, and each transfer's length is positive and its src and dst are non-negative, with src + length and dst + length at most 260.val · rig
- Rune.check_transposecheck_transpose p q f ~transpose args tests that transpose is f's transpose on three pairs of standard normal v, of args' structure, dtypes, shapes and placements, and w, of f v's, drawn from fixed keys (both parts of a complex array, zeros in an integer or boolean one): it is Ok () if <w, f v> = Re Σ w (f v) is within tol (default 1e-5) of <transpose w, v>, relative to the larger sum of the products' magnitudes where it is above 1, and Error msg naming the pair and both pairings otherwise.val · rune
- Quill.Session.checkpointcheckpoint s saves the current document to the undo history.val · quill
- Hugin.View.choicechoice name ~init is the key name of a choice among strings, initially init.val · hugin
- Nx.choleskycholesky a is the factor of the Hermitian positive definite matrix that a's triangle (default Lower) and the real parts of its diagonal name, which alone are read: L, lower triangular with a real and positive diagonal and a = L Lᴴ, or under ~triangle:Upper U = Lᴴ, upper triangular with a = Uᴴ U.val · nx
- Nx_cpu.choleskycholesky t ~dst a stores L with a = L Lᴴ under Lower, or U = Lᴴ under Upper, +0 in the other triangle and its diagonal real and positive.val · nx
- Nx_kernel.S.choleskycholesky t ~dst a stores L with a = L Lᴴ under Lower, or U = Lᴴ under Upper, +0 in the other triangle and its diagonal real and positive.val · nx
- Nx_kernel.Shape.choleskycholesky a is a, square.val · nx
- Norn.Bij.cholesky_corrcholesky_corr maps a vector of n (n - 1) / 2 coordinates to the Cholesky factor of an n × n correlation matrix.val · norn
- Hugin_kit.Ticks.choosechoose ~locale ~notation ~spacing ~length ~measure s is the most legible ticks of s for an axis length long on which the normalised values 0 and 1 lie at the ends, where measure l is the extent that label l needs along the axis, in the unit of length, including the clearance it wants from its neighbours, and spacing, 0.val · hugin
- Choosing an AlgorithmBrot supports 5 tokenization algorithms.guide · brot
- Kaun_datasets.cifar10cifar10 () is (train_x, train_y, test_x, test_y): 50,000 training and 10,000 test examples.val · kaun
- Hugin_gg.Path.circlecircle c r is the closed circle of centre c and radius r, starting at angle 0.val · hugin
- Hugin_kit.Symbol.circlecircle is a circle.val · hugin
- Ymir.Region.circlecircle p ~radius is the disc of radius about the origin of p's plane, in a unit of that plane.val · ymir
- Ymir.Frame.cirstype · ymir
- Ymir.Frame.cirscirs is the Celestial Intermediate Reference System of date: from GCRS by the IAU 2006/2000A precession-nutation, which takes the pole to the CIP, and the CIO locator s, which places the CIO (IERS Conventions 2010, 5.4.4).val · ymir
- Hugin_kit.Scheme.cividiscividis runs from dark blue through grey to yellow, and looks nearly the same to viewers with a red or green deficiency (Nuñez, Anderton and Renslow).val · hugin
- Rig.ClaimClaims on memory.module · rig
- ClaimsThis page shows a function that writes its result over its argument, and a caller that keeps the argument.guide · rig
- Nx.clampclamp ~min ~max x is minimum (maximum x min) max with each bound a constant, a bound left out not applied: max where min > max, and a NaN stays a NaN.val · nx
- Tolk.Compiler_cpu.clangclang arch is the compiler of C source to objects for arch, written MACHINE[-SYSTEM],CPU[,FEATURE]..., as in "x86_64,znver2,avx,-avx512f":val · tolk
- Tolk.Cstyle.clangclang target renders C for Clang, for a CPU.val · tolk
- Ymir_units.Codata.classical_electron_radiusclassical_electron_radius r is re in m.val · ymir
- Fehu.Buffer.clearclear buf removes all transitions, keeping storage allocated.val · fehu
- Tolk.Helpers.Diskcache.clearclear () removes every entry of every table, of every version, and the writes in progress last written an hour or more before.val · tolk
- Quill.Doc.clear_all_outputsclear_all_outputs d clears outputs from all code cells.val · quill
- Quill.Session.clear_all_outputsclear_all_outputs s clears outputs from all code cells.val · quill
- Kaun_hf.clear_cacheclear_cache () removes every cached file under cache_dir (defaulting as in cache_path).val · kaun
- Quill.Session.clear_outputsclear_outputs id s clears the outputs of cell id.val · quill
- Hugin_vg.Picture.clipclip ~rule q p paints p only inside the area of q under rule, which defaults to `Nonzero.val · hugin
- Fehu.Env.clip_actionclip_action env clamps continuous actions to the bounds of the inner environment's Space.Box action space.val · fehu
- Vega.clip_by_global_normclip_by_global_norm p ~max_norm grads scales grads so that their norm, Nx.Ptree.norm at float32, does not exceed max_norm.val · vega
- Vega.clip_by_valueclip_by_value p ~max grads clips every gradient element to the interval [-.val · vega
- Fehu.Env.clip_observationclip_observation ~low ~high env clamps observations to [low; high].val · fehu
- Rig_amd.clockThe type for how a device reads the GPU's clock, the clock of its stamps, against the host's clock (C functions).type · rig
- Rig_amd_abi.Thread_trace.clockclock g trace maps a shader time of trace to the GPU's clock (Capability.t.clock_hz) through trace's realtime markers.val · rig
- Rig_nv.clockThe type for how a device reads the GPU's timer, the clock of its stamps, against the host's clock (Submitting).type · rig
- Talon.Time.clockThe type for the witness of a stamp's clock, which a column type holds.type · talon
- Talon.Type.clockclock u is Clock in the unit u.val · talon
- Tolk.Shape.cloneclone ~device u is a copy of u into new storage on device (default u's).val · tolk
- Fehu.Env.closeclose env releases resources held by the environment.val · fehu
- Fehu.Vec_env.closeclose t closes all environments.val · fehu
- Hugin_gg.Path.closeclose p is p with its last subpath closed by a line back to its start, after which p has no current point.val · hugin
- Rig.closeclose d waits for the work submitted on d before the call, then loses d with the reason "closed" and returns once its driver's stop (Rig_edge.Driver.stop) has returned, which may be before d has stopped (Stops and forks).val · rig
- Rig_bnxt.closeclose nic unregisters the process from nic's firmware, turns its bus mastering off, frees its system memory, regions included, turns its relaxed ordering back on if open_ found it on, and releases its function.val · rig
- Rig_mlx5.closeclose nic destroys every object of nic and closes it.val · rig
- Rig_remote.closeclose j ends j in order.val · rig
- Rig_remote_proxy.Link.closeclose j ends j.val · rig
- Fehu.Env.closedclosed env is true iff the environment has been closed.val · fehu
- Nx.Placement.closedclosed p is p where the value is no sum: each device holds the window p gives it of the value itself.val · nx
- Sowilo.closingclosing ~kernel img is the dilation of img followed by the erosion by the reflected element, as opening.val · sowilo
- Hugin_font.Run.clustercluster r i is the byte index in text r of the first Unicode character that glyph i of r renders.val · hugin
- Nx_array.Layout.coalescecoalesce ls is layouts of one shape that reach the positions ls reach, in the same C order of indices, with their axes of extent 1 dropped and adjacent axes merged where every layout lays them out as one run.val · nx
- Talon.Expr.coalescecoalesce es is the first of es that is not null, or null.val · talon
- Tolk.CoalesceLate index simplification and memory coalescing.module · tolk
- Ymir_units.CodataCODATA releases of the measured constants.module · ymir
- Nx_array.Dtype.codecode dt is dt's code: its index in all and the constant NX_<NAME> of nx_dtype.h, as NX_FLOAT32 for Float32.val · nx
- Quill.Cell.codecode ?id ~language ?attrs source is a code cell in language with the given source.val · quill
- Rig_edge.codeThe type for a binary a driver took.type · rig
- Ymir.Transform.codeThe type for projections, by their FITS codes.type · ymir
- Rig_amd_abi.Code_objectCode objects, the programs AMD GPUs run.module · rig
- Tolk.CodegenCode generation: from a kernel to a program.module · tolk
- Tolk.Transcendental.cody_waite_reductioncody_waite_reduction d is (r, q) with d = q * pi/2 + r and |r| about pi/4 at most, for |d| <= 39800.val · tolk
- Jera.Piecewise.coefficientscoefficients p is p's coefficients, each leaf of shape [pieces; degree + 1] @ value.val · jera
- Talon.ColColumn handles.module · talon
- Talon.Ext.colcol e name is the column name read through e.val · talon
- Fehu.CollectTrajectory collection from environments.module · fehu
- Rune.Probe.collectcollect p f is f ()'s result, or the exception it raised with its backtrace, and the offers to p that reached it.val · rune
- Rune.Total.collectcollect t ~init f is (f (), total), total being init plus each addition to t that reaches it, in the order they reach it, and init itself if none does.val · rune
- Talon.Expr.collectcollect a is a's values in frame order as a list of a's type, with one element per row of the frame and a null element for a null value.val · talon
- Tolk.Realize.collectingcollecting () is true on a domain running kernels's function.val · tolk
- Collection and EvaluationThis guide covers trajectory collection, replay buffers, generalized advantage estimation, and policy evaluation.guide · fehu
- Hugin_gg.ColorColours.module · hugin
- Hugin_kit.Scheme.colorcolor ~unknown s u is the colour of the normalised value u in the continuous reading of s, and unknown if u is nan.val · hugin
- Hugin_text.Text.colorcolor c t is t with the colour c on every character that has none.val · hugin
- Hugin_vg_instances.Instances.colorcolor cs i ~own inherited is what paints a leaf of instance i: its own colour own cs.(i), or inherited if cs is None.val · hugin
- Tolk.Helpers.colorThe type for the colors of terminal text.type · tolk
- Tolk.Ops.Axis_type.colorcolor a is the colour a's letter is printed in.val · tolk
- Tolk.Helpers.coloredcolored c s is s wrapped in the ANSI escape sequences that paint it, or its background if background is true, in c: a 24-bit color for Rgb.val · tolk
- Tolk.Postrange.Scheduler.colored_shapecolored_shape k is the size of each of k's axes, right-aligned in four columns and coloured by its role (Ops.Axis_type.color).val · tolk
- Hugin_kit.Scheme.colorscolors n s is the n colours of the discrete reading of s, in class order, as a fresh array.val · hugin
- Talon.columncolumn t name is the column name of t, canonical (Column): t's own when it already is.val · talon
- Talon.ColumnColumns: one typed array of values, some of them null.module · talon
- Talon.Expr.columnThe type for functions of one column of any type.type · talon
- Ymir_fits.Fits.Table.columnThe type for column descriptions.type · ymir
- Norn.Summary.columnscolumns s is s's columns, each a vector with one element per row: mean, sd, q5, median, q95, mcse, then ess_bulk, ess_tail and rhat for chains, ess for weighted draws.val · norn
- Talon.Schema.columnscolumns s is the columns of s, in order.val · talon
- Ymir_fits.Fits.Table.columnscolumns t is t's columns in file order.val · ymir
- Nx.combineThe type for how scatter combines a target with its updates.type · nx
- Nx_kernel.combinetype · nx
- Nx.combine_patchescombine_patches ~output_size ~kernel_size ~stride ~dilation ~padding p is the adjoint of extract_patches from spatial axes output_size: p, of shape [lead…; P; L], is summed into [lead…; output_size…], each of its elements added at the position that extract_patches reads it from, and one that lands in the padding dropped.val · nx
- Tolk.Comgr_workermodule · tolk
- Coming from PythonThis page maps Python scientific computing concepts to their Raven equivalents.manual
- Rig_remote_proxy.Wire.commandThe type for the messages an agent reads from its controller.type · rig
- Ymir_fits.Fits.Header.commentarycommentary k h is the text of h's commentary records whose keyword is k (bytes 9–80, trailing spaces dropped), in order.val · ymir
- Tolk.Ops.commit_dtypecommit_dtype x is the type x is stored as: a weak integer commits to the narrowest type holding its bounds (Dtype.commit_int), a weak float to its default (Dtype.strong).val · tolk
- Tolk.Dtype.commit_intcommit_int ~default_int lo hi is the first of default_int, Int32, Int64 and Uint64 that holds every integer from lo to hi, and Int64 if none does.val · tolk
- Tolk.Uop_weak.commit_weak_constscommit_weak_consts u dt is u with each of its sources that is a weak constant converted to dt (Ops.ccast), and u if dt is None.val · tolk
- Talon.Type.commoncommon ts is the type at which operands of the types ts meet in an operation, and None if they do not meet or ts is empty.val · talon
- Tolk.Op.Set.commutativecommutative is the binary operations f with f x y = f y x, whose operands rewrites may swap: Add, Mul, Max, Cmpne, Cmpeq, Xor, And and Or.val · tolk
- Tolk.Shape.commutativecommutative orders the two operands of each commutative weak integer operation by Ops.compare_structure, least first, so that two sums of the same terms become the same node.val · tolk
- Ymir.Cosmology.comoving_distancecomoving_distance c z is χ, along the line of sight.val · ymir
- Ymir.Cosmology.comoving_volumecomoving_volume c z is the comoving volume within z per steradian, DH3 χ̂3 W(Ωk χ̂2) with χ̂ = χ/DH and W(x) = (S(4x) - 1)/(2x).val · ymir
- Ymir.Cosmology.comoving_volume_elementcomoving_volume_element c z is dVC/dz per steradian, c DM2/H.val · ymir
- Hugin_font.Font.comparecompare f f' is a total order on fonts compatible with equal.val · hugin
- Hugin_gg.Affine.comparecompare m n orders maps lexicographically by xx, yx, xy, yy, x0 and y0.val · hugin
- Hugin_gg.Box2.comparecompare a b orders boxes lexicographically by minx, miny, maxx and maxy.val · hugin
- Hugin_gg.Color.comparecompare c c' orders colours lexicographically by red, green, blue and alpha.val · hugin
- Hugin_gg.P2.comparecompare p q orders points by x, then by y.val · hugin
- Hugin_gg.Stroke.comparecompare s s' is a total order on stroke styles compatible with equal.val · hugin
- Hugin_kit.Time.comparecompare t t' orders instants from the past to the future.val · hugin
- Hugin_text.Text.comparecompare t t' is a total order on texts compatible with equal.val · hugin
- Nx_kernel.Prog.compareThe type for comparisons of two operands of one dtype, to booleans.type · nx
- Nx_kernel.Prog.E.comparecompare k x y is Op2 (Compare k, x, y).val · nx
- Talon.Time.comparecompare t0 t1 orders stamps of one clock chronologically.val · talon
- Talon.Time.Date.comparecompare t0 t1 orders dates chronologically.val · talon
- Talon.Time.Span.comparecompare d0 d1 orders spans from the most negative to the most positive.val · talon
- Tolk.Bigint.comparecompare orders integers by value.val · tolk
- Tolk.Dtype.comparecompare is the promotion order: by priority, then bitsize, then name.val · tolk
- Tolk.Dtype.Value.comparecompare v0 v1 is a total order on values by magnitude, where NaN is less than every other value and equal to itself.val · tolk
- Tolk.Op.comparecompare o0 o1 orders o0 and o1 by their declaration order in t.val · tolk
- Tolk.Ops.Axis_type.comparecompare orders axis types by their declaration order in t.val · tolk
- Tolk.Ops.comparecompare is a total order compatible with equal.val · tolk
- Tolk.Opt.comparecompare o0 o1 orders optimisations by kind (Tc, Split, Padto, Swap), then axis, then arguments, a split's target among them in the order of target.val · tolk
- Ymir_units.Unit.comparecompare u w totally orders units, compatibly with equal: it is the byte order of their canonical texts.val · ymir
- Rig_pci.Machine.compare_addresscompare_address a b orders bus addresses in bus order: by domain, then bus, device and function, each as a number.val · rig
- Tolk.Ops.compare_range_argcompare_range_arg r0 r1 orders the ranges r0 and r1 by argument: axis type (Axis_type.compare), then identity, lexicographically, a prefix before its extensions.val · tolk
- Tolk.Ops.compare_structurecompare_structure u0 u1 orders by operation, then argument as pp_arg prints it, then data type, then sources, recursively.val · tolk
- Talon.Type.compare_valuecompare_value t v0 v1 orders v0 and v1 by talon's total order on the values of t, the order that sorting, comparisons and grouping use:val · talon
- Tolk.Op.Set.comparisoncomparison is the operations whose result is a boolean whatever the type of their operands: Cmplt, Cmpne and Cmpeq.val · tolk
- Compilationjit s f is f compiled through tolk, raven's compiler.guide · rune
- Tolk.Codegen.compilecompile prg ren is the program linearize made, prg, completed for ren with the source ren writes and its binary, made anew and kept nowhere: neither the program, in memory or on disk, nor its binary, which ren's compiler makes (Renderer.Compiler.compile) without reading or filling its disk cache table, so that Setting.assert_compile does not apply.val · tolk
- Tolk.Renderer.Compiler.compilecompile c src is src compiled by c.val · tolk
- Tolk.Renderer.Compiler.compile_allcompile_all c srcs is List.map (compile_cached c) srcs, with the sources that c's disk cache does not hold, each once, compiled by one call of c's function: one of batch's compiles them in one run of its toolchain.val · tolk
- Tolk.Renderer.Compiler.compile_cachedcompile_cached c src is the binary of src held in c's disk cache table, or else compile c src (compile), which is then kept there, while Setting.ccache holds; otherwise it is compile c src.val · tolk
- Tolk.Renderer.Compiler.Compile_errorRaised by a compiler that rejects its source, with the toolchain's message.exception · tolk
- Tolk.Realize.compile_linearcompile_linear ~search ~targets linear is the schedule linear, whose devices compile for the targets targets gives, ready to link and run:val · tolk
- Tolk.Compiler_metal.compilercompiler () is the compiler of Metal source to Metal libraries, the binaries that Metal loads.val · tolk
- Tolk.Renderer.CompilerCompilers: source to binary.module · tolk
- Tolk.Compiler_amdThe compiler of HIP for AMD GPUs.module · tolk
- Tolk.Compiler_cpuThe compiler of C for processors.module · tolk
- Tolk.Compiler_cudaThe compiler of CUDA C++ for NVIDIA GPUs.module · tolk
- Tolk.Compiler_metalMetal: the compiler of Metal Shading Language for Apple GPUs.module · tolk
- Tolk.Compiler_workerCompiles in processes of their own.module · tolk
- Quill.Kernel.completecomplete k c ~pos is the completions at byte pos of c's source.val · quill
- Talon.Kit.completecomplete ks q is q with a row for each combination of the values of the columns ks that q lacks, its other columns null.val · talon
- Rig_edge.completionThe type for how a device's word advances.type · rig
- Rig_mlx5_abi.CompletionA completion is what the NIC writes once a work entry completes or fails.module · rig
- Quill.Kernel.completion_itemThe type for completion items.type · quill
- Quill.Kernel.completion_kindtype · quill
- Nx.complexcomplex dt ~re ~im is the complex value of parts re and im, broadcast together, each stored in dt's part format by the conversion rule: infinities, NaN and signed zeros stay, and one part never affects the other.val · nx
- Nx.complex128val · nx
- Nx.complex128_ttype · nx
- Nx.complex32_elttype · nx
- Nx_array.Dtype.complex32_eltComplex numbers of two binary32 components, real part first.type · nx
- Nx.complex64val · nx
- Nx.complex64_elttype · nx
- Nx_array.Dtype.complex64_eltComplex numbers of two binary64 components, real part first.type · nx
- Nx.complex64_ttype · nx
- Rig_nv_abi.Method.componentThe type for the sizes of a fill's components.type · rig
- Ymir.Cosmology.componentThe type for components.type · ymir
- Norn.Bij.composecompose b c maps u to b's image of c's image of u.val · norn
- Composition and LayoutFigures combine into figures.guide · hugin
- Talon_arrow.compressiontype · talon
- Talon_parquet.compressiontype · talon
- Ymir_units.Codata.compton_wavelengthcompton_wavelength r is λC, the electron's Compton wavelength, in m.val · ymir
- Fehu.Gae.computecompute ~rewards ~values ~terminated ~truncated ~next_values ~gamma ~lambda is (advantages, returns).val · fehu
- Rig_amd_abi.Capability.computeThe type for the packets a device's compute queue reads.type · rig
- Rig.computescomputes d is true iff d runs work, as every device but an io device does.val · rig
- Tolk.Decomp_dtype.computescomputes r is the data types of Dtype.all, in order, that r's target computes: those it supports (Renderer.supported_dtypes), and those pm_dtype_decomps emulates where it lacks them, the 8-bit floats, Dtype.t.Bfloat16, Dtype.t.Float16, Dtype.t.Int64 and Dtype.t.Uint64.val · tolk
- Nx_cpu.computes_oncomputes_on d is true iff these kernels compute on d's memory.val · nx
- Nx_kernel.S.computes_oncomputes_on d is true iff these kernels compute on d's memory.val · nx
- Brot.Encoding.concatconcat encs is the encoding with the tokens of each element of encs in order.val · brot
- Fehu.Collect.concatconcat trajs concatenates trajs into a single trajectory.val · fehu
- Hugin_text.Text.concatconcat ts is the characters of ts in order, each with its style.val · hugin
- Norn.Summary.concatconcat ss is the rows and findings of ss, in order: derived quantities summarised beside parameters.val · norn
- Nx_ragged.concatconcat rs is the rows of rs one after the other.val · nx
- Rig_nv_abi.Template.concatconcat ts is the bytes of ts one after another, with their patches moved with them.val · rig
- Talon.Expr.Str.concatconcat sep es is the texts of es joined by the literal sep, as String.concat sep joins strings: concat "_" [ session; unit ] of "s01" and "u3" is "s01_u3".val · talon
- Tolk.Ops.Pattern_matcher.concatconcat ms tries the rules of ms in order.val · tolk
- Nx.concatenateconcatenate ~axis xs is xs joined along their axis axis, in order.val · nx
- Nx.condcond dt a is, in dt, the condition number of each matrix of a in the norm p (default `Two), dropping its last two axes: under `Two its greatest singular value over its least, under `One and `Inf its matrix_norm of order p times that of its inv.val · nx
- Rune.condcond b p t f is t () where the 0-d p holds and f () where it does not, as switch on p.val · rune
- Talon.Join.condThe type for conditions, which are conjunctions of atoms.type · talon
- Tolk.Shape.Sint.condThe type for conditions on symbolic integers: known, or a boolean node.type · tolk
- Rune.cond'cond' p t f is cond over one array.val · rune
- Quill_project.configProject configuration.type · quill
- Rig_pci.Function.config16config16 f off is the 16-bit value at byte off of f's configuration space.val · rig
- Rig_pci.Function.config32config32 f off is the 32-bit value at byte off of f's configuration space.val · rig
- Rig_pci.Function.config8config8 f off is the byte at off of f's configuration space.val · rig
- Kaun.Metric.confusion_matrixconfusion_matrix predictions labels is the [classes; classes] matrix whose row i, column j entry counts the examples with label i and predicted class j.val · kaun
- Tolk.Shape.conjconj us is the conjunction of us.val · tolk
- Nx.conjugateconjugate x is x with each imaginary part negated, 0.val · nx
- Rig_bnxt.Qp.connectconnect qp e connects qp to the peer's queue pair e and makes it ready to send.val · rig
- Rig_mlx5.Qp.connectconnect qp a e connects qp to the peer's queue pair e, at the smaller of both ports' MTUs, and makes it ready to send.val · rig
- Rig_remote.connectconnect ~key agents starts a job with an agent at each host and port of agents, and opens each machine's host (hosts).val · rig
- Hugin.constconst v is v for every row, in the range of the role it is bound to (Roles): ~x:(const 0.5) is the middle of the panel, and ~fx:(const "a") draws its mark in the panel of the category "a" only (Facets).val · hugin
- Nx_kernel.Prog.E.constconst dt x is Const (dt, bits dt x) (Prog.bits).val · nx
- Talon.Expr.constconst v is v on every row.val · talon
- Tolk.Dtype.constThe type for constants: a value, or `Invalid, the value of an element that is masked out.type · tolk
- Tolk.Dtype.constconst dt c is c as a constant of dt.val · tolk
- Tolk.Ops.constconst ~dtype c is the constant c of type dtype.val · tolk
- Tolk.Ops.Upat.constconst ~dtype c matches the constant c.val · tolk
- Tolk.Ops.const_factorconst_factor u is a known integer that divides every value of u.val · tolk
- Tolk.Shape.const_likeconst_like ~dtype u c is the constant c of type dtype (default u's), expanded to u's shape.val · tolk
- Tolk.Shape.const_vconst_v u v is v as a constant of u's type (const_like).val · tolk
- Kaun.Init.constantconstant v fills with v.val · kaun
- Vega.Schedule.constantconstant lr is the schedule that always returns lr.val · vega
- Ymir_units.ConstantMeasured constants.module · ymir
- Ymir_units.Codata.constantsconstants r is every constant of r, in the order of NIST's table.val · ymir
- Norn_model.constrainconstrain m c is the latent values at the coordinates c.val · norn
- Tolk.Shape.constsconsts ~dtype cs is the stack of const ~dtype c for each c of cs.val · tolk
- Rig.Claim.consumableconsumable c b is true iff c's with_ is running, c claims b's memory, b is live and Buffer.spans it, and no hold holds it: consume c b then succeeds, unless another domain holds the memory meanwhile.val · rig
- Rig.Claim.consumeconsume c ~why b consumes b's memory with the reason why, so b and every buffer made over that memory until then are dead.val · rig
- Rig_mlx5_abi.Doorbell.consumedconsumed is the byte offset in a completion queue's record of its consumer count, a 32-bit word holding the count's low 24 bits.val · rig
- Fehu.Space.containscontains s v is true iff v is valid in s.val · fehu
- Kaun.Cache_index.contextcontext index is the number of columns a sequence can hold, the width of what extend returns: the table's on an index with one, plus seq under a draft, and seq on a whole index.val · kaun
- Kaun.Rope.contextcontext t is the number of positions t rotates: positions 0 to context t - 1.val · kaun
- Tolk.Setting.contextcontext bindings f is f (), run with each setting of bindings holding its bound value; when a setting is bound twice, the later binding wins.val · tolk
- Nx_array.Layout.contiguouscontiguous s is the layout of shape s in C order at offset 0: element k in C order is at position k.val · nx
- Tolk.Ops.contiguouscontiguous x is x materialised into its own buffer: x itself if it is weak, already staged, placed nowhere, or storage.val · tolk
- Tolk.Schedule.contiguous_mops_to_viewcontiguous_mops_to_view c src is, for the copy, stage or store c of the movements and bitcasts src of a buffer, c of a view of the buffer's elements (Prepare.contiguous_view) when they are contiguous, and that view for any other c; None when they are not contiguous or src's shape is symbolic, for a copy or a stage when the elements start at bound variables, and for a store into a buffer's state after earlier stores or of a value that reads the buffer short of the stored states of other storage, which the kernel split then stages.val · tolk
- Tolk.Prepare.contiguous_viewcontiguous_view u is Some (b, offset) if the elements of u, in row-major order, are the elements of b from offset on, in order, where u is b viewed through movements and bitcasts that neither reorder nor skip elements.val · tolk
- Hugin.contourcontour ~x ~y ~fill () draws the filled contours of fields sampled on grids.val · hugin
- Nx.contractcontract dt p a b is round_dt (init + Σ a · b) over p's summed names, for each index of p's result.val · nx
- Nx_cpu.contractcontract ~batch ~contracting ~segments ~acc ~dst ~init a b stores into dst the contraction of a with b, of the shape Nx_kernel.Shape.contract gives:val · nx
- Nx_kernel.S.contractcontract ~batch ~contracting ~segments ~acc ~dst ~init a b stores into dst the contraction of a with b, of the shape Nx_kernel.Shape.contract gives:val · nx
- Nx_kernel.Shape.contractcontract ~batch ~contracting ~segments (a, sa) (b, sb) is the shape of a contraction of a with b, sa and sb their scales' shapes: the batch axes in pair order, then a's free axes, then b's, in axis order.val · nx
- Tolk.Shape.contractcontract u rs is the stack of u with the upcast ranges rs substituted by each of their values, the last range varying fastest.val · tolk
- Hugin_gg.Color.contrastcontrast c is black or white, whichever has the higher WCAG 2 contrast ratio against c: black iff the relative luminance of c, 0.2126 R + 0.7152 G + 0.0722 B for its linear components R, G and B, is at least sqrt 0.0525 - 0.05, about 0.179, where both ratios are equal.val · hugin
- Kaun.Conv2-D convolution layers (NCHW).module · kaun
- Nx.convconv x w slides the kernel w, [o; k…; c / groups], over x, [n; s…; c], channels last, for K spatial axes s… and kernel extents k…: the result, [n; t…; o], is at (b, p…, j) the sum over i and the taps q… of x[b, p · stride + q · dilation - lo…, i] times w[j, q…, i], an element outside x reading zero.val · nx
- Nx.conv_transposeconv_transpose x w is conv's transpose with respect to its input: for the z of the least spatial extents whose conv z w, under the same arguments, has x's shape [n; t…; o], it is the [n; s…; c] value with sum (mul (conv z w) x) = sum (mul z (conv_transpose x w)) for every such z: s = (t - 1) stride + e - lo - hi, e as conv states, or 0 where that is negative or t = 0.val · nx
- Ymir.Direction.convertconvert e g d is d in frame g at e's instants, their batch axes broadcast.val · ymir
- Ymir.Time.convertconvert s t is t's instants on s, converted along UTC, TAI, TT, TDB one step at a time:val · ymir
- Ymir_units.Quantity.convertconvert u q is v u (value u q).val · ymir
- Tolk.Postrange.Scheduler.convert_loop_to_globalconvert_loop_to_global k makes global each weak range that an end of k's sink closes and no reduction does, whose bound a launch computes (Ops.known_at_launch), that every buffered value (Op.t.Stage) runs inside and that take threads (threadless), if k's renderer has local indices.val · tolk
- Ymir_units.Unit.convertibleconvertible u w is true iff u / w is dimensionless.val · ymir
- Hugin.coordcoord c f draws the panels of f in the coordinate system c.val · hugin
- Hugin.CoordCoordinate systems.module · hugin
- Nx_kernel.Prog.E.coordcoord i is Coord i.val · nx
- Norn.Dist.coordscoords d is the bijector onto d's support, from coordinates in which samplers and optimisers move: Bij.identity on the reals, Bij.exp on (0, inf), Bij.interval for beta, uniform and von_mises, Bij.simplex for dirichlet, Bij.cholesky_corr for lkj_cholesky, Bij.interval, Bij.greater or Bij.less for truncated, and for sorted d's bijector after Bij.ordered.val · norn
- Norn_model.coordsThe type for unconstrained latent values: one tensor per site, sized by its degrees of freedom (a simplex of K components has K - 1).type · norn
- Norn_model.coordscoords m is the structure of m's coordinates: m('s, 'd) latent structure.val · norn
- Rig_program.copiesThe type for how many copies of its memory a program keeps.type · rig
- Nx.copycopy x is x stored afresh: equal elements in new memory.val · nx
- Nx_array.copycopy a is a fresh C-contiguous array on a's device holding a's elements, the array create makes there filled by transfer.val · nx
- Nx_kernel.Prog.E.copyval · nx
- Rig.Buffer.copycopy ~src ~dst copies src's bytes into dst and returns once they are there: submit_copy ~src ~dst [| row (length src) |], then a wait for its point, holding its read claims until it returns.val · rig
- Rig.Submission.copycopy ~src ~dst copies src's bytes into dst.val · rig
- Rig_nv_abi.Method.copycopy ~dst ~src n copies the n bytes at src to dst on the copy engine, in one launch of one line.val · rig
- Tolk.Postrange.Scheduler.copycopy k is a scheduler of k's kernel and optimisations, which applying an optimisation to either does not change in the other.val · tolk
- Rig_nv_abi.Method.copy_releasecopy_release addr v writes the 64-bit v into the 8 bytes at addr, 8-byte aligned, once the copy engine's earlier copies are complete, so that any reader who sees v, the host and other devices included, sees their writes.val · rig
- Rig_nv_abi.Method.copy_release_stampcopy_release_stamp addr v is copy_release that also writes a stamp at addr + 8 (Semaphores).val · rig
- Tolk.Ops.copy_to_devicecopy_to_device u d is the copy of u to d.val · tolk
- Tolk.Heuristic.core_fitscore_fits ren ~dtype_in ~dtype_out ~rows ~outputs ~block is true iff a contraction of dtype_in operands of one free axis each, of extents rows and outputs, accumulated in dtype_out and summed over block of them at a time, leaves streaming (Streaming) to take a tensor core of ren: the lesser of rows and outputs is past the reuse that streams first, Setting.use_tc is 1, and a core of those dtypes, the one Setting.tc_select names if any (none where ren has no such core), a float32 core of CUDA only under Setting.allow_tf32, has its K dividing block and its M and N dividing rows and outputs, in either role.val · tolk
- Hugin.cornerThe type for the corners of a panel or of a scope's figure.type · hugin
- Ymir.Grid.cornerscorners g is each cell's four corners in the world, batch @ shape g @ [4] points, counter-clockwise in the pixel plane.val · ymir
- Ymir.Healpix.cornerscorners h p is each pixel's four corners, p's shape @ [4; 3], north, west, south and east, counter-clockwise seen from outside the sphere and within 2e-15 radians as centre; NaN rows for -1.val · ymir
- Nx.correlatecorrelate ~padding x kernel is the cross-correlation of x with kernel over the last K = ndim kernel axes of x, each sum of products taken without flipping kernel.val · nx
- Nx.coscos x is the cosine of x radians.val · nx
- Nx_kernel.Prog.E.cosval · nx
- Nx.coshcosh x is the hyperbolic cosine.val · nx
- Nx_kernel.Prog.E.coshval · nx
- Vega.Schedule.cosine_decaycosine_decay ~init_value ~decay_steps ?alpha () is cosine decay from init_value to alpha * init_value over decay_steps.val · vega
- Vega.Schedule.cosine_decay_restartscosine_decay_restarts ~init_value ~decay_steps ?t_mul ?m_mul ?alpha () is cosine decay that periodically resets to init_value (SGDR).val · vega
- Ymir.CosmologyThe background of a homogeneous expanding universe.module · ymir
- Ymir_units.Unit.coulombcoulomb is A s.val · ymir
- Nx.countcount ~axes m is the number of true elements of m, a bool or a bit value: 0 of none.val · nx
- Rig_amd_amdgpu.countcount () is the number n of buses, so GPUs are numbered 0 to n - 1, or their Error.val · rig
- Rig_amd_pci.countcount ~machine () is the number n of buses, so GPUs are numbered 0 to n - 1, or their Error.val · rig
- Rig_bnxt.countcount ~machine () is the number of buses, or their Error.val · rig
- Rig_cuda.countcount () is the number of GPUs CUDA sees, or Error why if the CUDA library cannot be loaded or initialised or CUDA cannot list its GPUs.val · rig
- Rig_metal.countcount () is the number of GPUs: Ok 1 on a Mac whose GPU supports Metal, and Ok 0 otherwise.val · rig
- Rig_nv_nvidia.countcount () is the number of buses, or their Error.val · rig
- Rig_nv_pci.countcount ~machine () is the number of buses, or their Error.val · rig
- Talon.Expr.countcount a is the number of non-null values of a, as int64.val · talon
- Talon.Join.countThe type for the number of matches each row of a side must have.type · talon
- Talon.Kit.count_bycount_by ks q is one row per group of q's rows that have the same keys in the columns ks, in order of first appearance: the columns ks, then "count", the group's number of rows as int64:val · talon
- Rig_amd_abi.CounterPerformance counters, which count events of a kernel's run.module · rig
- Ymir.Transform.coverscovers t x is Some m, m where apply t x is defined: false outside a stage's domain and at NaN, of x's batch shape.val · ymir
- Tolk.Helpers.cpu_objdumpcpu_objdump lib prints the instructions of the object file lib, as objdump -d prints them.val · tolk
- Rig_mlx5.CqA completion queue is a ring where the NIC reports the entries that completed.module · rig
- Rig_mlx5_abi.Uar.cq_doorbellcq_doorbell is the byte offset in an access region of the register a process stores Doorbell.arm in.val · rig
- Brot.Encoding.createcreate ~ids ~type_ids ~tokens ~words ~offsets ~special_tokens_mask ~attention_mask () is an encoding from the given arrays.val · brot
- Fehu.Buffer.createcreate ~capacity is an empty buffer that holds at most capacity transitions.val · fehu
- Fehu.Env.createcreate ~observation_space ~action_space ~reset ~step () makes a new environment.val · fehu
- Fehu.Space.Box.createcreate ~low ~high is a continuous space where element i satisfies low.(i) <= x.(i) <= high.(i).val · fehu
- Fehu.Space.Dict.createcreate fields is a dictionary space with named fields.val · fehu
- Fehu.Space.Discrete.createcreate ?start n is a discrete space with n choices in the range [start; start + n - 1].val · fehu
- Fehu.Space.Multi_binary.createcreate n is a binary vector space of length n.val · fehu
- Fehu.Space.Multi_discrete.createcreate nvec is a multi-discrete space where element i is in [0; nvec.(i) - 1].val · fehu
- Fehu.Space.Sequence.createcreate ?min_length ?max_length s is a sequence space over s.val · fehu
- Fehu.Space.Text.createcreate ?charset ?max_length () is a text space.val · fehu
- Fehu.Space.Tuple.createcreate spaces is a tuple space.val · fehu
- Fehu.Vec_env.createcreate envs creates a vectorized environment.val · fehu
- Nx.createcreate dt s vs is the host value of shape s holding vs in C order, a float stored as Dtype.of_float says.val · nx
- Nx_array.createcreate d dt s is a fresh C-contiguous array of shape s in d's memory of kind memory (defaults to Device), at offset 0, with unspecified elements.val · nx
- Quill.Session.createcreate ?history_capacity doc creates a session from doc.val · quill
- Quill_top.createcreate ?setup ~on_event () creates a new OCaml toplevel kernel, whose Quill.Kernel.language is "ocaml".val · quill
- Rig.Buffer.createcreate d n is a buffer of n new bytes of d's memory of kind memory (defaults to Device), with unspecified contents.val · rig
- Rig_amd.Lock.createcreate () is a new lock.val · rig
- Rig_nv.Lock.createcreate () is a new lock.val · rig
- Rig_pci.Memory.createcreate f tables ~bar is the memory of the GPU of f, whose page tables are tables and whose memory BAR is bar.val · rig
- Rig_pci.Page_table.createcreate ~base fmt s ~memory ~boot ~tables ~pages manages memory bytes of a GPU's physical memory: a boot pool of its first boot bytes, the page tables' pool if tables is Pool, and the main pool.val · rig
- Rig_pci.Space.createcreate ~base n is the n virtual addresses from base on, all free.val · rig
- Tolk.Support_memory.Tlsf_allocator.createcreate ~base ~block_size ~lv2_cnt size is an allocator of the size addresses from base (default 0) on, all free.val · tolk
- Rig_disk.create_filecreate_file path n is a new file at path of n zero bytes, as a buffer on device for reading and writing.val · rig
- Tolk.Schedule.create_linear_with_varscreate_linear_with_vars ~capturing ~by_storage big_sink is the schedule that realizes the values of big_sink, whose results are stored into buffers (Op.t.Buffer), with the value of each variable it reads, by name:val · tolk
- Tolk.Schedule.create_schedulecreate_schedule ~by_storage sched_sink is the Op.t.Linear of the kernel calls of the kernel graph sched_sink (Rangeify.get_kernel_graph), in an order that runs each kernel after those that write the storage it reads, and before those that overwrite the storage it reads.val · tolk
- Munin.Artifact.created_atcreated_at a is the time a was published.val · munin
- Ymir.Cosmology.critical_densitycritical_density c z is 3H(z)2/8πG, in 3/(8π) u2 over G's unit for h0 in u.val · ymir
- Hugin_gg.Path.cropcrop b p is what p draws within b, its edges included, subpath by subpath:val · hugin
- Sowilo.cropcrop ~y ~x ~height ~width img extracts a rectangular region starting at (y, x) with the given dimensions.val · sowilo
- Hugin_kit.Symbol.crosscross is a Greek cross: five equal squares, one at the centre and one on each of its sides.val · hugin
- Nx.crosscross ~axis a b is the cross product of the 3-vectors of a and b along axis (default -1) of their broadcast shape: component i is a(i+1) b(i+2) - a(i+2) b(i+1), indices modulo 3.val · nx
- Tolk.CstyleRenderers of the C family: kernels as C, Metal, CUDA and HIP source.module · tolk
- Brot.Decoder.ctcctc ~pad_token ~word_delimiter_token ~cleanup () is a decoder for CTC (Connectionist Temporal Classification) output.val · brot
- Hugin_vg_vector.Vector.ctxThe type for where the walk is: its frame and what enclosing instances set.type · hugin
- Tolk.Decomp_dtype.ctxThe type for the context of pm_dtype_decomps: the types a kernel uses that may need emulation, and the target's renderer.type · tolk
- Tolk.Decomp_dtype.ctxctx r is the context for emulating on r's target, with no type found yet.val · tolk
- Jera.Quad.cubaturecubature ~tol ~budget f box is the integral of f over each lane's box, of d dimensions with 2 ≤ d ≤ 10.val · jera
- Jera.Grid.cubiccubic ends ~axes values is the tensor-product cubic spline, with Piecewise.ends' ends along every axis: degree 3 along each axis.val · jera
- Jera.Piecewise.cubiccubic ends x y is the cubic spline through the samples, twice continuously differentiable: degree 3.val · jera
- Hugin_gg.Path.cubic_tocubic_to c c' pt p is p with a cubic Bézier from the current point to pt with control points c and c'.val · hugin
- Rig_nv_abi.CubinCubins: the ELF objects NVIDIA's compilers make for a GPU architecture.module · rig
- Tolk.Cstyle.cudacuda target renders CUDA C++ for an NVIDIA GPU, compiled with NVRTC.val · tolk
- Tolk.Tc.cudacuda arch is the tensor cores of the NVIDIA architecture arch, named sm_ followed by its compute capability: cuda_sm89 from 89, cuda_sm80 from 80, cuda_sm75 from 75, and none below.val · tolk
- Tolk.Setting.cuda_pathcuda_path is the directory of the CUDA toolkit whose headers Compiler_cuda.nvrtc includes, from CUDA_PATH, read when the compiler is made.val · tolk
- Tolk.Tc.cuda_sm75cuda_sm75 is NVIDIA's tensor cores from compute capability 7.5: 8 x 16 x 8 tiles of Dtype.t.Float16, into Dtype.t.Float32 or Dtype.t.Float16.val · tolk
- Tolk.Tc.cuda_sm80cuda_sm80 is those from compute capability 8.0: cuda_sm75, 8 x 16 x 16 tiles of Dtype.t.Float16 and Dtype.t.Bfloat16, and 8 x 16 x 8 tiles of Dtype.t.Float32.val · tolk
- Tolk.Tc.cuda_sm89cuda_sm89 is those from compute capability 8.9: cuda_sm80 and 8 x 16 x 32 tiles of the two 8-bit floats.val · tolk
- Tolk.Shape.cumalucumalu u axis op is the inclusive running op (Op.t.Add, Op.t.Mul or Op.t.Max) of u along axis, at u's type: each element reduces the window of the elements up to it.val · tolk
- Nx.cummaxcummax ~axis x is the running max, as cumsum.val · nx
- Nx.cummincummin ~axis x is the running min, as cummax.val · nx
- Nx.cumprodcumprod ~axis x is the running product, as cumsum.val · nx
- Nx.cumsumcumsum ~axis x is, at each index, the sum of x's elements along axis up to that index, inclusive; without axis, of its elements in C order up to that one, in x's shape.val · nx
- Jera.Quad.cumulativecumulative r f knots is the integral of f from the first knot to each, by r on each interval between knots: for knots of shape [n] @ shape, a result of the same shape whose row 0 is zero and row i the sum of the first i intervals' integrals.val · jera
- Talon.Kit.cumulativecumulative r is r over the frame's rows up to and including each:val · talon
- Nx.Ptree.Walk.cursorThe type for cursors of a walk that turns the parameter's positions from 'a into 'b and the fixed arrays from the set 'd to 'e.type · nx
- Hugin_kit.CurveCurves.module · hugin
- Hugin_kit.Scale.customcustom ~transform ~forward ~inverse () is a quantitative scale with transform forward, whose inverse is inverse, and which transform names in printed forms.val · hugin
- Tolk.Ops.custom_functionThe type for the arguments of Op.t.Custom_function.type · tolk
- Tolk.Ops.custom_functioncustom_function ~dtype ~shape ~fmt name src is the external function name, or with fmt (default false) the code name in the target's language (custom_function), which returns dtype (default Dtype.t.Void).val · tolk
- Rune.custom_jvpcustom_jvp p q rule args is fst (rule args), a value of structure q, whose derivative is the tangent map map = snd (rule args): map dargs is the result's tangent for the arguments' tangents dargs, of structure p, zeros for an array the differentiation does not track.val · rune
- Tolk.Shape.custom_kernelcustom_kernel args f calls the kernel f builds on placeholders of args, and is args, each that the kernel writes (written_params) ordered after it: an argument it only reads is itself.val · tolk
- Rune.custom_transposecustom_transpose p q f ~transpose args is f args, for an f linear in args whose transpose is transpose: <w, f v> = <transpose w, v> for every v and w (Derivatives).val · rune
- Rune.custom_vjpcustom_vjp p q rule args is fst (rule args), a value of structure q, whose reverse-mode derivative is the pullback pb = snd (rule args): pb cts is the arguments' cotangents, of structure p, for the result's cotangents cts, as vjp's pullback computes them (Derivatives).val · rune
- Talon.Expr.cutcut edges a is the bin of a among edges, as int64.val · talon
- Tolk.Ops.Upat.cvarcvar ~dtype ~arg name matches a constant, named name.val · tolk
d
- Nx.Devices.dThe set's brand.type · nx
- Nx.Host.dThe set's brand.type · nx
- Ymir_units.Codata.daltondalton r is mu, the atomic mass constant, which is the dalton, in kg.val · ymir
- Munin.History.damageddamaged h is the number of damaged lines read (Reading a log).val · munin
- Hugin.Theme.darkdark is v ~ink:(Color.gray 0.92) ~paper:(Color.gray 0.1) ~scheme (): light ink on dark paper.val · hugin
- Hugin_kit.Scheme.dark2dark2 is Brewer's Dark2, eight colours.val · hugin
- Hugin.Role.dashdash is a dash pattern.val · hugin
- Hugin_gg.Stroke.dashdash s is the dash pattern of s as given to v, [] if solid.val · hugin
- Hugin_kit.DashDash patterns.module · hugin
- Hugin_vg_vector.Vector.dash_decimalsdash_decimals d is the decimals of dash lengths in a frame of d decimals: three more, since the errors of the lengths add up along a subpath, so that a thousand of them stay within the accuracy.val · hugin
- Hugin_kit.Dash.dash_dotdash_dot is v [ 4.; 2.; 1.; 2.val · hugin
- Hugin_gg.Stroke.dash_offsetdash_offset s is the dash offset of s, reduced into [0;l[ for the length l of the even pattern, and 0.val · hugin
- Hugin_kit.Dash.dasheddashed is v [ 4.; 2.val · hugin
- Hugin_kit.Scale.dashesdashes s is the dash patterns of s, if set, as a fresh array.val · hugin
- Brot.dataThe type for training data sources.type · brot
- Kaun.DataMinibatch Seq.ts over in-memory tensors, with shuffling.module · kaun
- Rig_program.dataThe type for bytes with holes.type · rig
- Ymir.Observation.datadata o is o's data, 0 where invalid.val · ymir
- Ymir_fits.Fits.datadata hdu is hdu's data unit as stored, heap included, padding excluded: for a read HDU, a view of the file, which must not change while the view lives.val · ymir
- Ymir_fits.Fits.Table.datatype · ymir
- Tolk.Helpers.data64data64 x is (hi32 x, lo32 x).val · tolk
- Tolk.Helpers.data64_ledata64_le x is (lo32 x, hi32 x).val · tolk
- Hugin_kit.Time.dateThe type for civil dates (y, m, d): the year y, the month m in [1;12] and the day d, from 1 to the length of month m of year y.type · hugin
- Talon.Col.datedate name is v Kind.date name.val · talon
- Talon.Expr.datedate d is the date d.val · talon
- Talon.Kind.datedate reads date columns.val · talon
- Talon.Time.dateThe type for dates: a day of the proleptic Gregorian calendar, counted in days since 1970-01-01.type · talon
- Talon.Time.DateDates.module · talon
- Talon.Type.datedate is Date.val · talon
- Ymir.Time.datedate t is ((y, m, d), (h, mi, sec)), the date and time of day of t's one instant, as of_date reads them: of_date s (date t) is t.val · ymir
- Ymir.Time.dayday t is each instant's modified Julian date.val · ymir
- Ymir_units.Unit.dayday is 86400 s.val · ymir
- Hugin_kit.Time.daysdays k is seconds (86_400 * k).val · hugin
- Talon.Time.Span.daysdays n is like us in days of 86,400 seconds.val · talon
- Talon.Tz.dbThe type for zone databases: a directory of TZif files, read as zones are found.type · talon
- Tolk.Setting.dbgtvdbgtv, when not empty, makes code generation print the instructions of a program that fails its specification check (Codegen), from DBGTV.val · tolk
- Nx.dctdct x is the discrete cosine transform of type type_ (default Type_2) of x along each of axes: of n points x[j], type II is y[k] = 2 Σ_j x[j] cos (πk(2j + 1) / 2n), type I y[k] = x[0] + (-1)^k x[n-1] + 2 Σ_(0<j<n-1) x[j] cos (πjk / (n-1)), type III y[k] = x[0] + 2 Σ_(j>0) x[j] cos (πj(2k + 1) / 2n) and type IV y[k] = 2 Σ_j x[j] cos (π(2j + 1)(2k + 1) / 4n).val · nx
- Rig.Buffer.deaddead b is Some why if b is dead, where why is the reason its memory was consumed with (Claim.consume), or its borrow ended with (unborrow), and None otherwise.val · rig
- Tolk.Heuristic_common.debugdebug () is the level of Setting.debug.val · tolk
- Tolk.Setting.debugdebug is the verbosity of diagnostics printed on standard output, from DEBUG.val · tolk
- Tolk.Setting.debug_linearizedebug_linearize prints each node linearization places, with its position, operation, ranges and priority (Linearizer.linearize), from DEBUG_LINEARIZE.val · tolk
- Ymir_units.Unit.decadeca u is 10 u.val · ymir
- Ymir_units.Unit.decideci u is 10-1 u.val · ymir
- Hugin_kit.Locale.decimaldecimal l is the decimal separator of l.val · hugin
- Ymir_units.Unit.decimaldecimal s is the exact positive decimal s, as in "6.62607015e-34": digits ["." digits] ["e" ["-"] digits], with no sign, no + and no E.val · ymir
- Hugin_float_text.Float_text.decimalsdecimals n x is |x| rounded to a multiple of 10-n, to nearest, ties to even.val · hugin
- Hugin_kit.Number.decimalsdecimals dtype x is the least d >= 0 such that x written in Plain notation with d decimals, read as a decimal and rounded directly to dtype, is x rounded to dtype: the fewest decimals that reproduce x in its source dtype.val · hugin
- Hugin_vg_vector.Vector.decimalsdecimals k is the number of decimals that keeps a number of a frame the page magnifies k times within 0.0005 point once rounded: 3 for k <= 1., and one more per power of ten above, at most 17.val · hugin
- Munin_float_text.Float_text.decimalsdecimals n x is |x| rounded to a multiple of 10-n, to nearest, ties to even.val · munin
- Norn_float_text.Float_text.decimalsdecimals n x is |x| rounded to a multiple of 10-n, to nearest, ties to even.val · norn
- Rig_amd_abi.declaredeclare c is c under key, as the facts of an AMD device carry it (Rig_edge.facts).val · rig
- Rig_cuda_abi.declaredeclare c is c under key, as the facts of a CUDA device carry it (Rig_edge.facts).val · rig
- Rig_metal_abi.declaredeclare c is c under key, as the facts of a Metal device carry it (Rig_edge.facts).val · rig
- Rig_nv_abi.declaredeclare c is c under key, as the facts of an NVIDIA GPU carry it (Rig_edge.facts).val · rig
- Rig_remote_abi.declaredeclare c is c under key, as the facts of a proxy carry it (Rig_edge.facts).val · rig
- Brot.decodedecode t ids is the text obtained by decoding ids through t's vocabulary and decoder.val · brot
- Brot.Decoder.decodedecode decoder tokens applies decoder to tokens and returns the decoded text.val · brot
- Talon_csv.decodedecode f r is the table of the records that r reads, read as f says, in batches of about one MiB of input.val · talon
- Talon_json.decodedecode f r is the table of the rows that r reads, read as f says, in batches of about one MiB of input.val · talon
- Brot.decode_finishdecode_finish s is the text of the ids fed to s that no step has given out yet, decoded as the end of the ids.val · brot
- Brot.decode_stepdecode_step s id is (text, s'), where s' is s fed id and text is the text that no id fed after id changes, less what earlier steps gave.val · brot
- Brot.decode_streamdecode_stream t is a decoding by t that has been fed no token.val · brot
- Tolk.Stage.decodeddecoded o is true iff o converts integers it reads into floats, by a cast or a bitcast, outside any access's address.val · tolk
- Brot.decoderdecoder t is t's decoder, if any.val · brot
- Brot.DecoderDecoding tokens back to text.module · brot
- Tolk.Stage.decodesdecodes ren sink is the kernel sink, whose optimisations are applied, with its decoded operand's panel staged where ren targets the host ("CPU").val · tolk
- Tolk.Decomp_dtypeEmulated data types.module · tolk
- Tolk.Decomp_opOperations as the operations a target has.module · tolk
- Tolk.Helpers.dedupdedup (module H) l is l with each element kept only at its first occurrence, elements being compared with H.equal.val · tolk
- Hugin.Theme.defaultdefault is v ().val · hugin
- Hugin_kit.Locale.defaultdefault is v ().val · hugin
- Quill.Cell.default_attrsdefault_attrs is {collapsed = false; hide_source = false}.val · quill
- Quill_project.default_configDefault configuration with all fields empty/none.val · quill
- Tolk.Dtype.default_floatdefault_float () is the data type named, in any case, by the current value of the setting Setting.default_float, DEFAULT_FLOAT.val · tolk
- Tolk.Setting.default_floatdefault_float names the data type of floating-point values that do not state one, from DEFAULT_FLOAT.val · tolk
- Tolk.Dtype.default_intdefault_int () is the data type named, in any case, by the current value of the setting Setting.default_int, DEFAULT_INT.val · tolk
- Tolk.Setting.default_intdefault_int names the data type of integer values that do not state one, from DEFAULT_INT.val · tolk
- Tolk.Op.Set.definesdefines is the operations that declare storage or variables, and constant bytes: Param, Refparam, Buffer, Alloc and Binary.val · tolk
- Ymir_units.Unit.degreedegree is π/180 rad.val · ymir
- Jera.Ode.delaydelay y m ~tol ~budget ~pieces f ~lags ~history ~at y0 is the state at each time of at, increasing, of the solution of y' t = f t (y t) (y (t − τ)) for the constant lags τ of lags, 1-D, from y0 at at.(0), stacked as sample stacks them.val · jera
- Munin.Store.delete_artifactdelete_artifact s a deletes the version a and the aliases that name it.val · munin
- Munin.Store.delete_rundelete_run s id removes the run id and every file it holds.val · munin
- Jera.Linear.densedense materialises a as an n × n matrix, n the size of the vector, with one application to each vector of the standard basis, and solves it with Nx.solve: n + 1 applications with the check, and 2n³/3 flops.val · jera
- Norn.densityThe type for log densities over positions of 'u.type · norn
- Ymir.Cosmology.density_parameterdensity_parameter c i z is Ωi(z).val · ymir
- Tolk.Stage.depthdepth ~used ren tiles is the number of stages a core whose staged operands' tiles hold tiles bytes takes on ren, beside used bytes (default 0) of shared memory other cores take: 2 where ren copies asynchronously and two stages fit Renderer.t.shared_max, 1 where one does, and 0, staging nothing, where neither does, where tiles is empty or a tile holds less than one copy of 16 bytes, where ren has no local threads or no shared memory, or where Setting.stage is off.val · tolk
- Nx_quant.dequantdequant dt w is the values of w at dt, of shape shape w, each rounded once to dt.val · nx
- Rig_bnxt.Region.deregisterderegister r deregisters r and frees the memory the library allocated for it, the memory host made included, and unpins the memory pages pinned.val · rig
- Rig_mlx5.Region.deregisterderegister r deregisters r.val · rig
- Jera.Grid.derivativederivative ~axis g is the partial derivative of g along axis, of one degree less along it.val · jera
- Jera.Piecewise.derivativederivative p is the derivative of p in x on each piece, of degree one less (and 0 from degree 0).val · jera
- Derivativesrune differentiates in two modes.guide · rune
- Talon.Query.derivederive os q is q's rows with each of q's columns and the outputs os: an output replaces the column of its name in place, with the output's type, and the others follow q's columns, in order.val · talon
- Tolk.Shape.derived_dtypesderived_dtypes u src is the operands' meet and u's own type when u is broadcastable and sources src commit both to a width.val · tolk
- Talon.Order.descdesc name orders the column name descending, nulls last.val · talon
- Hugin_font.Font.descentdescent f is the distance below the baseline that lines of f reserve, from the table ascent reads.val · hugin
- Talon.Kit.describedescribe q summarizes each integer or float column of q, in order, as one row of the columns "column", its name; "count", its number of values that are not null; "nulls", its number of nulls; "mean"; "std"; "min"; "q25", "median" and "q75", its quartiles; and "max".val · talon
- Rig_bnxt.Qp.destroydestroy qp destroys qp and frees its rings.val · rig
- Rig_mlx5.Cq.destroydestroy cq destroys cq and frees its ring.val · rig
- Rig_mlx5.Qp.destroydestroy qp destroys qp and frees its ring.val · rig
- Nx.detdet a is the determinant of each matrix of a, the product of the diagonal of lu's u negated once per row interchange.val · nx
- Rig_amd_pci.detachdetach i takes GPU i of this machine from amdgpu as Rig_pci.Driver.detach does, and makes its memory BAR as large as the BAR and its bridge allow.val · rig
- Rig_bnxt.detachdetach i takes NIC i of this machine from its kernel driver as Rig_pci.Driver.detach does.val · rig
- Rig_nv_pci.detachdetach i detaches GPU i of this machine from its kernel driver, so that its opener can take it, unless it is detached already (Rig_pci.Driver.detach).val · rig
- Rig_pci.Driver.detachdetach d m i detaches function i of m from its kernel driver, so that a process can take it.val · rig
- Rune.detachdetach x is x with a zero derivative under every differentiation around the call: its tangent is zero.val · rune
- Munin.Provenance.detectdetect () is the current process's provenance.val · munin
- Ymir_units.Codata.deuteron_massdeuteron_mass r is md in kg.val · ymir
- Munin_sys.deviceThe type for what is known of a device at a call.type · munin
- Munin_sys.devicedevice name read is a reading that calls read and gives each field that is Some as a level, under dev/<name>/ and the key its field names.val · munin
- Nx.Placement.devicedevice s d holds the whole value on d alone, a device of s; on s where s has d alone.val · nx
- Nx_array.devicedevice a is the device of a's buffer, the device a lives on.val · nx
- Rig.Buffer.devicedevice b is the device b is on: d for a buffer that create, of_io or borrow made on d, host for one that of_string or of_bigarray made, and its buffer's device for a view and for Claim.consume's result.val · rig
- Rig.Image.devicedevice i is the device i is loaded on.val · rig
- Rig.Point.devicedevice p is Some d, p's device d, until d has stopped and a later device took its slot (Opening and closing), and None from then on.val · rig
- Rig_bnxt.Region.devicedevice nic runs is a region over another device's memory at runs, (bus address, bytes) pairs in order, such as a GPU's memory behind its BAR.val · rig
- Rig_disk.devicedevice is this machine's disk, named "DISK".val · rig
- Rig_remote_abi.deviceThe type for the record of another machine's device.type · rig
- Rig_remote_proxy.devicedevice l a c describes the proxy on l of the agent's device a, whose record is c.val · rig
- Tolk.DeviceDevices as a compiler sees them: the renderer that writes a device's programs, and the compiled program a runtime loads.module · tolk
- Tolk.Ops.deviceThe type for devices.type · tolk
- Tolk.Ops.devicedevice u is where u lives: its argument's device for storage and copies, its sources' first otherwise.val · tolk
- Nx.devicesdevices ~kernels ds mints a new set over ds, in their order, with a brand no other set has, computed eagerly by kernels.val · nx
- Nx.DevicesDevice sets.module · nx
- Nx.DevicesThe type for device sets as modules: a brand and its placements.module type · nx
- Nx.Placement.devicesdevices p is p's set.val · nx
- Rig_remote.devicesdevices h kind is a proxy of each device of kind, such as "CUDA", that the agent of h's machine serves (serve), in the agent's index order.val · rig
- Norn.DiagDiagnostics of draws.module · norn
- Nx.diagdiag ~k v is, for a vector v of n elements, the square of n + |k| whose kth diagonal (default 0, above the main one where positive) holds v and whose other elements are zero; for a matrix, its diagonal diagonal ~offset:k v.val · nx
- Quill.Kernel.diagnosticThe type for diagnostics.type · quill
- Quill.Kernel.diagnosticsdiagnostics k c is the parse and type errors of c's source.val · quill
- Norn.Gaussian.diagonaldiagonal u dtype ~mean ~scale is the Gaussian of mean mean and standard deviation scale, element by element.val · norn
- Nx.diagonaldiagonal ~offset ~axis1 ~axis2 x is the elements of x at index i along axis1 and i + offset along axis2 (default the last two axes, and 0), for each i where both lie in their axes; that axis replaces the two, last, the others kept in order.val · nx
- Rig_remote_proxy.Wire.dialdial fd ~key ~self ~peer runs the dialing end's handshake on the connected socket fd, as self, to peer.val · rig
- Hugin_kit.Symbol.diamonddiamond is two equilateral triangles joined by a side parallel to the x axis: a rhombus standing on a corner, taller than wide.val · hugin
- Fehu.Space.Dictmodule · fehu
- Rig_bin.Proc.die_bydie_by s restores s's default action and sends it to this process.val · rig
- Munin.Run.diffdiff r is the absolute path of r's git.diff (Files a run logged), if it had one when read.val · munin
- Nx.Ptree.Skeleton.diffdiff ~this k ~that k' is None iff equal k k', and otherwise the message of Maps and folds for their first differing visit, this and that placing each side: with ~this:"here" and ~that:"in the previous key", "window: int 8 here, int 4 in the previous key".val · nx
- Talon.Expr.Temporal.diffdiff a b is the span from b to a: a duration of their common unit for datetimes, durations and clocks, which meet as Type.common says, and duration[s] for dates.val · talon
- Talon.Schema.diffdiff s0 s1 is the changes from s0 to s1: first the Removed and Retyped columns in the order of s0, then the Added ones in the order of s1.val · talon
- Tolk.Op.Set.diffdiff s0 s1 is the set of the operations in s0 and not in s1.val · tolk
- Ymir.Time.diffdiff a b is a − b in seconds of their scale, their shapes broadcast, within half an ulp of the result and 7.3e-12 s.val · ymir
- Nx.digammadigamma x is ψ(x) = Γ'(x) / Γ(x), the mean of log X for X a gamma draw of shape x, on every float: -inf at +0., inf at -0.val · nx
- Munin.Artifact.digestdigest a is the digest of a's content, blake2b-256: followed by 64 hexadecimal digits (Blobs).val · munin
- Rig_pci.Firmware.digestdigest s is the lowercase hexadecimal BLAKE2b digest of s, 32 bytes long.val · rig
- Tolk.Source_digest.digestval · tolk
- Ymir_fits.Fits.digestdigest hdu is blake2b-256:<hex> of every header of hdu's file, each as Header.to_string gives it, in file order; of hdu's header for a constructed one.val · ymir
- Brot.Pre_tokenizer.digitsdigits () splits on digit boundaries.val · brot
- Sowilo.dilatedilate ~kernel img is img with each pixel replaced by the greatest value over its neighbourhood.val · sowilo
- Hugin.dimdim ~scale ~valid ~title ~labels k is the index along axis k of the shape of the mark it is bound in, as indexed categories: the datum (i0, …, in) is the category identified by ik.val · hugin
- Nx.dimdim a x is the extent of x's axis a.val · nx
- Nx_array.Layout.dimdim l i is the extent of l's axis i.val · nx
- Rig_nv_abi.Qmd.dimThe type for the sizes of a launch.type · rig
- Tolk.Tc.dimsdims tc is (n, m, k), the tile's dimensions N, M and K.val · tolk
- Munin.Run.dirdir r is the absolute path of r's directory.val · munin
- Brot.directionThe type for padding and truncation directions.type · brot
- Nx_kernel.directionThe type for a transform's direction.type · nx
- Talon.Kit.directionThe type for the directions of an as-of join.type · talon
- Ymir.DirectionDirections in a frame.module · ymir
- Rig_pci.Firmware.directorydirectory () is rig's firmware directory, the one rig firmware fills when given none: $XDG_DATA_HOME/rig/firmware, or $HOME/.local/share/rig/firmware where XDG_DATA_HOME is unset, empty or relative.val · rig
- Norn.Dist.dirichletdirichlet ~concentration is over vectors along the last axis of concentration, of positive components summing to one, with density proportional to prod_k x_k^(α_k - 1).val · norn
- Nx.Rng.dirichletdirichlet a is Dirichlet draws of concentrations a along its last axis, one draw per row: the gamma draws of a from the same key, each row divided by its sum.val · nx
- Tolk.Setting.disable_fast_idivdisable_fast_idiv keeps code generation from replacing integer division by a constant with a multiplication and a shift, from DISABLE_FAST_IDIV.val · tolk
- Tolk.Setting.disallow_broadcastdisallow_broadcast makes an elementwise operation on operands of different shapes fail instead of broadcasting them, from DISALLOW_BROADCAST.val · tolk
- Tolk.Renderer.Compiler.disassembledisassemble c lib prints the binary lib as instructions on standard output, if c can.val · tolk
- Rig_pci.Page_table.discarddiscard t ~va n is unmap of whatever the n bytes from va map, none or part of them included, as a map that an exception ended part-way leaves them.val · rig
- Fehu.Space.Discretemodule · fehu
- Tolk.Helpers.DiskcachePersistent key-value tables.module · tolk
- Rig_metal_abi.dispatchThe type for the dispatches an indirect command buffer records.type · rig
- Norn.DistProbability distributions over tensors.module · norn
- Rig_pci.Machine.distanceThe type for how far apart two functions of a machine are, nearest first.type · rig
- Rig_pci.Machine.distancedistance m a b is how far apart the functions at a and b of m are, or None if either is unread.val · rig
- Ymir.Cosmology.distance_modulusdistance_modulus c ~observed z is 5 log10(DL/10 pc), a magnitude, with DL as luminosity_distance gives it.val · ymir
- Talon.Kit.distinctdistinct q is the first of each set of q's rows that are the same on every column, by key identity (null is one key, and NaN is one), in order:val · talon
- Kaun.Init.distributionThe type for variance-scaling distribution families.type · kaun
- Nx.divdiv a b is a / b: the IEEE 754 quotient on floats and complex numbers; on integers the quotient truncated toward zero, 0 for a divisor of 0, and a signed dtype's least value for that value by -1.val · nx
- Nx_wide.divdiv a b is a / b within 10u² relative.val · nx
- Talon.Expr.Temporal.divdiv a b is how many of b fit in a, as float64, for example div (diff ts signup) (span (Time.Span.days 1)).val · talon
- Tolk.Bigint.divdiv a b is a / b rounded toward zero.val · tolk
- Tolk.Ops.divdiv ~rounding x y divides integers by Op.t.Truncdiv or Op.t.Div when rounding; otherwise it multiplies x, as a float, by y's reciprocal, then rounds as asked.val · tolk
- Tolk.Ops.Elementwise.divdiv ~rounding x y divides integers by Op.t.Truncdiv or Op.t.Div when rounding; otherwise it multiplies x, as a float, by y's reciprocal, then rounds as asked.val · tolk
- Tolk.Ops.Upat.divdiv ~rounding x y divides integers by Op.t.Truncdiv or Op.t.Div when rounding; otherwise it multiplies x, as a float, by y's reciprocal, then rounds as asked.val · tolk
- Ymir_units.Quantity.divdiv a b is the quotient of the payloads, as Nx.div computes it, in Unit.(unit a / unit b).val · ymir
- Tolk.Shape.div_and_mod_symbolicdiv_and_mod_symbolic rewrites divisions and remainders.val · tolk
- Nx.div_sdiv_s x c is div x c.val · nx
- Norn.Diag.divergent_shiftdivergent_shift u s d is, for each element, the mean of the draws whose transition diverged minus the mean of all draws, in standard deviations of all draws.val · norn
- Tolk.Shape.divide_exactdivide_exact u d is u / d if it divides exactly.val · tolk
- Tolk.Shape.dividesdivides u n is u / n if u is known to be a multiple of n.val · tolk
- Tolk.Bigint.divisibledivisible a b is true iff b divides a: a = b * q for an integer q.val · tolk
- Tolk.Heuristic_common.divisibledivisible s n is true iff s is a known size that n divides.val · tolk
- Rig.Buffer.dmaThe type for how another PCI function of a buffer's machine, such as a network adapter, reaches its memory by DMA (dma).type · rig
- Rig.Buffer.dmadma b ~fn is how the PCI function at the bus address fn of b's machine, such as a network adapter, reaches the whole allocation b's memory lies in, which starts offset bytes before b's first byte, as the driver of the device whose memory it is describes it, or Error why where that driver cannot (Rig_edge.Driver.dma).val · rig
- Rig_amd.dmadma d r ~fn is how the PCI function at the bus address fn of d's machine, such as "0000:41:00.0", reaches all of r's bytes by DMA, or Error why if it cannot, why naming the cause, such as a memory BAR smaller than the GPU's memory.val · rig
- Rig_cuda.dmadma d r ~fn is how the PCI function at the bus address fn of d's machine, such as "0000:41:00.0", reaches all of r's bytes by DMA, or Error why if it cannot, why naming the cause, such as a memory BAR smaller than the GPU's memory.val · rig
- Rig_edge.dmaThe type for how another PCI function of a device's machine, such as a NIC, reaches a region's memory by DMA.type · rig
- Rig_edge.Driver.dmadma d r ~fn is how the PCI function at the bus address fn of d's machine, such as "0000:41:00.0", reaches all of r's bytes by DMA, or Error why if it cannot, why naming the cause, such as a memory BAR smaller than the GPU's memory.val · rig
- Rig_metal.dmadma d r ~fn is how the PCI function at the bus address fn of d's machine, such as "0000:41:00.0", reaches all of r's bytes by DMA, or Error why if it cannot, why naming the cause, such as a memory BAR smaller than the GPU's memory.val · rig
- Rig_nv.dmadma d r ~fn is how the PCI function at the bus address fn of d's machine, such as "0000:41:00.0", reaches all of r's bytes by DMA, or Error why if it cannot, why naming the cause, such as a memory BAR smaller than the GPU's memory.val · rig
- Rig_remote_proxy.dmadma d r ~fn is how the PCI function at the bus address fn of d's machine, such as "0000:41:00.0", reaches all of r's bytes by DMA, or Error why if it cannot, why naming the cause, such as a memory BAR smaller than the GPU's memory.val · rig
- Tolk.Setting.dmcdmc keeps code generation from merging loads, and stores, of consecutive elements into vector accesses (Coalesce.memory_coalescing), from DMC.val · tolk
- Quill.DocNotebook documents.module · quill
- Quill.Session.docdoc s is the current document of session s.val · quill
- Hugin_kit.Scale.domainThe type for domains, one case per kind, so let Floats (a, b) = domain s is exhaustive.type · hugin
- Hugin_kit.Scale.domaindomain s is the domain of s: the one set, or the default.val · hugin
- Nx.donatedonate x is a handle of x (Donation and copies).val · nx
- Nx.Ptree.donatedonate s x is x with each array given up by its owner: Nx.donate of each, in walk order.val · nx
- Nx_array.doordoor ~written ~read f x claims the memory of written for writing and of read for reading, every claim or none, runs f x, and releases the claims when f returns or raises, before an exception of f propagates.val · nx
- Rig_mlx5_abi.DoorbellA doorbell record holds a count that a process stores for the NIC in its memory.module · rig
- Jera.Ode.dopri5dopri5 is Dormand and Prince's (1980) method of order 5 with an embedded order 4: six evaluations per step, its last stage the next step's first.val · jera
- Hugin.dotdot ~x ~y () draws a symbol at the point of each row, in row order, filled with fill and outlined with stroke at the theme's outline width.val · hugin
- Nx.dotdot a b is the sum of the products of a's last axis with b's only axis, for a 1-d b, or its second to last otherwise.val · nx
- Nx.Ptree.dotdot s dt x y is the 0-d inner product of x and y at dt: from 0, the sum in walk order, over each array a of the vector and y's array b at its path, of vdot (cast dt a) (cast dt b) for floats and of the same over the real parts, then the imaginary parts, stored in dt, for complex arrays.val · nx
- Hugin_kit.Dash.dotteddotted is v [ 1.; 2.val · hugin
- Kaun_hf.download_filedownload_file ~file repo_id is the local path to file from the Hub repository repo_id (e.g.val · kaun
- Hugin.Size.dpidpi d is the density of d dots per inch, d /.val · hugin
- Kaun.Cache_index.draftdraft ~slots ~sees index is index whose tokens are a draft verified in one call, a tree of guesses for instance, with:val · kaun
- Hugin.drawdraw ~prev ~density l paints l at density device pixels per point.val · hugin
- Hugin.DrawingDrawings.module · hugin
- Norn.DrawsEqual-weight draws from chains.module · norn
- Rig_edge.DriverDrivers.module type · rig
- Rig_pci.DriverUser-space drivers of a machine's PCI functions.module · rig
- DriversThis page traces one submission from Rig.submit through a driver to the hardware and back.guide · rig
- Rig_remote_proxy.Link.dropdrop l id sends the release of the agent's object id after every message sent before it, and returns once it is sent, or immediately if the job has failed or its close has begun.val · rig
- Talon.Kit.dropdrop sel q is q without the columns that sel selects:val · talon
- Kaun.DropoutDropout with an explicit ~training flag and optional Nx.Rng key.module · kaun
- Nx.dstdst x is the discrete sine transform of type type_ (default Type_2) along each of axes: type II is y[k] = 2 Σ_j x[j] sin (π(k + 1)(2j + 1) / 2n), type I y[k] = 2 Σ_j x[j] sin (π(j + 1)(k + 1) / (n + 1)), type III y[k] = (-1)^k x[n-1] + 2 Σ_(j<n-1) x[j] sin (π(j + 1)(2k + 1) / 2n) and type IV y[k] = 2 Σ_j x[j] sin (π(2j + 1)(2k + 1) / 4n).val · nx
- Norn.Dist.dtypedtype d is the dtype of d's log densities, and of its values for a continuous d.val · norn
- Nx.dtypeThe type for dtypes whose elements OCaml reads as 'v and which store them as 's.type · nx
- Nx.dtypedtype x is x's element format.val · nx
- Nx_array.dtypedtype a is a's dtype.val · nx
- Nx_array.DtypeElement formats.module · nx
- Nx_io.Gguf.dtypeThe type for GGUF storage types.type · nx
- Nx_kernel.Prog.dtypedtype p i is the dtype of p's node i: its operand's for In, int64 for Coord, the one it names for Const and Op1, its operands' for a Binary kind, bool for a Compare kind, and its last two operands' for Op3.val · nx
- Nx_kernel.Prog.E.dtypedtype e is e's dtype.val · nx
- Tolk.DtypeData types, their constants and their promotion.module · tolk
- Tolk.Ops.dtypedtype u is u's data type.val · tolk
- Tolk.Ops.Upat.dtypedtype p is the first type p requires, or Dtype.t.Void.val · tolk
- Tolk.Ops.dtype_ofdtype_of op src arg is the data type of the node op of src with argument arg.val · tolk
- Nx_kernel.dtypesThe type for reading the dtypes of a collection of arrays of type 'c in place: arrays c arrays, dtype c i the dtype of the ith.type · nx
- Talon.Type.durationduration u is Duration in the unit u.val · talon
- Vega.Loss_scale.dynamicdynamic at is a fresh adaptive scale at at, whole on each of its devices, adjusted by step.val · vega
- Rig_nv_abi.Launch.dynamic_shareddynamic_shared l is the most dynamic shared memory a block of l may take beside its own (Qmd.t.shared): 100 KiB less the block's own shared memory, the kernel's and the driver's 1 KiB rounded up to 128 bytes.val · rig
e
- Nx_kernel.Prog.EPrograms written as values.module · nx
- Talon.Expr.eacheach sel { column } is the outputs column n x for each name n that sel selects, in order, where x reads the column n at its own type, extensions included: each Sel.all { column = (fun n x -> n := rows - count x) }.val · talon
- Ymir.EarthThe Earth's orientation and motion.module · ymir
- Norn.Diag.ebfmiebfmi s is each chain's energy Bayesian fraction of missing information, shape [chain]: the mean squared change of the energy between draws over the energy's variance.val · norn
- Ymir.Frame.ecliptic_j2000type · ymir
- Ymir.Frame.ecliptic_j2000ecliptic_j2000 is the IAU 2006 mean ecliptic and equinox of J2000.0, on ICRS: the obliquity ε₀ = 84381.406″ applied to the IAU 2006 frame bias at J2000.0.val · ymir
- Ecosystem overviewRaven's libraries share one data type: Nx.t, the n-dimensional array.manual
- Quill_markdown.EditTypora-style cursor-aware markdown block segmentation.module · quill
- Ymir.Wcs.editThe type for keyword edits.type · ymir
- Tolk.Op.Set.effectfuleffectful is the operations that may act beyond their value each time they run: Store, Call and Barrier.val · tolk
- Norn.Evidence.effective_groupseffective_groups z is K = (Σ_g W_g)² / Σ_g W_g², a scalar, W_g the posterior mass of the independent group g: nested sampling's lineages, tempering's chains.val · norn
- Nx.eigeig dt a is (w, v) in dt, the […; n] eigenvalues of the real or complex a and columns of unit norm v, matmul a v being v with each column scaled by its eigenvalue.val · nx
- Nx_cpu.eigeig ~w ~v a stores the eigenvalues of a, square, into w, and columns v of unit norm with a v = v diag w.val · nx
- Nx_kernel.S.eigeig ~w ~v a stores the eigenvalues of a, square, into w, and columns v of unit norm with a v = v diag w.val · nx
- Nx.eigheigh dt a is (w, v), the […; n] eigenvalues in dt of the Hermitian a, ascending, and orthonormal columns v, matmul a v being v with each column scaled by its eigenvalue.val · nx
- Nx_cpu.eigheigh ~w ~v a stores the eigenvalues of a, square and Hermitian as its lower triangle and the real parts of its diagonal name it, which alone are read, ascending into w, and orthonormal columns v with a v = v diag w.val · nx
- Nx_kernel.S.eigheigh ~w ~v a stores the eigenvalues of a, square and Hermitian as its lower triangle and the real parts of its diagonal name it, which alone are read, ascending into w, and orthonormal columns v with a v = v diag w.val · nx
- Nx_kernel.Shape.eigheigh a is the shapes of w and v of a, square: […; n] and a, for S.eigh and S.eig alike.val · nx
- Nx.eigvalseigvals dt a is eig's w of a, computed without its vectors, as svdvals is of svd, in an order of its own, which need not be w's.val · nx
- Nx.eigvalsheigvalsh dt a is eigh's w of a, reading the same part, computed without its vectors, as svdvals is of svd: ascending.val · nx
- Nx.einsumeinsum ~sizes p [x1; …; xn] is round (Σ x1 ⋯ xn) over p's summed names, for each index of p's result, at the operands' dtype, each xi read where its names take their indices, so that one that repeats a name reads its diagonal.val · nx
- Ymir_units.Codata.electron_g_factorelectron_g_factor r is ge, negative, dimensionless.val · ymir
- Ymir_units.Codata.electron_magnetic_momentelectron_magnetic_moment r is μe, negative, in J T-1.val · ymir
- Ymir_units.Codata.electron_masselectron_mass r is me in kg.val · ymir
- Ymir_units.Unit.electronvoltelectronvolt is elementary_charge times volt: 1.602176634·10-19 J.val · ymir
- Fehu.Space.Box.elementContinuous vector represented as a float32 tensor.type · fehu
- Fehu.Space.Dict.elementNamed fields with different space types.type · fehu
- Fehu.Space.Discrete.elementDiscrete action represented as a scalar int32 tensor.type · fehu
- Fehu.Space.Multi_binary.elementBinary vector for multi-label scenarios.type · fehu
- Fehu.Space.Multi_discrete.elementMultiple discrete choices with independent cardinalities.type · fehu
- Fehu.Space.Sequence.elementVariable-length homogeneous sequence.type · fehu
- Fehu.Space.Text.elementString space for textual observations or actions.type · fehu
- Fehu.Space.Tuple.elementFixed-length heterogeneous sequence in Value.t form.type · fehu
- Ymir_fits.Fits.Image.elementelement t is the stored format, the standard's unsigned offsets included.val · ymir
- Tolk.Ops.element_sizeelement_size x is the size in bytes of one of x's elements.val · tolk
- Ymir_units.Unit.elementary_chargeelementary_charge is e, 1.602176634·10-19 A s.val · ymir
- Tolk.Op.Set.elementwiseelementwise is alu, Cast and Bitcast: the operations applied element by element.val · tolk
- Tolk.Ops.ElementwiseElementwise operations.module type · tolk
- Rig_nv_abi.Cubin.elfelf c is the object c was read from, which lays out its sections.val · rig
- Ymir.Region.ellipseellipse p ~a ~b ~angle is the ellipse about the origin of p's plane with semi-axis a along the direction at angle from the plane's +y toward +x, the position angle east of north on the sky, and semi-axis b across it.val · ymir
- Kaun.EmbeddingToken-id to dense-vector lookup tables.module · kaun
- Rune.Probe.emitemit p v offers v to the innermost open collect of p.val · rune
- Brot.Encoding.emptyempty is the encoding with no tokens.val · brot
- Fehu.Info.emptyempty is the empty dictionary.val · fehu
- Hugin.View.emptyempty binds no key, so every key has its initial value.val · hugin
- Hugin_gg.Path.emptyempty is the path with no subpaths.val · hugin
- Hugin_vg.Picture.emptyempty paints nothing.val · hugin
- Quill.Doc.emptyempty () is a document with no cells.val · quill
- Talon.Record.emptyempty is the record cell with no fields.val · talon
- Tolk.Shape.emptyempty ~device shape dt is uninitialised storage of shape for whoever realizes the graph to bind.val · tolk
- Ymir_fits.Fits.Header.emptyempty has no record.val · ymir
- Tolk.Shape.empty_likeempty_like u is uninitialised storage of u's shape, type and placement.val · tolk
- Hugin_vg_vector.Vector.emsems f is how far, in ems, the outline of a glyph of f reaches from its origin in any direction.val · hugin
- Tolk.Setting.emulated_dtypesemulated_dtypes names the data types code generation emulates with other types, as if the target did not support them, from EMULATED_DTYPES: a ,-separated list, empty items dropped.val · tolk
- Tolk.Decomp_dtype.emulatesemulates r dt is true iff code generation for r's target emulates dt: r does not support it (Renderer.supported_dtypes) or the setting Setting.emulated_dtypes names it, an unsigned 64-bit integer as the signed one that emulates it.val · tolk
- Rig.Profile.enabledenabled () is true iff some profile is being taken.val · rig
- Brot.encodeencode t text is the encoding of text by t.val · brot
- Talon_arrow.encodeencode f q w runs q with Talon.Query.fold and writes its rows to w as f says, then Bytesrw.Bytes.Slice.eod.val · talon
- Talon_csv.encodeencode ?eod f q w runs q and writes its rows to w as the records of a file of format f, which decode f reads back to q's rows.val · talon
- Talon_json.encodeencode ?eod f q w runs q and writes its rows to w as the rows of a file of format f, which decode f reads back to q's rows.val · talon
- Talon_parquet.encodeencode f q w runs q with Talon.Query.fold and writes its rows to w as a Parquet file of format f, then Bytesrw.Bytes.Slice.eod.val · talon
- Brot.encode_batchencode_batch t texts is the encoding of each text in texts.val · brot
- Brot.encode_batch_idsencode_batch_ids t texts is (ids, lengths): the token ids of every text end to end in one buffer, and the number of ids each of them holds.val · brot
- Brot.encode_batch_pairsencode_batch_pairs t pairs encodes a batch of sentence pairs.val · brot
- Brot.encode_idsencode_ids t text is Encoding.ids (encode t text), without building the encoding when it can be avoided: the alignment metadata is what costs, and this asks for none of it.val · brot
- Nx_io.encode_pngencode_png ?dpi ?srgb x is the file save_png writes, without touching the file system.val · nx
- Nx_io.encode_txtencode_txt ?sep x is the text save_txt ?sep path x writes to path, without touching the file system.val · nx
- Brot.EncodingTokenization encodings.module · brot
- Rig_remote_abi.end_The type for one machine's end of a rail.type · rig
- Rig_remote_abi.end_end_ ~outbound ~inbound ~counts ~ready ~ready_fn is a provider's new rail end.val · rig
- Tolk.Ops.end_end_ u rs is the Op.t.End of the ranges rs around u, or u if rs is empty.val · tolk
- Tolk.Ops.Upat.end_val · tolk
- Rig_bin.Line.endedended r is true once read met the end of fd r.val · rig
- Munin.Run.ended_atended_at r is the time of r's end, if r had ended.val · munin
- Tolk.Ops.ended_rangesended_ranges u is the ranges that u closes: those among its range sources, those a Op.t.Barrier or an Op.t.After's effects close, the loop a Op.t.Backedge repeats.val · tolk
- Rig_bnxt.Qp.endpointThe type for what a peer needs to connect to a queue pair.type · rig
- Rig_bnxt.Qp.endpointendpoint qp is what a peer's queue pair connects to qp with.val · rig
- Rig_mlx5.Qp.endpointThe type for what a peer needs to connect to a queue pair.type · rig
- Rig_mlx5.Qp.endpointendpoint qp is what a peer's queue pair connects to qp with.val · rig
- Jera.Piecewise.endsThe type for a cubic spline's end conditions, which change it near the ends:type · jera
- Rig_nv_abi.Method.engineThe type for the engines a channel binds.type · rig
- Norn.EnsembleEnsemble sampling by the stretch move, without derivatives.module · norn
- Rig.Image.entryentry fn is the value by which compiled code refers to fn (the code of Rig_edge.Driver.entry), such as a kernel descriptor's address, a CUfunction, an MTLComputePipelineState or the address of a host function.val · rig
- Rig.Submission.entryentry fn ~at is a ref whose 8 bytes hold, when the work runs, the value by which the device's work refers to fn (Image.entry).val · rig
- Rig_amd.entryentry i f is Entry e with what the driver needs to launch i's function f, or Absent if i has no function f.val · rig
- Rig_cuda.entryentry i f is Entry e with what the driver needs to launch i's function f, or Absent if i has no function f.val · rig
- Rig_edge.Driver.entryentry i f is Entry e with what the driver needs to launch i's function f, or Absent if i has no function f.val · rig
- Rig_edge.entryThe type for a function of an image, as a driver names it.type · rig
- Rig_metal.entryentry i f is Entry e with what the driver needs to launch i's function f, or Absent if i has no function f.val · rig
- Rig_mlx5_abi.EntryA work entry is one operation of a queue pair's send ring.module · rig
- Rig_nv.entryentry i f is Entry e with what the driver needs to launch i's function f, or Absent if i has no function f.val · rig
- Rig_nv_abi.Gpfifo.entryentry addr ~offset ~words is the entry of the segment of words words at addr + offset, which is 4-byte aligned and below 2{^40}.val · rig
- Rig_remote_proxy.entryentry i f is Entry e with what the driver needs to launch i's function f, or Absent if i has no function f.val · rig
- Rig_edge.entry_answerThe type for a driver's answer when asked for a function of an image.type · rig
- Fehu.EnvReinforcement learning environments.module · fehu
- Environments and WrappersThis guide covers creating custom environments, composing wrappers, rendering, and running vectorized environments.guide · fehu
- Brot.eos_tokeneos_token t is t's end-of-sequence token, if any.val · brot
- Fehu.Collect.episodesepisodes env ~policy ~n_episodes () collects complete episodes, one trajectory per episode.val · fehu
- Hugin_kit.Time.epochepoch is 1970-01-01 00:00:00 UTC.val · hugin
- Talon.Join.eqeq l r matches the rows whose left column l and right column r are the same key.val · talon
- Tolk.Ops.Elementwise.eqval · tolk
- Tolk.Ops.eqval · tolk
- Tolk.Ops.Upat.eqval · tolk
- Fehu.Value.equalequal a b is true iff a and b are structurally equal.val · fehu
- Hugin.Coord.equalequal c c' is true iff c and c' are the same coordinate system with equal parameters.val · hugin
- Hugin.Drawing.equalequal d d' is true iff d and d' have equal warnings and Renderable.equal renderables.val · hugin
- Hugin.equalequal f f' is true iff f and f' are made by the same combinators from equal arguments: tensors and functions compared physically, string and float arrays element by element, the constant of a parameter by the equality of its role (Role.param), and the values of the lower libraries by their equal.val · hugin
- Hugin.Layout.equalequal l l' is true iff l and l' lay out Resolved.equal figures in Theme.equal themes and have equal sizes, panels of equal ids, boxes and coordinate systems, equal frozen ticks, guides, headers and titles of equal contents in equal boxes, and equal warnings.val · hugin
- Hugin.Resolved.equalequal r r' is true iff r and r' resolve Hugin.equal figures under View.equal views to the same scopes, Scale.equal fitted scales and equal warnings.val · hugin
- Hugin.Size.equalequal s s' is true iff s and s' fix the same length to equal values.val · hugin
- Hugin.Theme.equalequal th th' is true iff th and th' have equal colours, sizes, locales, schemes (Scheme.equal) and lists of faces (Font.equal).val · hugin
- Hugin.View.equalequal view view' is true iff view and view' bind the same keys to equal values, floats compared by Float.equal.val · hugin
- Hugin_font.Font.equalequal f f' is String.equal (bytes f) (bytes f').val · hugin
- Hugin_font.Run.equalequal r r' is true iff r and r' have Font.equal fonts, and equal sizes, texts, glyphs, positions and clusters.val · hugin
- Hugin_gg.Affine.equalequal m n is true iff the coefficients of m and n are pairwise equal.val · hugin
- Hugin_gg.Box2.equalequal a b is true iff a and b have equal corners.val · hugin
- Hugin_gg.Color.equalequal c c' is true iff c and c' have equal components and alpha.val · hugin
- Hugin_gg.P2.equalequal p q is true iff p and q have equal coordinates.val · hugin
- Hugin_gg.Path.equalequal p q is true iff p and q were built from the same segments with equal points, non-finite ones included, a polyline counting as its point-by-point segments.val · hugin
- Hugin_gg.Stroke.equalequal s s' is true iff s and s' have equal widths, caps, joins, miter limits, dash patterns and dash offsets.val · hugin
- Hugin_gg_kit.Pgon2.equalequal p p' is true iff p and p' have Ring2.equal rings in the same order.val · hugin
- Hugin_gg_kit.Ring2.equalequal r r' is true iff r and r' have equal points in the same order from the same first point.val · hugin
- Hugin_kit.Curve.equalequal c c' is true iff c and c' are the same curve.val · hugin
- Hugin_kit.Dash.equalequal d d' is true iff d and d' have equal lengths, compared by Float.equal.val · hugin
- Hugin_kit.Locale.equalequal l l' is true iff l and l' have equal strings and group sizes.val · hugin
- Hugin_kit.Number.equalequal f f' is true iff f and f' have equal notations, precisions, trimming and grouping.val · hugin
- Hugin_kit.Scale.equalequal s s' is true iff s and s' set the same properties to equal values: floats by Float.equal, instants by Time.equal, colours by Hugin_gg.Color.equal, schemes by Scheme.equal, arrays element by element, symbols by Symbol.equal, dash patterns by Dash.equal, and the functions of custom transforms physically.val · hugin
- Hugin_kit.Scheme.equalequal s s' is true iff s and s' are the same named scheme or are made by the same constructor, ramp or palette, from colours that are pairwise Hugin_gg.Color.equal, and both or neither are reversed: a ramp through the table of viridis is not equal to viridis.val · hugin
- Hugin_kit.Symbol.equalequal s s' is true iff s and s' are the same symbol.val · hugin
- Hugin_kit.Ticks.equalequal t t' is true iff t and t' have equal major ticks, minor positions and notes, positions compared by Float.equal.val · hugin
- Hugin_kit.Time.equalequal t t' is true iff t and t' are the same instant.val · hugin
- Hugin_text.Text.equalequal t t' is true iff t and t' have the same characters with the same styles, size factors and shifts compared by Float.equal and colours by Hugin_gg.Color.equal.val · hugin
- Hugin_text.Text.Layout.equalequal l l' is true iff l and l' have equal boxes and fold equal runs (Hugin_font.Run.equal) at equal origins with equal colours, in the same order.val · hugin
- Hugin_vg.Picture.equalequal p p' is true iff p and p' have the same structure: the same cases, with fields compared by the equal function of their type (Hugin_gg.Path.equal, Hugin_font.Run.equal, which compares fonts by their bytes, Hugin_gg.Color.equal, Nx.Ptree.Path.equal for tag ids and so on), numbers by Float.equal, and image tensors by shape and elements, which reads them, so that equal raises what reading a tensor raises.val · hugin
- Hugin_vg.Renderable.equalequal r r' is true iff r and r' have equal widths and heights and Picture.equal pictures.val · hugin
- Norn.Support.equalequal s s' is true iff s and s' are the same constructor with equal arguments, floats compared with Float.equal.val · norn
- Nx.equalequal a b is a = b.val · nx
- Nx.Placement.equalequal p q is true iff p and q are of one set and every device holds the same window, and the same term of a partial value, at p and at q, whatever their brands.val · nx
- Nx.Ptree.Path.equalequal p q is true iff p and q have equal steps: the one field "a.b" and the field a then b differ, though they print alike.val · nx
- Nx.Ptree.Skeleton.equalequal k k' is true iff k and k' have equal visits, their paths compared by Path.equal.val · nx
- Nx_array.Dtype.equalequal dt dt' is true iff dt and dt' are the same dtype.val · nx
- Nx_array.Layout.equalequal l l' is true iff l and l' have the same shape, strides and offset: by the canonical form, iff they have one shape and map every index to the same position.val · nx
- Nx_kernel.Prog.E.equalval · nx
- Nx_wide.equalequal a b is a = b, exactly.val · nx
- Rig.equalequal d d' is d == d': each device is one value for the life of the process.val · rig
- Talon.equalequal t0 t1 is true iff t0 and t1 have equal schemas and the same keys row by row, by key identity (Type.compare_value, null being one more key), whatever their batches.val · talon
- Talon.Query.equalequal q0 q1 is true iff q0 and q1 are the same plan: the same steps with equal arguments, expressions compared by their identity, tables by key identity row by row, and sources physically, with equal requests (see pp).val · talon
- Talon.Schema.equalequal s0 s1 is true iff s0 and s1 have the same names with equal types (Type.equal) in the same order.val · talon
- Talon.Time.Date.equalequal t0 t1 is true iff t0 and t1 are the same day.val · talon
- Talon.Time.equalequal t0 t1 is true iff t0 and t1 are the same reading of one clock.val · talon
- Talon.Time.Span.equalequal d0 d1 is true iff d0 and d1 are the same span.val · talon
- Talon.Type.equalequal t0 t1 is true iff t0 and t1 are the same type: the same constructor with equal arguments.val · talon
- Talon.Tz.equalequal z0 z1 is true iff z0 and z1 have the same transitions and the same rule.val · talon
- Tolk.Bigint.equalequal n0 n1 is compare n0 n1 = 0.val · tolk
- Tolk.Dtype.equalequal d0 d1 is true iff d0 and d1 are the same data type.val · tolk
- Tolk.Op.equalequal o0 o1 is true iff o0 and o1 are the same operation.val · tolk
- Tolk.Op.Set.equalequal s0 s1 is true iff s0 and s1 have the same operations.val · tolk
- Tolk.Ops.Axis_type.equalequal a0 a1 is true iff a0 and a1 are the same axis type.val · tolk
- Tolk.Ops.equalequal u0 u1 is u0 == u1: structural equality, since nodes are hash-consed.val · tolk
- Tolk.Opt.equalequal o0 o1 is true iff o0 and o1 are the same optimisation.val · tolk
- Tolk.Shape.Sint.equalequal s0 s1 is true iff both are the same integer or the same node.val · tolk
- Tolk.Tc.equalequal tc0 tc1 is true iff tc0 and tc1 have the same types and fragments.val · tolk
- Ymir_fits.Fits.Header.equalequal h h' is true iff h and h' have the same records.val · ymir
- Ymir_units.Unit.equalequal u w is true iff u and w are the same unit, which is iff to_string u = to_string w.val · ymir
- Tolk.Ops.equal_argequal_arg a0 a1 is structural equality, with constants compared by Dtype.equal_const and nodes by ==.val · tolk
- Tolk.Tc.equal_bitequal_bit b0 b1 is true iff b0 and b1 are the same bit.val · tolk
- Tolk.Dtype.equal_constequal_const c0 c1 is true iff c0 and c1 are the same constant: the same constructor with the same payload.val · tolk
- Hugin_kit.Time.equal_intervalequal_interval i i' is true iff i and i' have the same boundaries at every offset.val · hugin
- Hugin_kit.Scale.equal_kindequal_kind k k' is Some Type.Equal iff k and k' are the same kind.val · hugin
- Nx.equal_sequal_s x c is equal x c.val · nx
- Fehu.Space.equal_specequal_spec a b is true iff a and b describe structurally identical spaces.val · fehu
- Nx_array.Dtype.equal_witnessequal_witness dt dt' is Some Equal iff equal dt dt'.val · nx
- Tolk.Shape.equalsequals u v is true iff the constant u holds the number v.val · tolk
- Ymir.Earth.eraera e is the Earth rotation angle in [0, 2π) radians, 2π (0.7790572732640 + 1.00273781191135448 T_u), T_u the UT1 Julian date less 2451545.0 (IERS Conventions 2010, 5.14), within 4e-15 radians of the exact value.val · ymir
- Nx.erferf x is the error function 2/√π ∫₀ˣ e{^-t²} dt: -1.val · nx
- Nx_kernel.Prog.E.erfval · nx
- Nx.erfcerfc x is 1 - erf x, to its own relative precision where erf x rounds to 1.: erfc 10.val · nx
- Nx.erfinverfinv p is the x with erf x = p, on [-1., 1.]: -inf at -1., -0.val · nx
- Sowilo.erodeerode ~kernel img is img with each pixel replaced by the least value over its neighbourhood.val · sowilo
- Hugin_font.Font.errorThe type for loading errors.type · hugin
- Jera.Solution.errorerror s is each lane's error estimate, of the answer's structure, as its solve defines it: an estimate, never a bound.val · jera
- Norn.Evidence.errorerror z is the estimate's standard error in ln Z, a scalar.val · norn
- Rig_mlx5_abi.Completion.errorThe type for the reasons an entry failed.type · rig
- Talon.ErrorFailures found in data and in the environment.module · talon
- Quill_book.Render.escape_htmlescape_html s escapes HTML special characters in s.val · quill
- Norn.Weighted.essess w is Kish's effective sample size (Σ w)² / Σ w², a scalar from 1 to n.val · norn
- Norn.Diag.ess_bulkess_bulk u d is the effective sample size of each element's normal scores, by Geyer's initial monotone sequence over the autocorrelations of the split chains.val · norn
- Norn.Diag.ess_tailess_tail u d is the smaller of the effective sample sizes of the indicators of each element being at most its 5% and its 95% quantile.val · norn
- Tolk.Realize.estimate_uopestimate_uop call is the cost of call, seen through its Op.t.Afters: a program's kernel estimates and, for a copy, its bytes as both memory touched and bytes loaded and stored.val · tolk
- Tolk.Ops.estimatesThe type for the cost estimates of a kernel.type · tolk
- Tolk.Renderer.EstimatesCost estimates of kernels.module · tolk
- Jera.Ode.eulereuler is the explicit Euler method: order 1, one field evaluation per step.val · jera
- Jera.Sde.euler_maruyamaeuler_maruyama is the Euler–Maruyama method, Itô, for any noise: strong order 1/2.val · jera
- Fehu.EvalPolicy evaluation.module · fehu
- Jera.Grid.evaleval g x is g at the points x of shape q @ [d]: of shape q @ value.val · jera
- Jera.Piecewise.evaleval p x is p at each point of x, of shape q: each leaf of shape q @ value.val · jera
- Quill.EvalBatch evaluation.module · quill
- Vega.Schedule.evaleval s step is the schedule's value at the host counter step, as an OCaml float.val · vega
- Jera.Piecewise.eval_ateval_at p i u is the series of piece i at the local coordinate u in [−1, 1], i and u of one shape q: each leaf of shape q @ value.val · jera
- Rune.eval_shapeeval_shape a b f x is f x's structure with each array a stand-in of its dtype, shape and placement, as jit traces it, computing nothing: Nx.shape, Nx.dtype and Nx.placement work on it, and an operation on it or a read raises Invalid_argument.val · rune
- Jera.Solution.evaluationsevaluations s is each lane's count of the points at which the problem's function was evaluated, of the lanes' shape.val · jera
- Jera.Ode.eventevent y m ~tol ~budget f ~event ~t0 ~t1 y0 is the solve from y0 at t0 toward t1 that ends at the first crossing of an event: the time, the state there, and the flat index of the component of event that crossed.val · jera
- Quill.Kernel.eventtype · quill
- Rig.Profile.eventThe type for profile events.type · rig
- Rig_mlx5.eventThe type for events of a NIC.type · rig
- Kaun.Cache_index.everyevery m index is index read in blocks of m positions, for a layer that keeps one entry per block (a compressed key, a summary): column j is the slot the index's table of blocks of m positions names for block j, positions j * m to j * m + m - 1, and it stands at the block's last position.val · kaun
- Hugin_vg_instances.Instances.everywhereeverywhere is the box from -1e15 to 1e15 along both axes.val · hugin
- Norn.EvidenceEvidence: the normalising constant of a posterior, with its draws.module · norn
- Norn.Nested.evidenceevidence u k s is the evidence of the dead points of s and of its live points added as dead, k simulating the shrinkage sequences of its error and resampling its lineages.val · norn
- Norn.Smc.evidenceevidence s is the evidence of the tempered density at s.beta, the posterior's at β = 1, with the particles as its sample.val · norn
- Nx.ewmaewma ~alpha x is the exponentially weighted moving average of x along axis: y₀ = x₀, then yᵢ = (1 - alpha) yᵢ₋₁ + alpha xᵢ; without axis, of its elements in C order, in x's shape.val · nx
- Ymir_units.Unit.exaexa u is 1018 u.val · ymir
- Hugin_float_text.Float_print.exactexact x is the decimal of fewest digits that reads back as x, to nearest, ties to even, the nearest to x of those (Float_text.shortest), written as g writes it at the larger of 15 and its number of digits: 0.1, 0.30000000000000004, 1e+15 and 5e-324.val · hugin
- Munin_float_text.Float_print.exactexact x is the decimal of fewest digits that reads back as x, to nearest, ties to even, the nearest to x of those (Float_text.shortest), written as g writes it at the larger of 15 and its number of digits: 0.1, 0.30000000000000004, 1e+15 and 5e-324.val · munin
- Nx_array.Dtype.exactexact a b is true iff every value of a is a value of b, so that a store of one into b (Stores) gives the same number, the same infinity or a NaN.val · nx
- Tolk.Ops.exactexact dt vs is true iff dt is not a committed integer type, or each of vs is one of its values.val · tolk
- Munin_float_text.Float_print.exact_floatexact_float x is exact x with .0 added when it has neither a fraction nor an exponent, as 1.0 and -0.0, so that a reader that tells integers from floats by their text reads a float.val · munin
- Tolk.Ops.exact_powerexact_power b e is true iff b ** e, for integers b and e, can be computed exactly in exec_alu when nothing truncates: a base of 0, 1 or -1, a negative e, or a power of at most max_exact_shift bits.val · tolk
- Rig.Claim.exclusiveexclusive c b is true iff c holds b's memory exclusive, so that the caller may consume it and write the buffer consume returns in place (Claims).val · rig
- Brot.execThe type for executors, which run the tasks of a batch on the program's domains.type · brot
- Tolk.Ops.exec_aluexec_alu ~truncate_output op dt args computes the arithmetic operation op on the scalars args, one lane of a vector operation, exactly, then truncates the result to dt (Dtype.truncate) if truncate_output (default true).val · tolk
- Quill.Kernel.executeexecute k c runs the code cell c on k's backend if c's language is language k.val · quill
- Execution ModesQuill provides five execution modes: the interactive REPL, the terminal notebook UI, a web frontend, batch execution (with optional watch), and clean.guide · quill
- Hugin_vg_instances.Instances.existsexists f p is true iff f holds of p or of a picture within it.val · hugin
- Norn.Bij.expexp maps u to exp u, onto (0, inf).val · norn
- Nx.expexp x is e{^x}.val · nx
- Nx_kernel.Prog.E.expval · nx
- Nx.exp2exp2 x is 2{^x}.val · nx
- Nx_kernel.Prog.E.exp2val · nx
- Tolk.Ops.Elementwise.exp2val · tolk
- Tolk.Ops.exp2val · tolk
- Tolk.Ops.Upat.exp2val · tolk
- Tolk.Shape.expandexpand u shape broadcasts u to shape; a size of -1 keeps u's.val · tolk
- Tolk.Prepare.expand_bitcastexpand_bitcast bc is the Op.t.Bitcast bc of a value to a type of another size as shifts and casts of unsigned integers, lane by lane along its last axis: to a wider type, each group of narrow lanes is shifted into place and summed; to a narrower one, each wide lane is shifted and cast into each of its narrow lanes, the least significant first.val · tolk
- Tolk.Setting.expand_ssaexpand_ssa makes C-style code generation (Cstyle) give every value a local variable, from EXPAND_SSA.val · tolk
- Nx_array.expectexpect dt (Any a) is a at type dt if its dtype is dt.val · nx
- Norn.Evidence.expectationexpectation u f z is the posterior expectation of f, a map of one draw to a tensor, under z's sample, and its standard error, element by element.val · norn
- Munin.Run.experimentexperiment r is the experiment r is an attempt at.val · munin
- Nx.expm1expm1 x is e{^x} - 1, accurate near 0: -0.val · nx
- Nx_kernel.Prog.E.expm1val · nx
- Tolk.Transcendental.exponent_biasexponent_bias dt is the bias of the exponent field of the float dt: 2e-1 - 1 for e exponent bits (Dtype.finfo), and one more for Dtype.t.Fp8e4m3fnuz and Dtype.t.Fp8e5m2fnuz.val · tolk
- Norn.Dist.exponentialexponential ~rate has density rate exp (-rate x) on (0, inf).val · norn
- Nx.Rng.exponentialexponential dt shape is exponential draws of rate 1, by inverting the distribution at 1 - u for a uniform u: finite, and never negative.val · nx
- Nx.Rng.exponential_atexponential_at p dt shape is exponential dt shape at p.val · nx
- Vega.Schedule.exponential_decayexponential_decay ~init_value ~decay_rate ~decay_steps is init_value * decay_rate{^ (step / decay_steps)}.val · vega
- Rig.Buffer.exportexport b k is b's bytes as elements of k for good, once every use of b's memory is done, for code that keeps them, such as a library's array.val · rig
- Brot.export_tiktokenexport_tiktoken t ~merges_path ~vocab_path exports t's BPE merges and vocabulary in tiktoken-compatible format.val · brot
- Talon.ExprExpressions: typed computations over the columns of a frame.module · talon
- Tolk.Ops.exprexpr u is the name of the parameter or buffer u.val · tolk
- Expressions and FramesAn expression computes values from the columns of a frame: the ordered rows that a verb, or an enclosing expression, gives it.guide · talon
- Talon.ExtExtension declarations.module · talon
- Talon.Type.extThe type that extension cells read as.type · talon
- Talon.Type.extext ~name ?metadata storage is Ext named name, with metadata (defaults to "") and stored as storage.val · talon
- Jera.Piecewise.extendextend e p is p defined at every finite point: `Hold takes the value at the nearest end, and `Polynomial continues the end pieces' series.val · jera
- Kaun.Attention.Cache.extendextend index cache k v is (k', v', cache'): cache with the call's keys k and values v stored, both of shape [| batch; kv_heads; seq; head_dim |], and what the call's tokens attend over, of shape [| batch; kv_heads; context; head_dim |].val · kaun
- Kaun.Cache_index.extendextend index values pool is (seen, pool'), with pool of shape [| slots; ...val · kaun
- Hugin.Mark.extentextent rows `X is (a, b), the normalised interval each row covers along x, from a.(i) to b.(i) in either order:val · hugin
- Nx.extractextract c x is the elements of x where c is not zero, in C order, c and x flattened: a NaN is not zero, -0.val · nx
- Talon.Expr.Str.extractextract ~group p a is group group of the first match of p in a, 0, the whole match, by default.val · talon
- Tolk.Bigint.extractextract n off len is the non-negative integer of bits off to off + len - 1 of n.val · tolk
- Nx.extract_patchesextract_patches ~kernel_size ~stride ~dilation ~padding x is the windows of x over its last K = Array.length kernel_size axes, padded with zeros by padding's (lo, hi) pairs.val · nx
- Nx_kernel.extremetype · nx
- Nx.eyeeye ~m ~k dt n is the n × m host value (m defaults to n) whose element at row i and column j is one where j - i = k (default 0) and zero elsewhere: k > 0 is a diagonal above the main one.val · nx
- Nx.eye_ateye_at ~m ~k p dt n is eye ~m ~k dt n at p.val · nx
f
- Hugin_float_text.Float_print.ff decimals x is x as C's %.*f writes it, with decimals digits after the point, at least 0: 0.12 for 0.125 at two.val · hugin
- Munin_float_text.Float_print.ff decimals x is x as C's %.*f writes it, with decimals digits after the point, at least 0: 0.12 for 0.125 at two.val · munin
- Norn_float_text.Float_print.ff decimals x is x as C's %.*f writes it, with decimals digits after the point, at least 0: 0.12 for 0.125 at two.val · norn
- Tolk.Ops.Upat.ff p o matches the node o of the one source p matches: f p Cast matches a cast of p to any type.val · tolk
- Kaun.Metric.f1f1 ?average predictions labels is the harmonic mean of precision and recall per class — 2 tp / (2 tp + fp + fn) — combined according to average, which defaults to `Macro.val · kaun
- Tolk.Decomp_dtype.f2ff2f ~sat v fr to_ converts between a narrow float and Dtype.t.Float32, v being the bits of a float of type fr, as the unsigned integer of its width:val · tolk
- Norn_model.factorfactor l adds l to the log likelihood.val · norn
- Norn.Dist.factorsfactors d x is the log density of each independent unit of x: one per element for the elementwise families, per row for dirichlet and mvn, per matrix for lkj_cholesky, one for sorted, mixture and hmm, with iid's axes in front.val · norn
- Nx.factorsThe type for how much of a matrix's unitary factors a factorisation gives: k columns of the left one and k rows of the right one, or the whole squares, of m and of n.type · nx
- Nx_kernel.factorsThe type for how much of a matrix's unitary factors Shape.qr and Shape.svd shape: k columns of the left one and k rows of the right one, or the square ones, of m and n.type · nx
- Rig_amd.factsfacts d is d's facts.val · rig
- Rig_cuda.factsfacts d is d's facts.val · rig
- Rig_edge.Driver.factsfacts d is d's facts.val · rig
- Rig_edge.factsThe type for the facts of a device.type · rig
- Rig_metal.factsfacts d is d's facts.val · rig
- Rig_nv.factsfacts d is d's facts.val · rig
- Rig_remote_proxy.factsfacts d is d's facts.val · rig
- Rig.failfail why fails the process.val · rig
- Rig_remote_proxy.Link.failfail j why fails j with the root cause why, unless it has failed or closed.val · rig
- Rig_pci.Function.failedfailed f is why f can no longer be reached, if it cannot: its machine has failed (Machine.failed), or f has left the bus and its vendor ID reads 0xffff.val · rig
- Rig_pci.Machine.failedfailed m is the reason m can no longer be reached, if it cannot.val · rig
- Rig_remote.failedfailed j is Some why if j has failed, why the root cause this process recorded, and None otherwise.val · rig
- Rig_remote_proxy.Link.failedfailed j is Some why iff wait j ~ms:0 is Failed why.val · rig
- Rig.failurefailure () is the process's first failure, if any: the first loss of a device that no close or fork ended, as Lost prints it, or why of fail if that came first.val · rig
- Hugin_font.Font.familyfamily f is the typographic family name of f (name ID 16) if it has one and its family name (name ID 1) otherwise, or "" if f names no family.val · hugin
- Norn.Dist.familyfamily d is d's family, such as "normal".val · norn
- Ymir_units.Unit.faradfarad is A2 s4 kg-1 m-2.val · ymir
- Ymir_units.Unit.faradayfaraday is F, avogadro * elementary_charge.val · ymir
- Tolk.Decomp_op.fast_idivfast_idiv r x d is x divided by the positive constant d, rounding towards zero, as a multiplication and a right shift: (x * m) >> s, for x's values from 0 to its upper bound.val · tolk
- Rig_amd.FaultFault why reports that the device faulted.exception · rig
- Rig_cuda.FaultFault why reports that the device faulted.exception · rig
- Rig_edge.Driver.FaultFault why reports that the device faulted.exception · rig
- Rig_edge.Io.FaultFault why reports that the device failed, such as a closed connection.exception · rig
- Rig_metal.FaultFault why reports that the device faulted.exception · rig
- Rig_nv.FaultFault why reports that the device faulted.exception · rig
- Rig_remote_proxy.FaultFault why reports that the device faulted.exception · rig
- Rig_bin.Line.fdfd r is r's descriptor.val · rig
- Rig_bin.Proc.fd_numberfd_number fd is fd's number, as RIG_REMOTE_REPORT names it.val · rig
- Nx_kernel.Prog.E.fdivval · nx
- Tolk.Bigint.fdivfdiv a b is a / b rounded toward negative infinity.val · tolk
- fehuFehu is a reinforcement learning environment toolkit for OCaml.library · fehu
- FehuReinforcement learning environments and utilities.module · fehu
- Fehu vs. Gymnasium -- a practical comparisonThis guide explains how Fehu's reinforcement learning API relates to Python's Gymnasium (and Stable Baselines3 for collection/buffer/GAE), focusing on:guide · fehu
- Fehu_envsBuilt-in environments for testing and learning.module · fehu
- Ymir_units.Unit.femtofemto u is 10-15 u.val · ymir
- Rig_bin.Firmware.fetchfetch list dir fills dir, or Rig_pci.Firmware.directory if None, with the images of list, a driver's firmware.tsv as Images holds it: a line per image, its path, digest and URL separated by tabs, after comment lines starting with #.val · rig
- Nx.fftfft x is the forward transform of x along axes.val · nx
- Nx.fft_normThe type for the scaling of a transform and its inverse.type · nx
- Nx.fftfreqfftfreq dt ~d n is the host value of the frequency of each of fft's n bins for samples d apart (default 1.), in cycles per unit of d: i / (d n) for i in [0, (n - 1) / 2], then (i - n) / (d n) for the rest, the negative frequencies.val · nx
- Nx.fftfreq_atfftfreq_at p dt ~d n is fftfreq dt ~d n at p, an _at form as Creation states.val · nx
- Nx.fftshiftfftshift x is x rolled along each of axes (default the last) by n / 2 places toward the end, n its extent there: the zero frequency of a transform moves to the middle, after the negative frequencies.val · nx
- Jera.Ode.fieldThe type for fields: f t y is the derivative of y at the scalar time t, a value of y's structure, dtypes and shapes.type · jera
- Nx.Ptree.Walk.fieldfield c name walk x is walk c' x, c' at c's path then Field name.val · nx
- Talon.Expr.fieldfield k n r is the field n of r read as k, null where r is null or the field is: field Kind.float "flux" band.val · talon
- Talon.Expr.Temporal.fieldThe type for calendar fields:type · talon
- Talon.Expr.Temporal.fieldfield f a is the field f of a, as int64: field `Year orderdate.val · talon
- Talon.Record.fieldfield k name r is the field name of r read as k: Some v if it holds v, and None if it is null.val · talon
- Hugin_gg_kit.Field2Scalar fields sampled on grids, and their isolines and isobands.module · hugin
- Hugin.Size.figurefigure w h is a figure w points wide and h points high.val · hugin
- Quill_markdown.figuresThe type for how a notebook stores its images.type · quill
- Munin.History.fileThe type for files a run logged.type · munin
- Talon_arrow.filefile ?format path is the source of the file or stream at path, which it maps (Rig_disk.of_file), read as format, or as sniff gives it when format is absent.val · talon
- Talon_csv.filefile ?format ?nulls path is the source of the CSV file path, named csv "path", whose errors name path: as format says, or as sniff ?nulls infers from the file when format is absent.val · talon
- Talon_json.filefile ?format path is the source of the JSON file path, named json "path", whose errors name path: as format says, or as sniff infers from the file when format is absent.val · talon
- Talon_parquet.filefile ?format path is the source of the file path, which it maps (Rig_disk.of_file), read as format, or as sniff gives it when format is absent.val · talon
- Munin.History.filesfiles h is the files the run logged that are not orphans, in log order.val · munin
- Hugin.Role.fillfill is the colour that fills shapes and glyphs.val · hugin
- Hugin_vg.Picture.fillfill ~rule c q paints the area of q under rule with c.val · hugin
- Rig.Buffer.fillfill b c sets every byte of b to c and returns once they are set: submit_fill b c, then a wait for its point, holding its read claim until it returns.val · rig
- Rig.Submission.fillfill ~refs f ~params calls the C function at f from the driver's submit (The C functions) with the queue's context, the host address of its argument and the work's value.val · rig
- Rig_nv_abi.Method.fillfill c ~dst v n writes n components of c's size from dst on the copy engine, each holding the low bytes of the 32-bit v, least significant first, in one launch that reads no memory.val · rig
- Rig_nv_abi.Template.fillfill value t is t.bytes with each patch filled with value slot.val · rig
- Rig_pci.Window.fillfill w off n c stores n bytes c from byte off of w.val · rig
- Talon.Kit.fill_backwardfill_backward x is x, or where x is null its next value in the frame:val · talon
- Talon.Kit.fill_forwardfill_forward x is x, or where x is null its last earlier value in the frame:val · talon
- Hugin_kit.Symbol.filledfilled is the symbols designed to be filled: circle, cross, diamond, square, star, triangle and wye.val · hugin
- Rig_amd_abi.Counter.filterfilter g ~vmid is the 32-bit filter word that counts the waves of VMID vmid alone: 1 lor (1 lsl (vmid + 1)).val · rig
- Talon.Query.filterfilter p q is the rows of q on which p is true, in order, so a row on which p is null is dropped.val · talon
- Talon.Source.Pred.filterfilter s p is the filter that p is over a table of the columns s describes: it is true, false or null where p is, so Query.filter of it keeps the rows that pass p.val · talon
- Sowilo.filter2dfilter2d kernel img applies a custom 2D convolution kernel to img.val · sowilo
- Fehu.Info.findfind k t is the value bound to k in t, if any.val · fehu
- Nx_io.Archive.findfind name a is a's entry name, if any.val · nx
- Quill.Doc.findfind id d is the cell with identifier id in d, or None.val · quill
- Rig_pci.Firmware.findfind dirs name ~digest is the first file dir/name, for dir in dirs in order, whose digest is digest.val · rig
- Talon.Schema.findfind s name is the type of the column name of s, or None if s has no such column.val · talon
- Talon.Tz.findfind db name is the zone name of db, read and decoded from the file name under db's directory, and named name (name).val · talon
- Ymir_fits.Fits.Header.findfind is get with Ok None for an absent or undefined keyword.val · ymir
- Talon.Expr.Str.find_allfind_all p a is the matches of p in a, in order, as a list[string], found as split finds them: the empty list where a has no match.val · talon
- Quill.Doc.find_displayfind_display id d is the first cell of d holding a display with id id, or None.val · quill
- Fehu.Info.find_exnfind_exn k t is the value bound to k in t.val · fehu
- Quill.Doc.find_indexfind_index id d is the zero-based index of cell id in d.val · quill
- Tolk.Helpers.find_llvm_objdumpfind_llvm_objdump () is the llvm-objdump to run: Homebrew's on macOS, and elsewhere the first of ROCm's and those named llvm-objdump-21, llvm-objdump-20 and llvm-objdump on PATH.val · tolk
- Norn.Summary.findingThe type for findings.type · norn
- Norn.Summary.findingsfindings s is s's findings, the elements' in row order, then the transitions'.val · norn
- Tolk.C.findlibfindlib ~extra_paths name paths is the file of the library name, the first of:val · tolk
- Ymir_units.Codata.fine_structurefine_structure r is α, the fine-structure constant, dimensionless.val · ymir
- Hugin_vg_vector.Vector.finerfiner k d is k decimals more than d, at most 17.val · hugin
- Tolk.Dtype.finfofinfo dt is (exponent, mantissa), the number of exponent bits and of explicit mantissa bits of dt.val · tolk
- Munin.Session.finishfinish s stops s's samplers, waiting for a reading in progress, writes the run's end with status, syncs the log, writes the run's statistics hint (Statistics of an ended run) and releases the lock.val · munin
- Quill.Session.finish_executionfinish_execution id ~success s marks cell id as idle and increments its execution count.val · quill
- Rig_pci.Function.finish_releasesfinish_releases () gives back each function of this process whose release has begun and that is not given back, such as one whose give-back an exception ended part-way: what it had not given back, and nothing twice.val · rig
- Munin.Metric.finiteThe type for the statistics of a series' finite samples.type · munin
- Rig_bin.Firmwarerig firmware: a directory filled with a driver's pinned firmware images.module · rig
- Rig_bin.Help.firmwarefirmware is rig firmware's page.val · rig
- Rig_pci.FirmwareFirmware images, verified by digest.module · rig
- Rig_gpus.firstfirst ~firmware () is the first GPU of paths that opens, or Error why, where why names each path's refusal and each GPU's.val · rig
- Talon.Expr.firstfirst a is a's first non-null value in frame order.val · talon
- Hugin_kit.Scale.fitfit observed s is s with nice and zero unset and its domain set to:val · hugin
- Norn.Gaussian.fitThe type for fits of Gaussians to a density.type · norn
- Nx_array.Layout.fitsfits s is true iff s is a layout's shape: at most max_rank extents, none negative, and those other than 0 multiplying to at most max_numel.val · nx
- Tolk.Heuristic_host.fitsfits ?upcast ?unroll k is whether k stays within the host's budget once upcast splits n lanes off an axis, given as (axis, n), and unroll unrolls an axis by n lanes, 0 for whole.val · tolk
- Tolk.Shape.fitsfits x v is true unless x is of a signed integer type that does not hold v: the rules that make a constant from others make it only where it fits (symbolic_simple, symbolic).val · tolk
- Ymir_fits.Fitsmodule · ymir
- Tolk.Bigint.fits_intfits_int n is true iff n is in int's range.val · tolk
- Tolk.Ops.fits_no_intfits_no_int u is true iff u's bounds are known and neither Dtype.t.Int64 nor Dtype.t.Uint64 holds them: no machine integer can carry every value of u.val · tolk
- Tolk.Shape.fits_startfits_start ~param start is true iff storage that starts at start (storage_phase) starts where a parameter that assumes param takes it: known modulo at least the parameter's alignment, at its phase.val · tolk
- Norn_model.fixfix sel v m is m whose site at sel is the constant v, with no density term.val · norn
- Jera.Quad.fixedfixed r f range is r's sum for the integral of f over each element of range, of range's shape.val · jera
- Talon.Tz.fixedfixed o is the zone whose offset is o seconds east of UTC at every instant.val · talon
- Ymir.Frame.fixedThe type of a frame at a constant orientation from ICRS.type · ymir
- Brot.Pre_tokenizer.fixed_lengthfixed_length n cuts the text into pretokens of n characters.val · brot
- Ymir.Frame.fk5_j2000type · ymir
- Ymir.Frame.fk5_j2000fk5_j2000 is FK5, mean equator and equinox of J2000.0, oriented at epoch J2000.0: the rotation from ICRS by the vector (−19.9, −9.1, +22.9) mas of Mignard and Froeschlé (2000), 31.7 mas in all.val · ymir
- Tolk.Compiler_cpu.flagsflags ~version arch is the arguments that make a Clang whose --version states version, such as "Ubuntu clang version 18.1.3 (1ubuntu1)", compile for arch as clang reads it: its target, processor, features and the processor it tunes for, in that order.val · tolk
- Hugin_gg.Path.flattenflatten ~tolerance m ~move ~line ~close acc p is fold over transform m p, with every cubic replaced by line segments that stay within tolerance of it.val · hugin
- Nx.flattenflatten ~start_dim ~end_dim x is x with its axes start_dim to end_dim, both included, merged into one.val · nx
- Nx.Ptree.flattenflatten s x is x's arrays in walk order and x's skeleton.val · nx
- Tolk.Shape.flattenflatten ~start ~stop u merges the axes from start (default 0) to stop (default -1), included.val · tolk
- Nx.flipflip ~axes x is x with the order of elements along each axis of axes reversed; without axes, along every axis.val · nx
- Tolk.Shape.flipflip u axes reverses axes of u, negative axes counting from the end.val · tolk
- Nx_io.Archive.floatfloat ~shape dt name a is a's entry name at dt.val · nx
- Talon.Col.floatfloat name is v Kind.float name, which binds every float type.val · talon
- Talon.Expr.floatfloat x is the float x.val · talon
- Talon.Kind.floatfloat reads float16, float32 and float64 columns.val · talon
- Tolk.Ops.Elementwise.O.floatfloat x is the weak literal x.val · tolk
- Tolk.Ops.floatfloat ~dtype x is const ~dtype (`Float x).val · tolk
- Tolk.Ops.O.floatfloat x is the weak literal x.val · tolk
- Tolk.Ops.Upat.floatfloat x is the weak literal x.val · tolk
- Tolk.Ops.Upat.O.floatfloat x is the weak literal x.val · tolk
- Tolk.Setting.floatfloat ~reach key default is a setting whose initial value is the number the variable key holds, or default if key is unset.val · tolk
- Ymir_fits.Fits.Value.floatfloat is integer or real text (FITS 4.0 §4.2.4), read correctly rounded.val · ymir
- Nx.float16val · nx
- Talon.Type.float16float16 is Float16.val · talon
- Nx.float16_elttype · nx
- Nx_array.Dtype.float16_eltIEEE 754 binary16.type · nx
- Nx.float16_ttype · nx
- Nx.float32val · nx
- Rig.Submission.Run.float32val · rig
- Talon.Type.float32float32 is Float32.val · talon
- Nx.float32_elttype · nx
- Nx_array.Dtype.float32_eltIEEE 754 binary32.type · nx
- Nx.float32_ttype · nx
- Nx.float4_e2m1fnval · nx
- Nx.float4_e2m1fn_elttype · nx
- Nx_array.Dtype.float4_e2m1fn_eltfloat4 E2M1FN: 2 exponent bits, 1 fraction bit, exponent bias 1, no infinity and no NaN.type · nx
- Nx.float4_e2m1fn_ttype · nx
- Nx.float64val · nx
- Rig.Submission.Run.float64val · rig
- Talon.Type.float64float64 is Float64.val · talon
- Nx.float64_elttype · nx
- Nx_array.Dtype.float64_eltIEEE 754 binary64.type · nx
- Nx.float64_ttype · nx
- Nx.float8_e4m3fnval · nx
- Nx.float8_e4m3fn_elttype · nx
- Nx_array.Dtype.float8_e4m3fn_eltfloat8 E4M3FN: 4 exponent bits, 3 fraction bits, exponent bias 7, no infinity; S.1111.111 is NaN.type · nx
- Nx.float8_e4m3fn_ttype · nx
- Nx.float8_e5m2val · nx
- Nx.float8_e5m2_elttype · nx
- Nx_array.Dtype.float8_e5m2_eltfloat8 E5M2: 5 exponent bits, 2 fraction bits, exponent bias 15, with infinities and NaNs as in IEEE 754.type · nx
- Nx.float8_e5m2_ttype · nx
- Nx_kernel.float_carrierfloat_carrier dt is carrier of the float dt, the compute type in which a float kind of dt computes before it rounds once to dt (Prog).val · nx
- Nx_array.Dtype.float_formatThe type for the facts of a float format.type · nx
- Nx_array.Dtype.float_formatfloat_format dt is the facts of dt's format.val · nx
- Hugin_float_text.Float_printFloats as decimal text, the same on every system.module · hugin
- Munin_float_text.Float_printFloats as decimal text, the same on every system.module · munin
- Norn_float_text.Float_printFloats to a number of digits, the same text on every system.module · norn
- Munin_float_text.Float_readDecimal text as floats, the same on every system.module · munin
- Hugin_float_text.Float_textDecimal digits of floats, the same on every system.module · hugin
- Munin_float_text.Float_textDecimal digits of floats, the same on every system.module · munin
- Norn_float_text.Float_textDecimal digits of floats, the same on every system.module · norn
- Hugin.floatsfloats ~scale ~title a is the quantities of the elements of a, one datum per element, with scale and title as for num: its shape is [|Array.length a|], and NaN and infinities are missing.val · hugin
- Tolk.Dtype.floatsfloats is the floats of known width: fp8s, then Float16, Bfloat16, Float32 and Float64.val · tolk
- Hugin_kit.Time.floorfloor ~tz_offset_s i t is the latest boundary of i at tz_offset_s that is not after t.val · hugin
- Nx.floorfloor x rounds toward negative infinity.val · nx
- Nx_kernel.Prog.E.floorval · nx
- Nx_wide.floorfloor w is the greatest integer at most w, exactly.val · nx
- Talon.Expr.Temporal.floorfloor step a is the first value of the period of step that holds a, a date or a datetime.val · talon
- Tolk.Ops.Elementwise.floorfloor x is trunc x, less one where that is above x.val · tolk
- Tolk.Ops.floorfloor x is trunc x, less one where that is above x.val · tolk
- Tolk.Ops.Upat.floorfloor x is trunc x, less one where that is above x.val · tolk
- Tolk.Helpers.floordivfloordiv x y is x / y rounded down, and 0 if y is 0.val · tolk
- Tolk.Helpers.floormodfloormod x y is x - floordiv x y * y: the remainder of the division rounded down, with the sign of y.val · tolk
- Jera.Split.flowThe type for flows: flow h s is the state s moved by its part over the duration h: a scalar, or in step a tensor that broadcasts against the state's leaves.type · jera
- Rig_pci.Window.flushflush w makes the stores made through w before it reach the function before any access made after it, through any window.val · rig
- Nx.fmafma a b c is a * b + c rounded once.val · nx
- Nx_kernel.Prog.E.fmaval · nx
- Tolk.Ops.Elementwise.fmodfmod x y is the remainder of div ~rounding:`Trunc x y.val · tolk
- Tolk.Ops.fmodfmod x y is the remainder of div ~rounding:`Trunc x y.val · tolk
- Tolk.Ops.Upat.fmodfmod x y is the remainder of div ~rounding:`Trunc x y.val · tolk
- Tolk.Dtype.fmtfmt dt is the format character of a dt value in a byte string: '?' for Bool, 'b', 'h', 'i' and 'q' for the signed integers, 'B', 'H', 'I' and 'Q' for the unsigned ones, and 'e', 'f' and 'd' for Float16, Float32 and Float64.val · tolk
- Kaun.FnActivation functions (relu, gelu, silu, ...) and sampling masks.module · kaun
- Nx.Ptree.fnThe type for signatures of functions of type 'f on the set 'd.type · nx
- Rig_pci.Machine.fnThe type for the operations on a function that a transport took.type · rig
- Talon.Expr.fnThe type for elementwise nx functions of one argument that preserve its dtype, such as Nx.exp, on host values as columns hold.type · talon
- Talon.Expr.fn2The type for elementwise nx functions of two arguments of one dtype that preserve it, such as Nx.atan2.type · talon
- Hugin_gg.Path.foldfold ~move ~line ~cubic ~close acc p folds over what p draws, in drawing order: move acc x y starts a subpath, line acc x y and cubic acc c1x c1y c2x c2y x y extend it and close acc closes it.val · hugin
- Hugin_text.Text.Layout.foldfold f acc l folds f over the runs of l in text order: f acc c at r for the run r with origin at and colour c, None if its characters have none.val · hugin
- Nx.Ptree.foldfold s f x acc is f pn an (… (f p1 a1 acc)), a1, …, an x's arrays and p1, …, pn their paths.val · nx
- Nx.Ptree.Payload.foldfold (module U) f x acc is f pn an (… (f p1 a1 acc)), a1, …, an x's payloads and p1, …, pn their paths, in walk order.val · nx
- Nx_cpu.foldfold p ~dst x stores into dst the adjoint of a Padded load by p of dst's shape, where x has the shape that load gives.val · nx
- Nx_kernel.S.foldfold p ~dst x stores into dst the adjoint of a Padded load by p of dst's shape, where x has the shape that load gives.val · nx
- Talon.Query.foldfold q ~init f runs q: it reads q's sources, computes q's rows in batches and folds f over them, in order, from init.val · talon
- Tolk.Ops.Nodes.foldfold f s acc folds f over the nodes of s in the order they joined it.val · tolk
- Tolk.Ops.Pattern_matcher.foldfold rules is the matcher of the rules rules () returns, which it calls on its first rewrite as v does, and whose results are not the nodes they match.val · tolk
- Nx.Rng.fold_infold_in k i is the key of k indexed by i: distinct i give independent keys.val · nx
- Nx.Rng.fold_in_tensorfold_in_tensor k i is fold_in k of the indices held in i, for an index known only as data.val · nx
- Hugin_vg_instances.Instances.fold_strokesfold_strokes f m p acc folds f over the strokes of p, in order, each with m composed with the transforms above it within p.val · hugin
- Talon.Kit.foldsfolds ~seed ~by k q is q with a last column "fold", an int64 in [0;k - 1], that gives every row of a group of by one fold, by key identity (null is one group and NaN one).val · talon
- Munin.History.followfollow h is h with the whole lines its log gained since folded in, or, with limit, up to the first whole line that ends past limit bytes after what h read.val · munin
- Hugin_font.FontOpenType fonts.module · hugin
- Hugin_font.Run.fontfont r is the font of r.val · hugin
- Hugin.Theme.fontsfonts th is the faces of th.val · hugin
- Fehu.Render.Pixel.formatThe type for pixel formats.type · fehu
- Hugin_float_text.Float_text.formatThe type for binary floating-point formats.type · hugin
- Munin_float_text.Float_text.formatThe type for binary floating-point formats.type · munin
- Norn_float_text.Float_text.formatThe type for binary floating-point formats.type · norn
- Rig_pci.Page_table.formatThe type for a vendor's page-table format and the access to the GPU's memory that holds the tables.type · rig
- Talon.Expr.Temporal.formatformat fmt a writes a, a date, datetime or clock, in the format fmt, as parse reads it: a datetime as its clock reads, a clock on 1970-01-01; %f writes nine digits and %z writes Z.val · talon
- Talon_arrow.formatThe type for Arrow formats: the framing, compression and record-batch size encode writes with, the schema's custom metadata, and each column's name, Arrow type, custom metadata and the talon type it reads as.type · talon
- Talon_arrow.formatformat ?framing ?compression ?batch ?metadata s is the format of what encode writes from a query of schema s.val · talon
- Talon_csv.formatThe type for CSV formats: a dialect, and the columns of the file's records, in order, each with its name and the type its fields read as.type · talon
- Talon_csv.formatformat ?sep ?quote ?header ?nulls columns is the format of a file whose records hold columns, in order, each a name and the type its fields read as, with:val · talon
- Talon_json.formatThe type for JSON formats: a framing and the file's columns, in order, each with its name and the type its values read as.type · talon
- Talon_json.formatformat ?framing columns is the format of a file whose rows hold columns, in order, each a name and the type its values read as.val · talon
- Talon_parquet.formatThe type for Parquet formats: the columns of a file, in schema order, each with its name, its Parquet type and the talon type it reads as, and the codec and row-group size encode writes with.type · talon
- Talon_parquet.formatformat ?compression ?row_group s is the format of a file that encode writes from a query of schema s.val · talon
- FormatsFiles enter a query as sources: data whose columns are known when the source is built and whose rows arrive in batches when the query runs.guide · talon
- Norn.Bij.forwardforward b u is (x, ld): the value x that b maps the coordinates u to, and log |det J| of the map at u, one element per unit.val · norn
- Rig_pci.Machine.forwardsforwards m ~provider ~client is true iff the bus of m carries the requests of the function at client to a BAR of the function at provider, as Memory.reaches states Linux's p2pdma decides, for a process that reads every bridge's ACS controls.val · rig
- Tolk.Dtype.fp8_fnuzfp8_fnuz is [Fp8e4m3fnuz; Fp8e5m2fnuz].val · tolk
- Tolk.Dtype.fp8_ocpfp8_ocp is [Fp8e4m3; Fp8e5m2].val · tolk
- Tolk.Dtype.fp8sfp8s is fp8_ocp @ fp8_fnuz.val · tolk
- Tolk.Tc.frag_coordsfrag_coords tc is, for A, B and C in turn, the coordinates of the element that each lane holds at each element index: (row, column) for A and C, (k, column) for B.val · tolk
- Tolk.Tc.fragmentThe type for fragments: where an operand's elements are held.type · tolk
- Hugin.frameframe () outlines the domain of each panel with stroke, the ink by default, at the theme's outline width: under theme th it draws what rect ~stroke:(const (Theme.ink th)) () draws.val · hugin
- Rig_program.frameThe type for a run's frame: its inputs, by index, the first words of its ints, and where the run reports each step's point.type · rig
- Ymir.Direction.frameframe d is d's frame.val · ymir
- Ymir.FrameCelestial frames.module · ymir
- Ymir.Healpix.frameframe h is h's frame.val · ymir
- Talon_arrow.framingtype · talon
- Talon_json.framingtype · talon
- Rig.Buffer.freefree ~why b tries to return b's memory early, and is false, changing nothing, if b does not span its memory (spans), another claim holds it, or it is outside the claims (Claims).val · rig
- Rig_amd.freefree d r gives back r, which is a region alloc made, a mapping map_peer or map_host made, or d's word of the driver's own (Rig_edge.facts).val · rig
- Rig_cuda.freefree d r gives back r, which is a region alloc made, a mapping map_peer or map_host made, or d's word of the driver's own (Rig_edge.facts).val · rig
- Rig_edge.Driver.freefree d r gives back r, which is a region alloc made, a mapping map_peer or map_host made, or d's word of the driver's own (Rig_edge.facts).val · rig
- Rig_edge.Io.freefree d r gives r back.val · rig
- Rig_metal.freefree d r gives back r, which is a region alloc made, a mapping map_peer or map_host made, or d's word of the driver's own (Rig_edge.facts).val · rig
- Rig_nv.freefree d r gives back r, which is a region alloc made, a mapping map_peer or map_host made, or d's word of the driver's own (Rig_edge.facts).val · rig
- Rig_pci.Memory.freefree m r gives back what alloc, map_host or map_peer gave.val · rig
- Rig_pci.Page_table.freefree t x m unmaps m, which alloc gave with the hold x, and flushes.val · rig
- Rig_pci.Space.freefree x gives back x: its range goes back with the last hold on it.val · rig
- Rig_remote_proxy.freefree d r gives back r, which is a region alloc made, a mapping map_peer or map_host made, or d's word of the driver's own (Rig_edge.facts).val · rig
- Tolk.Support_memory.Tlsf_allocator.freefree a x frees the block that alloc returned at x.val · tolk
- Rig.free_cachefree_cache d first returns d's staging memory that no copy holds (Copying), then moves d's collected memory (Reclamation), then returns d's cached memory to its driver once the work using it is done.val · rig
- Rig_pci.Function.free_dmafree_dma f w frees w, and f reaches it no more.val · rig
- Kaun.Rope.frequenciesfrequencies t is a copy of t's inverse frequencies, in pair order.val · kaun
- Rig_remote_proxy.Link.freshfresh () is an id the process gave no object before: the controller names each object it makes on an agent by one (Wire.request).val · rig
- Quill.Cell.fresh_idfresh_id () is a fresh unique identifier.val · quill
- Tolk.Transcendental.frexpfrexp v is (m, e) with |v| = |m| * 2e and 0.5 <= |m| < 1, for a normal v.val · tolk
- Jera.Quad.Range.fromfrom a is the half-line from a to +∞, at unit scale: a rule's nodes spread over distances near 1 from a.val · jera
- Brot.from_filefrom_file path is a tokenizer loaded from a HuggingFace tokenizer.json file.val · brot
- Brot.from_model_filefrom_model_file ~vocab () loads a tokenizer from HuggingFace model files.val · brot
- Norn_model.from_priorfrom_prior m ~n k is n prior draws of the coordinates, stacked on a leading axis.val · norn
- Tolk.Dtype.from_storage_scalarfrom_storage_scalar dt s is the value stored by s, the inverse of to_storage_scalar: the `Float that the low bits of s encode for Bfloat16 and the 8-bit floats, and s itself otherwise.val · tolk
- Nx.fullfull dt s v is the host value of shape s whose every element is v, a float stored as Dtype.of_float says.val · nx
- Tolk.Spec.fullfull is the rules of the forms between stages (ends of loops over any integers, any ordering, any load or store), then tensor's and program's rules, in that order.val · tolk
- Nx.full_atfull_at p dt s v is full dt s v at p.val · nx
- Nx.full_likefull_like x v is full (dtype x) (shape x) v.val · nx
- Nx_kernel.full_precisionfull_precision dt is true iff dt is float32, float64, complex64 or complex128, the dtypes the matrix routines, S.cholesky to S.solve_triangular, take.val · nx
- Tolk.Codegen.full_rewrite_to_sinkfull_rewrite_to_sink ~optimize ~beam ast ren is the kernel ast lowered for ren, ready to be linearized.val · tolk
- Tolk.Postrange.Scheduler.full_shapefull_shape k is the size of each of k's axes, simplified.val · tolk
- Rig.Image.funcThe type for a function of a loaded image.type · rig
- Rig_pci.FunctionPCI functions the process has taken.module · rig
- Tolk.Ops.function_namefunction_name k is k.name as an identifier (Helpers.to_function_name).val · tolk
- Rig_pci.Machine.functionsfunctions m is the PCI functions of m, in bus order (compare_address).val · rig
- Brot.Decoder.fusefuse is a collapsing decoder that concatenates all tokens into a single string with no delimiter.val · brot
- Hugin.Role.fxfx is the column of facet panels.val · hugin
- Hugin.Role.fyfy is the row of facet panels.val · hugin
g
- Hugin_float_text.Float_print.gg ~precision x is x as C's %.*g writes it, with precision significant digits, 6 by default and at least 1: positional for a decimal exponent from -4 to below precision, otherwise with an exponent of at least two digits, without trailing zeros, as 135.562, 1e+16 and 1.5e-05.val · hugin
- Hugin_gg.Color.gg c is the encoded green component of c.val · hugin
- Munin_float_text.Float_print.gg ~precision x is x as C's %.*g writes it, with precision significant digits, 6 by default and at least 1: positional for a decimal exponent from -4 to below precision, otherwise with an exponent of at least two digits, without trailing zeros, as 135.562, 1e+16 and 1.5e-05.val · munin
- Norn_float_text.Float_print.gg p x is x as C's %.*g writes it, with p significant digits, at least 1: positional for a decimal exponent from -4 to below p, otherwise with an exponent of at least two digits, without trailing zeros, as 135.562, 1e+16 and 1.5e-05 at six.val · norn
- Fehu.GaeGeneralized Advantage Estimation.module · fehu
- Ymir.Frame.galactictype · ymir
- Ymir.Frame.galacticgalactic is Galactic coordinates as the Hipparcos Catalogue defines them on ICRS: the north Galactic pole at α = 192.85948°, δ = 27.12825°, and the north celestial pole at l = 122.93192°.val · ymir
- Norn.Dist.gammagamma ~concentration ~rate has density proportional to x^(α - 1) exp (-rate x) on (0, inf), α the concentration.val · norn
- Nx.Rng.gammagamma a is gamma draws of concentration a and unit rate, elementwise: Marsaglia and Tsang's rejection, below a concentration of 1 through Gamma(a) = Gamma(a + 1) U{^1/a}.val · nx
- Nx.gammaincgammainc a x is the regularised lower incomplete gamma function P(a, x) = (1/Γ(a)) ∫₀ˣ t{^a-1} e{^-t} dt, the distribution function at x of the gamma distribution of shape a and unit rate, for a, x ≥ 0.val · nx
- Nx.gammainccgammaincc a x is Q(a, x) = 1 - P(a, x) (gammainc), the upper tail, to its own relative precision where P rounds to 1.: gammaincc 1.val · nx
- Nx.gammainccinvgammainccinv a q is the x with gammaincc a x = q, as gammaincinv from the upper tail: inf at q = 0., 0.val · nx
- Nx.gammaincinvgammaincinv a p is the x with gammainc a x = p, the gamma distribution's quantile, for a ≥ 0 and p in [0., 1.].val · nx
- Ymir_units.Unit.gas_constantgas_constant is R, avogadro * boltzmann.val · ymir
- Tolk.Ops.gate_kernel_sinkgate_kernel_sink u is false for a linear program and a kernel's sink, so walks gated by it do not enter kernels.val · tolk
- Tolk.GaterMoving gates from indices to memory accesses.module · tolk
- Nx_cpu.gathergather ~axis ~dst idx x stores into dst at each index j x's element at j with axis axis replaced by idx's element at j.val · nx
- Nx_kernel.S.gathergather ~axis ~dst idx x stores into dst at each index j x's element at j with axis axis replaced by idx's element at j.val · nx
- Nx_kernel.Shape.gathergather ~axis idx x is the shape of a gather of x at positions of shape idx: idx, which has x's rank and extents but along axis.val · nx
- Nx_cpu.gather_rangesgather_ranges ~dst ~starts ~ends x packs runs of x along its axis a = rank starts - 1 into dst.val · nx
- Nx_kernel.S.gather_rangesgather_ranges ~dst ~starts ~ends x packs runs of x along its axis a = rank starts - 1 into dst.val · nx
- Nx_kernel.Shape.gather_rangesgather_ranges ~size starts ends x is the shape of a range gather's destination: x with size elements along axis a = rank starts - 1.val · nx
- Jera.Quad.Rule.gaussgauss n is the n-point Gauss–Legendre rule, exact for polynomials of degree 2n − 1.val · jera
- Norn.GaussianGaussians over structures, diagonal plus low rank.module · norn
- Sowilo.gaussian_blurgaussian_blur ~sigma img applies isotropic Gaussian blur using separable convolution.val · sowilo
- Munin.Store.gcgc s removes the blobs of s that no version names, the .-named work that crashed writers left under blobs/ and artifacts/, and what a delete_run cut short left under runs/, and is the number of entries it removed.val · munin
- Tolk.Bigint.gcdgcd a b is the greatest common divisor of a and b, never negative: gcd a zero is abs a.val · tolk
- Tolk.Heuristic_common.gcdgcd a b is the greatest common divisor of a and b.val · tolk
- Tolk.Shape.gcdgcd us is a common divisor of us: their common factors times the greatest common divisor of their constant coefficients, a scalar constant whatever the shape of us.val · tolk
- Tolk.Heuristic_common.gcd_withgcd_with a n is the greatest common divisor of n with the positive a.val · tolk
- Ymir.Frame.gcrstype · ymir
- Ymir.Frame.gcrsgcrs is the Geocentric Celestial Reference System: ICRS's axes at the Earth's centre.val · ymir
- Talon.Join.gege l r is lt, as Expr.(>=) compares.val · talon
- Tolk.Ops.Elementwise.gege x y is logical_not (lt x y).val · tolk
- Tolk.Ops.gege x y is logical_not (lt x y).val · tolk
- Tolk.Ops.Upat.gege x y is logical_not (lt x y).val · tolk
- Kaun.Fn.gelugelu x is the exact Gaussian error linear unit x * Φ(x), computed as 0.5 * x * (1 + erf(x / sqrt 2)) where Φ is the standard normal CDF.val · kaun
- Kaun.Fn.gelu_approxgelu_approx x is the tanh approximation of gelu: 0.5 * x * (1 + tanh(sqrt(2/π) * (x + 0.044715 * x³))).val · kaun
- Ymir.Frame.geocentricThe type of a frame whose directions are as seen from the Earth's centre.type · ymir
- Nx.geomspacegeomspace dt ~endpoint start stop n is the host value of n values in geometric progression from start to stop, stop included where endpoint (default true): the value at i is start (stop / start)^(i / (n - 1)), or ^(i / n) without endpoint.val · nx
- Nx.geomspace_atgeomspace_at p dt ~endpoint start stop n is geomspace dt ~endpoint start stop n at p.val · nx
- Tolk.Bigint.geqgeq n0 n1 is n0 >= n1.val · tolk
- Hugin.Mark.getget rows r is Some vs, the value of each row for the role r in its range, positions normalised and a constant repeated, or None if the mark binds no channel to r.val · hugin
- Hugin.View.getget k view is the value bound to k in view, or the initial value of k if view binds none or binds the name of k with another sort.val · hugin
- Jera.Solution.getget s is the answer if every lane converged.val · jera
- Nx.getget p x is slice (List.map (fun i -> I i) p) x, a view.val · nx
- Nx_array.getget a i is a's element at index i.val · nx
- Rig.Local.getget k d is k's value on d.val · rig
- Rig_nv.getget p (at, n) is the field of n bytes at byte at of p, in the host's byte order, as the RM lays out its parameters: unsigned for 1, 2 and 4 bytes, and for 8 the value Int64.to_int gives, its top bit dropped.val · rig
- Tolk.Helpers.Diskcache.getget ~table key is the value of key in table, if any.val · tolk
- Ymir_fits.Fits.getget ~ver name hdus is the HDU whose EXTNAME is name and whose EXTVER is ver, an absent EXTVER being 1.val · ymir
- Ymir_fits.Fits.Header.getget v k h is the value of keyword k read with v.val · ymir
- Rig_pci.Window.get32get32 w off is the unsigned 32-bit word at byte off of w.val · rig
- Rig_pci.Window.get64get64 w off is the 64-bit word at byte off of w.val · rig
- Rig_pci.Window.get8get8 w off is the byte at off of w.val · rig
- Tolk.Realize.get_call_arg_uopsget_call_arg_uops call is the buffer arguments of call: its arguments without its scalar ones, the variables it binds and the values it passes.val · tolk
- Tolk.Realize.get_call_kernel_argsget_call_kernel_args call prg is what call passes for each parameter of the program prg (Shape.kernel_params), in the order the program takes them: for a buffer, call's argument in its slot; for a scalar parameter that a slot names, call's argument in that slot, an expression of the loops around call; for a named variable, the variable call binds to its name, whose value holds where a run does not bind the name, or else the variable, at slot -1, that a run binds by name.val · tolk
- Tolk.Realize.get_call_nameget_call_name ~var_vals call bufs is how diagnostics name call, whose buffers are bufs: a program's kernel name, as its kernel gives it, or, for a copy, "copy S, D <- E" in yellow, where S is the size copied with the variables of var_vals (default []) replaced by their values, and D and E the first seven characters of the destination's and the source's devices.val · tolk
- Tolk.Realize.get_call_outs_insget_call_outs_ins call is the positions among call's arguments of the buffers it writes and of those it reads: a program's (Ops.program_info.outs and Ops.program_info.ins), ([0], [1]) for a copy, and ([], []) for any other call.val · tolk
- Tolk.Ops.get_idxget_idx u is u without its validity condition.val · tolk
- Tolk.Search.get_kernel_actionsget_kernel_actions ~max_up k is the kernels that one action makes of k: (a, k') for an action a of actions () and k' a copy of k it was applied to, in the order of actions.val · tolk
- Tolk.Rangeify.get_kernel_graphget_kernel_graph sink is the kernel graph of the prepared tensor graph sink (Prepare.prepare_rangeify).val · tolk
- Talon.Error.get_okget_ok r is v if r is Ok v.val · talon
- Tolk.Postrange.Scheduler.get_optimized_astget_optimized_ast ~name_override k is k's kernel with its argument listing the applied optimisations and marking it optimised (optimized).val · tolk
- Tolk.Helpers.get_single_elementget_single_element l is the element of l.val · tolk
- Tolk.Ops.get_validget_valid u is u's validity condition: true unless u is `Invalid.val · tolk
- Getting StartedThis guide covers installation, data types, array creation, slicing, broadcasting, and basic operations.guide · nx
- Getting StartedThis guide walks through the whole kaun pattern once: define a model as a record, write the training step yourself, run the loop, evaluate.guide · kaun
- Getting StartedThis guide shows you how to describe parameters, initialize an optimizer's state, and run optimization steps.guide · vega
- Getting StartedThis guide covers the basics: encoding text to token IDs, decoding back to text, configuring the pipeline, and training tokenizers from scratch.guide · brot
- Getting StartedThis page builds a table, queries it, reads files, and shows what happens when a query is wrong.guide · talon
- Getting StartedThis page draws a first figure, explains how a mark turns tensors into rows, saves figures to files, shows them in a notebook, and reads the errors and warnings Hugin reports.guide · hugin
- Getting StartedThis guide covers the REPL, creating a notebook, executing it in different modes, and viewing results.guide · quill
- Getting StartedThis guide covers the basics: creating environments, running the step loop, understanding spaces, and using the built-in environments.guide · fehu
- Getting StartedThis guide covers loading images, understanding image conventions, building your first processing pipeline, and saving results.guide · sowilo
- Nx_io.GgufGGUF files.module · nx
- Ymir_units.Unit.gigagiga u is 109 u.val · ymir
- Munin.Provenance.gitThe type for the state of a git work tree.type · munin
- Kaun.Init.glorot_normalglorot_normal samples a truncated normal with target variance 2 / (fan_in + fan_out).val · kaun
- Kaun.Init.glorot_uniformglorot_uniform samples from U(-limit, limit) with limit = sqrt (6 / (fan_in + fan_out)).val · kaun
- Hugin_font.Font.glyphThe type for glyph ids.type · hugin
- Hugin_font.Font.glyphglyph f u is the glyph f's character map gives u, and 0 if it gives none or gives a glyph f does not have.val · hugin
- Hugin_font.Run.glyphglyph r i is the glyph at index i of r.val · hugin
- Hugin_font.Font.glyph_countglyph_count f is the number of glyphs of f, at least 1.val · hugin
- Hugin_vg.Picture.glyphsglyphs c at r paints the glyphs of the run r with c, the origin of r at at.val · hugin
- Jera.Linear.gmresgmres ~restart ~rel ~budget ~precondition is restarted GMRES from u = 0, preconditioned on the right by precondition, for any non-singular a.val · jera
- Ymir.Earth.gmstgmst e is the IAU 2006 Greenwich mean sidereal time in [0, 2π) radians: era plus a polynomial of TT (IERS Conventions 2010, 5.32), within 4e-15 radians of the exact value.val · ymir
- Hugin_kit.Scheme.gnbugnbu is Brewer's GnBu, green to blue.val · hugin
- Ymir.Transform.gnomonicgnomonic c is the same with Tan: great circles map to lines.val · ymir
- Munin.History.goalgoal h k is the goal of k's latest metric record, if it gave one.val · munin
- Munin.Metric.goalThe type for goals: whether lower or higher values are better.type · munin
- Rig_nv_abi.GpfifoRing entries.module · rig
- Rig_amd_abi.GpuAMD GPUs, as their formats depend on them.module · rig
- Rig_amd_amdgpu.gpugpu i is the opener of GPU i, named "AMD" for 0 and "AMD:i" otherwise.val · rig
- Rig_amd_pci.gpugpu ~machine ~firmware i is the opener of GPU i of machine (defaults to Rig_pci.Machine.this), named "AMD-PCI" for 0 and "AMD-PCI:i" otherwise.val · rig
- Rig_cuda.gpugpu i is the opener of GPU i, the ith by PCI address among the GPUs CUDA sees, whatever CUDA_DEVICE_ORDER says.val · rig
- Rig_gpus.gpuThe type for a GPU a path sees.type · rig
- Rig_metal.gpugpu i is the opener of GPU i, named "METAL" for 0 and "METAL:i" otherwise.val · rig
- Rig_nv.gpuThe type for what a path reads of a GPU, from which make builds its Rig_nv_abi.Gpu.t.type · rig
- Rig_nv_abi.GpuNVIDIA GPUs, as their formats depend on them.module · rig
- Rig_nv_nvidia.gpugpu i is the opener of GPU i, named "NV" for 0 and "NV:i" otherwise.val · rig
- Rig_nv_pci.gpugpu ~machine ~firmware i is the opener of GPU i of machine (defaults to Rig_pci.Machine.this), named "NV-PCI" for 0 and "NV-PCI:i" otherwise.val · rig
- GPU hardware notesWriting rig's drivers taught us facts about Apple, NVIDIA and AMD GPUs, and the software around them, that the vendors document poorly or not at all.page · rig
- Tolk.GpudimsLaunch dimensions.module · tolk
- GPUs without a driverOn Linux, rig opens an AMD or NVIDIA GPU that no kernel driver holds.guide · rig
- Rune.gradgrad p f params is the gradient of the real scalar f params at params: the pullback of 1, a value of params' structure, dtypes and shapes.val · rune
- Rune.grad'grad' f x is grad of one array.val · rune
- Ymir_units.Unit.gramgram is 10-3 kg.val · ymir
- Rig_cuda_abi.graphThe type for graphs.type · rig
- Rig_cuda_abi.graphgraph ~handle ~nodes ~collect is the value of a native graph that a driver made.val · rig
- Tolk.GraphUOp graphs as text: the format of the programs the disk cache keeps (Codegen.to_program) and of the goldens, which tolk's test/README.md specifies.module · tolk
- Tolk.Ops.graph_rewritegraph_rewrite ~calls ~pass ~ctx u rules is u rewritten by rules as pass says: each node is rebuilt on its rewritten sources and rewritten by the first rule that applies, and shared nodes stay shared.val · tolk
- Hugin_gg.Color.graygray ~alpha l is v ~alpha l l l.val · hugin
- Ymir_units.Unit.graygray is m2 s-2.val · ymir
- Norn.Bij.greatergreater ~low maps u to low + exp u, onto (low, inf).val · norn
- Nx.greatergreater a b is less b a.val · nx
- Nx.greater_equalgreater_equal a b is less_equal b a.val · nx
- Nx.greater_equal_sgreater_equal_s x c is greater_equal x c.val · nx
- Nx.greater_sgreater_s x c is greater x c.val · nx
- Hugin_gg.Color.greengreen is v 0.val · hugin
- Hugin_kit.Scheme.greensgreens is Brewer's Greens.val · hugin
- Hugin_kit.Scheme.greysgreys is Brewer's Greys.val · hugin
- Hugin.gridgrid rows arranges figures in a grid of rows, from the top, of cells, from the left: grid [ [ a; b ] ] puts a left of b, and grid [ [ a ]; [ b ] ] puts a above b.val · hugin
- Jera.GridTensor-product series over grids.module · jera
- Ymir.GridCells seen through a transform.module · ymir
- Ymir.Observation.gridgrid o is o's grid.val · ymir
- Fehu_envs.Grid_world5x5 grid navigation with obstacle.module · fehu
- Hugin_kit.Locale.groupgroup l is the digit group separator of l.val · hugin
- Hugin_vg.Picture.groupgroup ps paints the pictures of ps in order, each over those before it.val · hugin
- Tolk.Ops.groupgroup us is the empty Op.t.Stack ordered after us (after): the effects us merged without a value.val · tolk
- Tolk.Postrange.Scheduler.group_for_reducesgroup_for_reduces k is the number of k's reduce axes that are warp or local axes.val · tolk
- Tolk.Gpudims.group_gpudimsgroup_gpudims r sink is the kernel sink with its global loops replaced by workgroup indices (grouped_dims ~reverse:true with prefix "gidx") and its warp and local loops by thread indices (prefix "lidx"), within r's bounds.val · tolk
- Tolk.Gpudims.grouped_dimsgrouped_dims ~reverse prefix dims max_sizes is the index of each loop of sizes dims, in terms of hardware indices, one per axis, whose sizes fit max_sizes if given.val · tolk
- Hugin_kit.Locale.groupinggrouping l is the sizes of the digit groups of l, from the decimal separator leftwards, the last size repeating, without the repeats of the last size that end the list it was given: grouping (v ~grouping:[3; 3] ()) is [3].val · hugin
- Nx.groupsThe type for n keys in k groups of equal keys, numbered in order of first appearance.type · nx
- Rig.Submission.Run.groupsgroups r b x y z sets the grid of b to x by y by z groups.val · rig
- Hugin_gg.Box2.growgrow d b is b with each side moved outwards by d, inwards if d is negative.val · hugin
- Rig_amd.growgrow d makes what d's queues need to hold the parts of the submission whose room last answered RIG_GROW on the calling thread (The C functions).val · rig
- Rig_cuda.growgrow d makes what d's queues need to hold the parts of the submission whose room last answered RIG_GROW on the calling thread (The C functions).val · rig
- Rig_edge.Driver.growgrow d makes what d's queues need to hold the parts of the submission whose room last answered RIG_GROW on the calling thread (The C functions).val · rig
- Rig_metal.growgrow d makes what d's queues need to hold the parts of the submission whose room last answered RIG_GROW on the calling thread (The C functions).val · rig
- Rig_nv.growgrow d makes what d's queues need to hold the parts of the submission whose room last answered RIG_GROW on the calling thread (The C functions).val · rig
- Rig_remote_proxy.growgrow d makes what d's queues need to hold the parts of the submission whose room last answered RIG_GROW on the calling thread (The C functions).val · rig
- Rig_edge.grownThe type for a driver's answer when asked to grow a device's queues (Driver.grow).type · rig
- Ymir.Earth.gstgst e is the Greenwich apparent sidereal time in [0, 2π) radians, the angle from the true equinox to the terrestrial origin: era less the equation of the origins, the CIO's right ascension of the equinox (IERS Conventions 2010, 5.30).val · ymir
- Talon.Join.gtgt l r is lt, as Expr.(>) compares.val · talon
- Tolk.Bigint.gtgt n0 n1 is n0 > n1.val · tolk
- Tolk.Ops.Elementwise.gtgt x y is lt y x.val · tolk
- Tolk.Ops.gtgt x y is lt y x.val · tolk
- Tolk.Ops.Upat.gtgt x y is lt y x.val · tolk
- Rig_amd.Lock.guardedguarded fn f is f () run with the calling thread's signals blocked but inside the path's waits, except on Windows, where OCaml has no signal masks and it blocks none.val · rig
- Rig_nv.Lock.guardedguarded fn f is f () run with the calling thread's signals blocked but inside the path's waits (Drivers), except on Windows, where OCaml has no signal masks and it blocks none.val · rig
- Nx.Rng.gumbelgumbel dt shape is standard Gumbel draws, of distribution exp (-exp (-x)): -log e for an exponential draw e of 53 random bits at float64 and 32 otherwise, computed so that e keeps its relative precision near 0., so the draws are finite.val · nx
- Nx.Rng.gumbel_atgumbel_at p dt shape is gumbel dt shape at p.val · nx
h
- Hugin_gg.Box2.hh b is maxy b -.val · hugin
- Hugin_vg.Renderable.hh r is the height of the page of r, in points.val · hugin
- Rig_bin.Agent.halfhalf ~firmware address is the half of the agent at address, HOST:PORT as written, which passes firmware on to its agent.val · rig
- Norn.Dist.half_cauchyhalf_cauchy ~scale is the distribution of |x|, x Cauchy with location 0, on (0, inf).val · norn
- Norn.Dist.half_normalhalf_normal ~scale is the distribution of |x|, x normal with location 0, on (0, inf).val · norn
- Hugin_text.Text.Layout.halignThe type for horizontal alignments (Alignment).type · hugin
- Tolk.Heuristic.hand_coded_optimizationshand_coded_optimizations k is a copy of k with the optimisations these rules pick; k itself is left as it is.val · tolk
- Rig.Buffer.handlehandle b is the driver's object for b's memory (Rig_edge.location), such as an MTLBuffer, which b starts offset bytes into.val · rig
- Rig_remote_proxy.Wire.handoverThe type for handovers: the work of one value on a device's timeline.type · rig
- Rig_pci.Driver.hang_mshang_ms is 30_000, 30 seconds.val · rig
- Nx.hannhann dt n is the host value of the periodic Hann window of n points, 0.5 - 0.5 cos (2πi / n), computed at float64 and rounded once: the point that would open the next period is dropped, so shifted copies sum to a constant.val · nx
- Nx.hann_athann_at p dt n is hann dt n at p, an _at form as Creation states.val · nx
- Ymir_units.Codata.hartree_energyhartree_energy r is Eh in J.val · ymir
- Tolk.Ops.has_buffer_identityhas_buffer_identity ~after_ok u is true iff u names storage, through reshapes and, if after_ok, an Op.t.After.val · tolk
- Tolk.Compiler_cpu.has_featurehas_feature arch f is true iff the flags clang writes for arch enable the feature f by name: its last mention among arch's FEATUREs is f and not -f, since Clang applies them in order, or, on an arm64 Mac whose CPU is native, f is fp16 and macOS lists it, with no -fp16 after.val · tolk
- Tolk.Ops.has_unbound_outputshas_unbound_outputs c is true iff the call c still has outputs that no buffer is bound to.val · tolk
- Nx.Ptree.Skeleton.hashhash k hashes k: equal skeletons hash equal.val · nx
- Nx_array.Layout.hashhash l is a hash of l such that equal l l' implies hash l = hash l'.val · nx
- Talon.Expr.hashhash ~seed a is the hash of each row of a (Hashing), as int64, and is never null.val · talon
- Tolk.Dtype.hashhash dt is a hash of dt, compatible with equal.val · tolk
- Tolk.Ops.hashhash u is a hash of u, compatible with equal.val · tolk
- Tolk.Dtype.hash_consthash_const c is a hash of c, compatible with equal_const.val · tolk
- Ymir_units.Unit.hbarhbar is ℏ, planck / (int 2 * pi * radian): the action per radian.val · ymir
- Ymir_fits.Fits.hduThe type for HDUs: a header and the bytes of its data unit.type · ymir
- Ymir_fits.Fits.Image.hduhdu ~tiles h t is t as an image HDU with h's cards that are not structural (write); BLANK is dropped from a float image.val · ymir
- Ymir_fits.Fits.Table.hduhdu h columns is a BINTABLE of columns in order, with h's cards that are neither structural nor column-numbered.val · ymir
- Kaun.Init.he_normalhe_normal samples a truncated normal with target variance 2 / fan_in.val · kaun
- Kaun.Init.he_uniformhe_uniform samples from U(-limit, limit) with limit = sqrt (6 / fan_in).val · kaun
- Talon.Kit.headhead n q is the first n rows of q:val · talon
- Ymir_fits.Fits.headerheader hdu is hdu's header.val · ymir
- Ymir_fits.Fits.HeaderHeaders.module · ymir
- Ymir.HealpixHEALPix pixels and spherical harmonics.module · ymir
- Ymir_units.Unit.hectarehectare is 104 m2.val · ymir
- Ymir_units.Unit.hectohecto u is 102 u.val · ymir
- Tolk.Heuristic.held_registersheld_registers k file is the registers of file that k's reduces' accumulators take: one accumulator for each output lane of each reduce.val · tolk
- Tolk.Heuristic_host.held_registersheld_registers k file is the registers of file that k's reduces' accumulators take.val · tolk
- Ymir.Earth.heliocentricheliocentric t is (p, v), the Earth's position and velocity from the Sun's centre at TDB t, each [...; 3] in au and au/d, within 5e-15 au and 1e-16 au/d of iauEpv00's.val · ymir
- Ymir_units.Codata.helion_masshelion_mass r is mh, the mass of the helium-3 nucleus, in kg.val · ymir
- Rig_bin.HelpThe pages of rig --help and of each command's --help.module · rig
- Tolk.HelpersCompilation targets, integer and list utilities, terminal text, the disk cache and programs.module · tolk
- Ymir_units.Unit.henryhenry is kg m2 s-2 A-2.val · ymir
- Jera.Piecewise.hermitehermite x ~values ~slopes is the piecewise cubic with the given values and slopes at the knots: degree 3, once continuously differentiable.val · jera
- Ymir_units.Unit.hertzhertz is s-1.val · ymir
- Tolk.HeuristicHand-coded kernel optimisations.module · tolk
- Tolk.Heuristic_commonWhat Heuristic's rules read of a kernel and apply to it.module · tolk
- Tolk.Heuristic_hostThe host's rules of the hand-coded optimisations.module · tolk
- Nx.hffthfft dt z is the forward transform over axes of the Hermitian signal h, with h(-j) = conj h(j) for every index j over axes taken modulo their lengths, whose first n/2 + 1 samples along the last of axes are z, read as irfft reads its bins: its spectrum is real.val · nx
- Sowilo.hfliphflip img flips horizontally (left to right).val · sowilo
- Nx_wide.hihi w is w rounded to nearest.val · nx
- Tolk.Helpers.hi32hi32 x is x shifted right by 32 bits, keeping its sign.val · tolk
- Rig_amd_abi.Code_object.hiddenThe type for the implicit arguments a kernel reads, which its dispatcher writes among its arguments.type · rig
- Tolk.Compiler_amd.hiphip arch is the compiler of HIP source to executable code objects for the GPU architecture arch, as in "gfx1100".val · tolk
- Tolk.Cstyle.hiphip target renders HIP C++ for an AMD GPU, compiled with comgr.val · tolk
- Hugin_kit.Stats.histogramhistogram ~bins v counts the values along the last axis of v into bins bins of equal width, one histogram per index of v's other axes.val · hugin
- Nx.histogramhistogram [ (e1, x1); …; (ed, xd) ] counts points into the cells of the edges e1 to ed: point i is (x1.{i}, …, xd.{i}), one per element of the coordinates, which share one shape.val · nx
- Munin.HistoryWhat a run's log says, folded.module · munin
- Talon.Source.hivehive paths is the table of the keys and values in paths, one row per path in its order: the column "path", then one string column per key of the first path, in the order they appear in it.val · talon
- Norn.HmcHamiltonian Monte Carlo over many chains, with one trajectory length.module · norn
- Norn.Dist.hmmhmm ~initial ~transition n d is n observations of a hidden Markov chain over k states, its states summed out: state z_1 is j with probability softmax initial, state z_(t+1) is j with probability softmax transition_i after z_t = i, and observation t is a draw of d_(z_t), the slice of d at z_t on its leading axis.val · norn
- Rig.Claim.holdThe type for claims that keep buffers in use until they are released or their value is unreachable.type · rig
- Rig.Claim.holdhold bs takes a read claim on the memory of each of bs that no donation writes and no consume ends (consumable).val · rig
- Rig_edge_primitives.holdhold () counts one more lock of the calling thread, from before its take (rig_hold).val · rig
- Rig_pci.Driver.holdThe type for the process's hold on one function.type · rig
- Rig_pci.Space.holdThe type for one holder's hold on one range.type · rig
- Rig_pci.Space.holderThe type for holders of ranges.type · rig
- Rig_pci.Space.holderholder () holds no range.val · rig
- Nx_array.Dtype.holdsholds dt x is false iff x is an int below min_value dt or above max_value dt: the only values a store into dt refuses.val · nx
- Rig_edge_primitives.holdsholds () is the number of locks the calling thread holds (rig_holds).val · rig
- Talon.Type.holdsholds t v is true iff t holds v, that is iff v can be stored as t:val · talon
- Tolk.Heuristic_common.holdsholds c is the truth of the comparison c.val · tolk
- Rig_program.holeThe type for holes: places in a description's bytes that receive a value when the program loads or runs.type · rig
- Munin_sys.hosthost () is a new reading of the machine.val · munin
- Nx.hostThe brand of Host.type · nx
- Nx.HostThe process's host, Rig.host, computed by nx.cpu.module · nx
- Nx_array.hosthost dt s is create Rig.host dt s: a fresh C-contiguous array of shape s in the host's memory, which write fills.val · nx
- Rig.hosthost is this process's device, named "CPU".val · rig
- Rig_bnxt.Region.hosthost nic n is a region over new, zeroed system memory of nic's machine, of n bytes rounded up to the machine's page (Rig_pci.Machine.page), which the process reads and writes with read and write.val · rig
- Rig_pci.Memory.hosthost r is the process's window on r's bytes, if it has one: the system memory alloc gave, the BAR window on Bar memory, the bytes map_host borrowed, or the window of the region a map_peer region maps.val · rig
- Rig_remote_abi.hostThe type for the record of another machine's host.type · rig
- Tolk.Host_configmodule · tolk
- Tolk.Heuristic_host.host_elementwisehost_elementwise k is true iff k renders for the host and holds no reduce.val · tolk
- Rig.host_ofhost_of d is the host of d's machine: host for this machine's devices, and the device open_host opened for another machine's.val · rig
- Rig_bin.Address.host_porthost_port s is the host, its brackets taken off, and the port of s, written HOST:PORT.val · rig
- Rig_remote.hostshosts j is the host of each of j's machines, in the order of their agents in connect's list or in RIG_REMOTE_AGENTS (Launchers).val · rig
- Ymir_units.Unit.hourhour is 3600 s.val · ymir
- Hugin_kit.Time.hourshours k is seconds (3600 * k).val · hugin
- Talon.Time.Span.hourshours n is like us in hours.val · talon
- Sowilo.hsv_to_rgbhsv_to_rgb img converts HSV back to RGB [0, 1].val · sowilo
- Ymir.Cosmology.hubblehubble c z is H(z), in h0's unit.val · ymir
- Kaun.Loss.huberhuber ?delta predictions targets is the reduction over all elements of the Huber loss of the error d = predictions - targets: d²/2 where |d| <= delta and delta * (|d| - delta/2) beyond.val · kaun
- huginHugin draws figures from nx tensors.library · hugin
- HuginFigures.module · hugin
- Hugin_float_textmodule · hugin
- Hugin_fontFonts and glyph runs.module · hugin
- Hugin_ggGeometry and colour.module · hugin
- Hugin_gg_kitPlanar geometry.module · hugin
- Hugin_kitVisual encoding.module · hugin
- Hugin_literalmodule · hugin
- Hugin_textRich text and its layout.module · hugin
- Hugin_vgPictures and renderables.module · hugin
- Hugin_vg_instancesmodule · hugin
- Hugin_vg_pdfPDF rendering.module · hugin
- Hugin_vg_rasterRaster rendering.module · hugin
- Hugin_vg_svgSVG rendering.module · hugin
- Hugin_vg_vectormodule · hugin
- Nx.hypothypot x y is sqrt (x{^2} + y{^2}) with no intermediate overflow: finite wherever the result is.val · nx
i
- Nx.i0ei0e x is e{^-|x|} I₀(x), I₀ the modified Bessel function of the first kind of order 0: even, 1.val · nx
- Nx.i1ei1e x is e{^-|x|} I₁(x), I₁ the modified Bessel function of the first kind of order 1: odd, so -0.val · nx
- Rig_metal_abi.icbThe type for indirect command buffers.type · rig
- Rig_metal_abi.icbicb ~handle ~commands ~collect is the value of a native indirect command buffer that a driver made.val · rig
- Ymir.Frame.icrstype · ymir
- Ymir.Frame.icrsicrs is the International Celestial Reference System.val · ymir
- Fehu.Env.idid env is the environment's identifier, if any.val · fehu
- Hugin.idThe type for the ids of a figure's nodes.type · hugin
- Hugin.Mark.idid rows is the id of the mark, or of the legend in a swatch (v).val · hugin
- Hugin_gg.Affine.idid is the identity map.val · hugin
- Munin.Run.idid r is r's id (Names).val · munin
- Munin.Session.idid s is the id of s's run.val · munin
- Quill.Cell.idThe type for cell identifiers.type · quill
- Quill.Cell.idid c is the unique identifier of cell c.val · quill
- Rig_pci.Machine.idThe type for the identity of a function.type · rig
- Ymir.Transform.idid maps every point to itself.val · ymir
- Brot.id_to_tokenid_to_token t id is the token string for id in t, if any.val · brot
- Nx.idctidct x is the inverse of dct under the same type_ and norm: types I and IV are their own inverses up to scale, II and III each other's.val · nx
- Tolk.Op.Set.idempotentidempotent is the binary operations f with f x x = x: Or, And and Max.val · tolk
- Brot.Normalizer.identityidentity s is the alignment of s on itself, for text that was not normalized.val · brot
- Norn.Bij.identityidentity maps u to u, onto the reals.val · norn
- Tolk.C.identityidentity file names the library file, as findlib finds it: its path, size and time of last modification, or its path alone if it cannot be read, as a library of macOS's shared cache, and "none" for None.val · tolk
- Tolk.Ops.identity_elementidentity_element op dt is the value e of type dt with op e x = x for every x of type Dtype.strong dt (Dtype.strong), the width an accumulator of dt takes: 0 for Op.t.Add, 1 for Op.t.Mul, and that type's least value for Op.t.Max.val · tolk
- Nx_kernel.Prog.E.idivval · nx
- Brot.Encoding.idsids enc is the token ID array.val · brot
- Brot.idsThe type for flat token id buffers: the rows of a batch, one after the other.type · brot
- Nx.idstidst x is the inverse of dst under the same type_ and norm.val · nx
- Tolk.Heuristic_common.idxidx b is the index of the access b.val · tolk
- Talon.Expr.if_if_ c a b is a where c is true and b where c is false or null.val · talon
- Nx.ifftifft x is the inverse transform of x along axes: under one norm, ifft (fft x) is x to rounding.val · nx
- Nx.ifftshiftifftshift x undoes fftshift along the same axes: x rolled by n / 2 places toward the start.val · nx
- Tolk.Setting.ignore_beam_cacheignore_beam_cache makes the beam search ignore the results it cached, from IGNORE_BEAM_CACHE.val · tolk
- Nx.ihfftihfft dt x is the first n/2 + 1 samples, along the last of axes, of the Hermitian signal whose transform over axes is the real x: conjugate (rfft dt x) scaled as an inverse transform.val · nx
- Norn.Dist.iidiid s d is s independent draws of d, its value of shape s followed by d's.val · norn
- Nx.imagimag dt z is z's imaginary parts, as real.val · nx
- Fehu.Render.imageThe type for rendered frames.type · fehu
- Fehu.Render.imageimage ~width ~height data constructs a frame.val · fehu
- Hugin.imageimage px draws the images of px.val · hugin
- Hugin_vg.Picture.imageimage b px paints the pixels of px over the box b.val · hugin
- Hugin_vg_vector.Vector.imageThe type for images cropped to the cut.type · hugin
- Rig.ImageCode loaded on a device.module · rig
- Rig.Image.imageimage fn is the image fn belongs to.val · rig
- Rig_amd.imageThe type for loaded code.type · rig
- Rig_amd.imageimage d b takes the binary b, or is Error why if d refuses it.val · rig
- Rig_amd_abi.Code_object.imageimage co is co's image: size bytes, laid out as the module's preamble says.val · rig
- Rig_cuda.imageThe type for loaded code.type · rig
- Rig_cuda.imageimage d b takes the binary b, or is Error why if d refuses it.val · rig
- Rig_edge.Driver.imageThe type for loaded code.type · rig
- Rig_edge.Driver.imageimage d b takes the binary b, or is Error why if d refuses it.val · rig
- Rig_metal.imageThe type for loaded code.type · rig
- Rig_metal.imageimage d b takes the binary b, or is Error why if d refuses it.val · rig
- Rig_nv.imageThe type for loaded code.type · rig
- Rig_nv.imageimage d b takes the binary b, or is Error why if d refuses it.val · rig
- Rig_pci.Firmware.imageThe type for firmware images found.type · rig
- Rig_program.imageThe type for images: binary in device's format, loaded in the description's order (Rig.Image.load).type · rig
- Rig_remote_proxy.imageThe type for loaded code.type · rig
- Rig_remote_proxy.imageimage d b takes the binary b, or is Error why if d refuses it.val · rig
- Ymir_fits.Fits.ImageImages.module · ymir
- Hugin_vg_raster.image_samplesimage_samples is 4, the most points per axis that a pixel averages an image shown smaller than its pixels at: the image pixels under (p + (i + 0.5) / 4), for device pixel p and i in [0;3], along each axis (Accuracy).val · hugin
- Rig_bin.ImagesThe firmware images each driver-less driver boots with: its gen/headers/firmware.tsv, from which gen/gen.py also makes the driver's pins.module · rig
- Hugin_kit.Scale.implyimply i s is s with each property that s leaves unset taken from i, except the name, which imply keeps from s or keeps absent.val · hugin
- Norn.Gaussian.importanceimportance u lp k ~n g is n draws x_s of g weighted for the density lp, each log weight the raw log importance ratio lp x_s - log g (x_s).val · norn
- Jera.Sde.Brownian.incrementincrement w s t is W t − W s and the space–time Lévy area over [s, t], H = (1 / (t − s)) ∫_s^t (W r − W s − (r − s) / (t − s) (W t − W s)) dr, each of the path's shape; H is zero when s = t.val · jera
- Quill.Cell.increment_execution_countincrement_execution_count c increments the execution count of a code cell.val · quill
- Hugin.indexindex ~scale ~title k is the index along axis k of the shape of the mark it is bound in, as quantities: the datum (i0, …, in) has the value float ik.val · hugin
- Hugin.Mark.indexindex rows is the datum of each row, as its index in the mark's shape flattened in row-major order.val · hugin
- Nx.indexThe type for selections along one axis.type · nx
- Nx.Ptree.Walk.indexindex c i walk x is walk c' x, c' at c's path then Index i.val · nx
- Tolk.Ops.indexindex u idxs is the Op.t.Index of u by idxs; the ith source of a stack if u is a stack and idxs the constant i.val · tolk
- Tolk.Ops.Upat.indexval · tolk
- Tolk.Heuristic_common.indexesindexes r b is true iff the index of b computes from r, r aside.val · tolk
- Tolk.IndexingRanges: from whole tensors to indexed elements.module · tolk
- Tolk.Coalesce.indexing_simplifyindexing_simplify rewrites an index where cond x invalid into a buffer to where cond x' invalid, with x' the simplification of x given that cond holds (Symbolic.uop_given_valid), when the condition simplifies x further than Shape.simplify does alone.val · tolk
- Hugin_kit.Scheme.infernoinferno runs from black through purple and orange to pale yellow, with more contrast at the high end than magma.val · hugin
- Nx.InfixOperators over values, for a local open.module · nx
- Fehu.InfoStep metadata dictionaries.module · fehu
- Nx_io.Gguf.infoinfo name g is the description of g's array name.val · nx
- Norn.Evidence.informationinformation z is H, the posterior's Kullback-Leibler divergence from the prior in nats, E_post (ln L) - ln Z: how much the data narrowed the prior.val · norn
- Rig_bin.Proc.inheritedinherited fd f is f (), with fd passed on to the processes f starts.val · rig
- Kaun.Attention.initinit ~key ~embed_dim is make ~key ~embed_dim Nx.float32: Glorot-uniform weights, zero biases.val · kaun
- Kaun.Batch_norm.initinit ~features at is a fresh layer for features-dimensional features at at: parameters with gamma all ones and beta all zeros, and statistics with mean all zeros and var all ones (so an untrained layer's eval mode is close to the identity of a standardized input).val · kaun
- Kaun.Conv.initinit ~key ~in_channels ~out_channels ~kernel_size is make ~key ~in_channels ~out_channels ~kernel_size Nx.float32: Glorot-uniform weights, zero bias.val · kaun
- Kaun.Embedding.initinit ~key ~vocab ~dim is make ~key ~vocab ~dim Nx.float32: rows drawn from N(0, 1).val · kaun
- Kaun.InitWeight initializers (Glorot, He, LeCun, variance scaling).module · kaun
- Kaun.Layer_norm.initinit ~dim at is make ~dim Nx.float32 at.val · kaun
- Kaun.Linear.initinit ~key ~inputs ~outputs is make ~key ~inputs ~outputs Nx.float32: Glorot-uniform weights, zero bias.val · kaun
- Kaun.Rms_norm.initinit ~dim at is make ~dim Nx.float32 at.val · kaun
- Norn.Ensemble.initinit u ?ensembles lp start is the state of walkers at start, a position, in ensembles ensembles, by default 1.val · norn
- Norn.Hmc.initinit u ?accept ?rank ?geometry lp start is the state of chains at start, a position.val · norn
- Norn.Nested.initinit u ?batch ?tolerance ~budget ~prior ~likelihood live is the state with the live points live, n independent draws of the prior.val · norn
- Norn.Nuts.initinit u ?max_depth ?accept ?rank ?geometry lp start is the state of chains at start, a position.val · norn
- Norn.Smc.initinit u ?move ?resampled ~prior ~likelihood start is the state at β = 0 of the particles start, N independent draws of the prior.val · norn
- Norn_model.initinit ?from m y ~chains k is a start of chains chains with a finite log density.val · norn
- Norn_model.InitInitialisation strategies.module · norn
- Nx.initinit dt s f is the host value of shape s whose element at index i is f i.val · nx
- Quill_top.initialize_if_neededinitialize_if_needed () ensures the OCaml toplevel environment is initialized.val · quill
- Hugin.Theme.inkink th is the ink of th.val · hugin
- Hugin_font.Font.inkink f g is Path.bounds (outline f g), the smallest box containing the ink of g, or None if g has no ink.val · hugin
- Hugin_text.Text.Layout.inkink l is the union of the Hugin_font.Run.bounds of the runs of l at their origins, or None if no glyph has ink.val · hugin
- Nx.innerinner a b is the sum of the products of a's last axis with b's last axis; the result's axes are a's others, then b's.val · nx
- Nx_kernel.Prog.E.inputinput k dt is In k, operand k of dtype dt.val · nx
- Rig_program.inputThe type for inputs: a buffer of device of at least bytes bytes.type · rig
- Input/OutputNx_io loads and saves arrays in image, NumPy, SafeTensors, and delimited text formats, and loads GGUF.guide · nx
- Tolk.Ops.insins u i is the instruction i with u's sources and type, unless given.val · tolk
- Quill.Doc.insertinsert ~pos c d inserts c at position pos.val · quill
- Quill.Session.insert_cellinsert_cell ~pos cell s inserts cell at position pos.val · quill
- Hugin.insideThe type for the places of a legend inside a data area: a corner, the middle of a side, or the centre.type · hugin
- Talon.Join.insideinside p (s, e) matches the rows whose left column p is from the right column s, included, to the right column e, excluded.val · talon
- Nx_top.install_printerval · nx
- Quill_top.install_printerinstall_printer name installs a toplevel pretty-printer by evaluating #install_printer name;;.val · quill
- InstallationRaven requires OCaml 5.5 or later and opam.manual
- Hugin_vg_instances.InstancesThe rules of stamps that every renderer follows: which instances show and which colours they paint with, and the walks of a stamp's picture that decide them.module · hugin
- Talon.Col.instantinstant name is v Kind.instant name, which binds every datetime type of instants.val · talon
- Talon.Expr.instantinstant t is the instant t.val · talon
- Talon.Kind.instantinstant reads datetime columns of every unit whose clock is UTC (Instants and wall clocks).val · talon
- Talon.Time.instantThe type for instants.type · talon
- Talon.Type.instantinstant u is Datetime of instants in the unit u.val · talon
- Nx.Ptree.instantiateinstantiate (module U) is U with arrays at its parameter's positions: nest (module U) tensor.val · nx
- Rig.instruction_setinstruction_set d is the instruction set of d's arch: "x86_64" for a level of x86-64, "arm64" for an arm64 arch, and the arch itself for any other.val · rig
- Nx.Ptree.Walk.intint c n reports n at c's path and is n.val · nx
- Talon.Col.intint name is v Kind.int name, which binds every integer type.val · talon
- Talon.Expr.intint n is the integer n.val · talon
- Talon.Kind.intint reads the integer columns, int8 through int64 and uint8 through uint64.val · talon
- Tolk.Ops.Elementwise.O.intint n is the weak literal n.val · tolk
- Tolk.Ops.intint ~dtype n is const ~dtype (`Int n).val · tolk
- Tolk.Ops.O.intint n is the weak literal n.val · tolk
- Tolk.Ops.Upat.intint n is the weak literal n.val · tolk
- Tolk.Ops.Upat.O.intint n is the weak literal n.val · tolk
- Tolk.Setting.intint ~reach key default is a setting whose initial value is the integer the variable key holds, or default if key is unset.val · tolk
- Ymir_fits.Fits.Value.intint is integer text within int's range.val · ymir
- Ymir_units.Unit.intint n is the number n, factored into primes exactly.val · ymir
- Nx.int16val · nx
- Talon.Type.int16int16 is Int16.val · talon
- Nx.int16_signed_elttype · nx
- Nx_array.Dtype.int16_signed_eltSigned 16-bit integers.type · nx
- Nx.int16_ttype · nx
- Nx.int16_unsigned_elttype · nx
- Nx_array.Dtype.int16_unsigned_eltUnsigned 16-bit integers.type · nx
- Nx.int32val · nx
- Rig.Submission.Run.int32val · rig
- Talon.Type.int32int32 is Int32.val · talon
- Nx.int32_elttype · nx
- Nx_array.Dtype.int32_eltSigned 32-bit integers.type · nx
- Nx.int32_ttype · nx
- Nx.int4val · nx
- Nx.int4_elttype · nx
- Nx_array.Dtype.int4_eltSigned 4-bit integers, two's complement.type · nx
- Nx.int4_ttype · nx
- Nx.int64val · nx
- Rig.Submission.Run.int64val · rig
- Talon.Type.int64int64 is Int64.val · talon
- Nx.int64_elttype · nx
- Nx_array.Dtype.int64_eltSigned 64-bit integers.type · nx
- Nx.int64_ttype · nx
- Nx.int8val · nx
- Talon.Type.int8int8 is Int8.val · talon
- Nx.int8_signed_elttype · nx
- Nx_array.Dtype.int8_signed_eltSigned 8-bit integers.type · nx
- Nx.int8_ttype · nx
- Nx.int8_unsigned_elttype · nx
- Nx_array.Dtype.int8_unsigned_eltUnsigned 8-bit integers.type · nx
- Tolk.Setting.int_optionint_option ~reach key is a setting whose initial value is Some n if the variable key holds the integer n, and None if key is unset: the setting's documentation states what None stands for.val · tolk
- Jera.Grid.integralintegral ~axis g is the antiderivative of g along axis, zero at that axis's first break, of one degree more along it.val · jera
- Jera.Piecewise.integralintegral p is the antiderivative of p that is zero at the first break and continuous across breaks, of degree one more.val · jera
- Ymir.Observation.integralThe type for integrals of an observation over a region, each of the region's and window's batch axes broadcast.type · ymir
- Jera.Quad.integrandThe type for integrands: f x is f at each point of x: of x's shape over a range, and over a box of x's shape without its last, coordinate axis.type · jera
- Ymir.Observation.integrateintegrate r o weighs o's cells with Region.weights and sums each term once in double-word arithmetic at 'e, with w the weight.val · ymir
- Hugin_gg.Box2.interinter a b is the box of the points in both a and b, or None if they have none.val · hugin
- Talon.Sel.interinter s0 s1 selects s0's columns that are in s1, in s0's order.val · talon
- Talon.Kit.interpolateinterpolate ~on x is x where it holds a value.val · talon
- Sowilo.interpolationThe type for interpolation methods.type · sowilo
- Quill.Kernel.interruptinterrupt k requests that the code k runs stop.val · quill
- Rig_nv_abi.Method.interruptinterrupt raises a non-stalling interrupt when the channel reaches it, without waiting, so it may reach the host before the write of a release just before it is visible.val · rig
- Rig_pci.Function.interruptinterrupt f ms waits at most ms milliseconds for an interrupt of f, with the OCaml runtime released, and is true iff one arrived.val · rig
- Rig_bin.Proc.interruptedinterrupted () is the first interrupt the process got, at the first call after a wait it woke, and None at every other call.val · rig
- Talon.Kit.intersect_intervalsintersect_intervals ~by (s, e) rest q is the values that both q's and rest's set of a key hold, written q |> Kit.intersect_intervals (s, e) rest, in normal form (Normal form).val · talon
- Hugin.View.intervalinterval name ~init is the key name of an interval or none, such as a brushed selection, initially init.val · hugin
- Hugin_kit.Time.intervalThe type for calendar intervals.type · hugin
- Norn.Bij.intervalinterval ~low ~high maps u to low + (high - low) sigmoid u, onto (low, high).val · norn
- Hugin_kit.Stack.intervalsintervals ~offset ~columns ~series lengths is (starts, ends), where the length lengths.(r), in column columns.(r) and series series.(r), covers the interval from starts.(r) to ends.(r).val · hugin
- IntroductionRaven is a project to bring modern scientific computing to the OCaml programming language.manual
- Tolk.Dtype.intsints is uints @ sints.val · tolk
- Nx.invinv a is solve a of the identity: NaN for a matrix solve finds singular.val · nx
- Tolk.Ops.invalidinvalid is const `Invalid.val · tolk
- Tolk.Shape.invalid_gateinvalid_gate matches where cond x i with i the constant Ops.invalid, naming its sources "cond", "x" and "i".val · tolk
- Tolk.Shape.invalid_patinvalid_pat matches the constant `Invalid, named "i".val · tolk
- Rig_mlx5_abi.Completion.invalidateinvalidate b at marks the completion at byte at of b as one the NIC has not written, so that owned is false for it at every count until the NIC writes it.val · rig
- Rig_nv_abi.Method.invalidate_cachesinvalidate_caches invalidates the compute engine's instruction, data and constant caches, which would hide earlier writes, the host's and other devices' included, from launches scheduled after it.val · rig
- Norn.Bij.inverseinverse b x is the coordinates that b maps to x, for x in the support.val · norn
- Nx.Pattern.inverseinverse p is the one-operand pattern that undoes p: its sides exchanged.val · nx
- Ymir.Transform.inverseinverse t reverses t's stages and flips each one's sense.val · ymir
- Norn.Dist.inverse_gammainverse_gamma ~concentration ~scale is the distribution of 1 / x, x gamma with rate scale.val · norn
- Hugin.Coord.invertinvert p pt is Some (x, y), the normalised position that p puts at the point pt of the page, possibly outside the unit square, or None if p puts no position there.val · hugin
- Hugin_gg.Affine.invertinvert m is Some m', with m * m' the identity up to rounding, if the determinant xx *.val · hugin
- Hugin_kit.Scale.invertinvert s u is the value that s normalises to u:val · hugin
- Sowilo.invertinvert img is 1.0 - img.val · sowilo
- Rig.Buffer.ioio b k is the region b's memory lies in, if it is an io device's memory of a library whose key is k, and None otherwise.val · rig
- Rig_edge.IoDevices of memory that the host reads and writes through calls.module type · rig
- Rig_edge.io_openerThe type for openers of named io devices.type · rig
- Nx.irfftirfft dt x is the real inverse of rfft along axes, of n points along the last of them from its first n/2 + 1 bins, which are cut or extended with zeros to that many.val · nx
- Tolk.Op.Set.irreducibleirreducible is the operations that index arithmetic treats as atoms: Const, Special, Range, Param and Getaddr.val · tolk
- Jera.Solution.isis st s is true in each lane whose status is st, of the lanes' shape.val · jera
- Nx_array.Dtype.isis k dt is true iff kind dt is k.val · nx
- Hugin_vg_vector.Vector.is_axialis_axial m is true iff the linear part of m scales the axes by positive factors and turns nothing, up to a relative 1e-9.val · hugin
- Tolk.Dtype.is_boolis_bool dt is true iff dt is Bool.val · tolk
- Rig.Buffer.is_borrowedis_borrowed b is true iff b lies over memory that something other than b's device holds, either a bigarray (of_bigarray) or another device's memory (borrow).val · rig
- Tolk.Shape.is_bound_varis_bound_var u is true iff u is a variable bound to a value.val · tolk
- Quill.Kernel.is_completeis_complete k c is true iff c's source holds a complete phrase, ready to run.val · quill
- Tolk.Shape.is_constis_const u is true iff u is a constant node.val · tolk
- Nx_array.Layout.is_contiguousis_contiguous l is true iff element k of l in C order is at position offset l + k.val · nx
- Hugin_vg_vector.Vector.is_dashedis_dashed d pen is true iff a dash length of pen is written non-zero with dash_decimals d decimals.val · hugin
- Tolk.Ops.is_disjointis_disjoint u is true iff u is the Op.t.After of an after_disjoint.val · tolk
- Tolk.Ops.is_disk_deviceis_disk_device d is true iff d is a disk: its name, up to its first :, is DISK in any case.val · tolk
- Nx_array.Layout.is_distinctis_distinct l is true iff l has no element or, with l's axes of extent above 1 ordered by |stride|, ties by axis, each stride exceeds the reach Σ (d{_j} - 1)·|s{_j}| of the axes before it.val · nx
- Brot.Encoding.is_emptyis_empty enc is true iff enc has no tokens.val · brot
- Fehu.Info.is_emptyis_empty t is true iff t has no bindings.val · fehu
- Hugin_gg.Path.is_emptyis_empty p is equal p empty.val · hugin
- Hugin_vg_vector.Vector.is_evenis_even m is true iff the linear part of m is a positive multiple of the identity, up to a relative 1e-9.val · hugin
- Tolk.Dtype.is_floatis_float dt is true iff dt is in floats or is Weak_float.val · tolk
- Fehu.Buffer.is_fullis_full buf is true iff size buf = capacity.val · fehu
- Tolk.Indexing.is_gatheris_gather u is whether u is a gather: an Op.t.Index of a tensor by a value with axes, which reads the tensor at the row each element of the value names.val · tolk
- Rig_amd.is_gpuis_gpu ~vendor ~class_ is true iff a PCI function of vendor vendor and 24-bit class code class_ is an AMD GPU: AMD's vendor 0x1002 and a display controller or a processing accelerator, base class 0x03 or 0x12.val · rig
- Rig_nv.is_gpuis_gpu ~vendor ~class_ is true iff a PCI function of vendor vendor and 24-bit class code class_ is an NVIDIA GPU: NVIDIA's vendor 0x10de and a display controller, base class 0x03.val · rig
- Tolk.Ops.is_groupis_group u is true if and only if u is an Op.t.After of the empty Op.t.Stack, as group makes of several effects.val · tolk
- Nx_kernel.Prog.is_identityis_identity p is true iff p has one operand and one output, and its output is that operand's In node, so that p gives its operand back.val · nx
- Talon.Expr.is_inis_in vs a is true iff a is the same key as one of vs (see Type.compare_value), so it is false, never null, where a is null: not (is_in vs a) is true there, whereas not (a = v0 || a = v1) is null.val · talon
- Tolk.Ops.is_inline_callis_inline_call u is true iff u calls a plain sink that is not compiled on its own.val · tolk
- Tolk.Dtype.is_intis_int dt is true iff dt is in ints or is Weak_int.val · tolk
- Tolk.Ops.is_invalidis_invalid u is true iff u is the constant `Invalid.val · tolk
- Tolk.Ops.is_joinis_join u is true iff u is an Op.t.After whose dependencies are all Op.t.Afters: a join, which orders a read of its first source after them and writes nothing.val · tolk
- Quill.Cell.is_languageis_language s is true iff s names a language.val · quill
- Tolk.Shape.is_launchis_launch u is true iff u is a launch: a kernel's Op.t.Sink whose sources are calls of kernels compiled on their own (Ops.call_info), its members.val · tolk
- Tolk.Ops.is_literalis_literal u is true iff u is a constant or a cast of one, whose value value reads.val · tolk
- Talon.Expr.is_nullis_null a is true where a is null and false elsewhere.val · talon
- Tolk.Ops.is_numberis_number u i is true iff source i of u is a number and no value of its peer's type, so that it never wraps at that type's width: the count of a Op.t.Shl or Op.t.Shr, or the weak integer exponent of an Op.t.Pow of an integer base.val · tolk
- Quill_project.is_placeholderis_placeholder nb is true iff nb has no file (empty path).val · quill
- Nx_kernel.Prog.is_singleis_single p is true iff single p is Some n, where n is then p's last node.val · nx
- Tolk.Schedule.is_store_afteris_store_after u is true iff u is an Op.t.After that orders a store into storage: storage other than call-local storage (Op.t.Alloc), or any storage whose first effect is a store.val · tolk
- Tolk.Dtype.is_unsignedis_unsigned dt is true iff dt is in uints.val · tolk
- Tolk.Shape.is_variableis_variable u is true iff u is a scalar variable with a range.val · tolk
- Tolk.Ops.is_virtualis_virtual u is true iff u cannot back a buffer as it is: it has no device, or its type is weak.val · tolk
- Tolk.Shape.is_weakis_weak dt is true iff dt is one of Dtype.weaks.val · tolk
- Nx.isfiniteisfinite x is true where x is neither an infinity nor a NaN, and for a complex number where both parts are.val · nx
- Nx.isinfisinf x is true where x is an infinity, or a complex number with an infinite part.val · nx
- Nx.isnanisnan x is true where x is a NaN, or a complex number with a NaN part.val · nx
- Hugin_gg_kit.Field2.isobandisoband ~lo ~hi f is the polygon whose surface is R lo without R hi: where f is at least lo and below hi.val · hugin
- Hugin_gg_kit.Field2.isolineisoline l f is the isoline of f at level l: the boundary of R l less its parts on the boundary of the domain, as curves.val · hugin
- Nx.istftistft dt ~window z is the signal whose stft under the same window, step and win is z: each frame inverted by irfft, multiplied by win, summed back where it was taken, and divided by the sum of win's squares landing there.val · nx
- Hugin_text.Text.italicitalic t is t with every character italic.val · hugin
- Hugin_font.Font.italic_angleitalic_angle f is the angle of f's upright strokes from the vertical, in radians, negative when they lean right (the sign of OpenType and PDF, opposite to the plane's convention for angles), from the post table.val · hugin
- Nx.itemitem i x is x's element at i, one position per axis, a negative one counting from the end; item [] x reads a 0-d value.val · nx
- Tolk.Dtype.itemsizeitemsize dt is the number of bytes of a dt value, bitsize dt rounded up to a whole byte.val · tolk
- Tolk.Worker.iteriter f g l applies g x (f x) to each element x of l, g on the calling domain, in the order the applications of f end, while up to p - 1 other domains apply f to the elements not yet started.val · tolk
- Tolk.Device.Tiny_elf.iter_sigiter_sig ~offset signature is where the value of each parameter of signature lies when their values are packed in order from byte offset, each aligned to its size: its byte offset and its type, a scalar's own and a buffer's address's Dtype.t.Uint64.val · tolk
- Rune.iterateiterate c ~max ~until f init applies the step f to the carry, from init, until until carry holds, and returns that carry.val · rune
- Rune.iterate'iterate' ~max ~until f x is iterate of one array.val · rune
- Jera.Minimize.iteratesiterates x m ~steps f x0 is the first steps estimates of m's search from x0, x0 first, detached, stacked on a new leading axis of every tensor.val · jera
j
- Rune.jacfwd'jacfwd' f x is the Jacobian of f at x, of shape shape (f x) @ shape x and f x's dtype: its column k is the tangent of f along element k of x in C order.val · rune
- Rune.jacrev'jacrev' f x is the Jacobian of f at x, of shape shape (f x) @ shape x and x's dtype: its row k is the pullback of element k of the result in C order.val · rune
- Ymir.Time.jdjd t is (j1, j2), each instant's Julian date j1 + j2: j1 the Julian date of its day's 0h, exact, and j2 the fraction of the day, in [0, 1], rounded once.val · ymir
- jeraNumerical methods for OCaml, built on Nx and Rune.library · jera
- JeraNumerical methods.module · jera
- Rune.jitjit ~beam ~parallel s f is f compiled, s its signature, its kernels searched as Search states.val · rune
- Tolk.JitCaptured schedules.module · tolk
- Rune.jit'jit' ~beam ~parallel f is jit of a function of one array.val · rune
- Tolk.Jit.jit_lowerjit_lower ~beam ~search ~targets ~held_bufs ~inputs ~input_vars linear is the captured schedule linear, an Op.t.Linear of calls, ready to link:val · tolk
- Tolk.Setting.jitbeamjitbeam is the width of the beam search for the kernels of a captured schedule (Jit), from JITBEAM.val · tolk
- Rig_remote_proxy.Link.jobThe type for the jobs of this process.type · rig
- Rig_remote_proxy.Link.jobjob () is a new open job, with no link.val · rig
- Rig_remote_proxy.Link.job_ofjob_of l is l's job.val · rig
- Hugin_gg.Stroke.joinThe type for line joins.type · hugin
- Hugin_gg.Stroke.joinjoin s is the line join of s.val · hugin
- Talon.JoinJoins of two queries by conditions on their columns.module · talon
- Talon.Query.joinjoin ?kind ?each_left ?each_right ~on right left pairs left's rows with the rows of right that on matches, written left |> join ~on right.val · talon
- Vega.Schedule.joinjoin segments sequences schedules end-to-end.val · vega
- Talon.Kit.join_asofjoin_asof ~by ~direction ~strict ~on:(l, r) right left is each row of left, in order, with its pick in right (As-of joins), written left |> Kit.join_asof ~on right.val · talon
- Joinsleft |> Query.join ~on right pairs the rows of two queries that a condition matches.guide · talon
- Ymir_units.Unit.joulejoule is kg m2 s-2.val · ymir
- Ymir.Units.julian_yearjulian_year is 365.25 Unit.day, 31557600 s.val · ymir
- Rune.jvpjvp p q f params tangents is (f params, dy), dy the tangent of f params where params moves along tangents: the Jacobian-vector product of f at params, computed in one forward pass.val · rune
- Rune.jvp'jvp' f x t is jvp of one array.val · rune
k
- Ymir_units.Unit.katalkatal is mol s-1.val · ymir
- kaunKaun is a neural network library for OCaml, trained with rune autodiff.library · kaun
- KaunNeural networks as typed parameter records.module · kaun
- Kaun_datasetsDataset loaders for kaun.module · kaun
- Kaun_hfHuggingFace Hub integration.module · kaun
- Talon.Expr.keepkeep sel outputs the columns that sel selects, unchanged and under their names.val · talon
- Tolk.Ops.keepWhat later passes may do to a buffer an Op.t.Stage makes.type · tolk
- Kaun.Fn.keep_top_kkeep_top_k ~k logits is logits with every entry below the k-th largest of its row set to negative infinity; the last axis is the vocabulary.val · kaun
- Kaun.Fn.keep_top_pkeep_top_p ~p logits is logits with the least likely entries of each row set to negative infinity, keeping the fewest entries whose softmax probabilities sum to at least p (nucleus sampling).val · kaun
- Ymir_units.Unit.kelvinkelvin is the symbol K.val · ymir
- Tolk.Renderer.Compiler.keptkept c src is the binary of src held in c's disk cache table while Setting.ccache holds, if any: what compile_cached returns without compiling.val · tolk
- Quill.KernelCode execution kernels.module · quill
- Rig_amd_abi.Code_object.kernelThe type for kernels, as their code object describes them.type · rig
- Rig_amd_abi.Code_object.kernelkernel co name is the kernel name of co, whose descriptor is at the symbol name ^ ".kd", if co defines it (kernels).val · rig
- Rig_cuda_abi.kernelThe type for the kernel launches a graph records.type · rig
- Rig_nv_abi.Cubin.kernelThe type for what a launch of a kernel needs from its cubin.type · rig
- Rig_nv_abi.Cubin.kernelkernel c name is what a launch of the kernel name of c needs, or None if c lacks the section .text.name.val · rig
- Tolk.Realize.kernelThe type for a kernel lower_and_compile compiled.type · tolk
- Tolk.Spec.kernel_graphkernel_graph accepts the graph of kernel calls: a sink of calls of opaque bodies over storage and parameters, with the constants, stacks, casts and bitcasts that make their arguments, and orderings of storage.val · tolk
- Tolk.Ops.kernel_infoThe type for the arguments of a kernel's Op.t.Sink.type · tolk
- Tolk.Ops.kernel_infokernel_info () is the argument with these fields.val · tolk
- Tolk.Shape.kernel_paramskernel_params prg is the parameters of the linearized Op.t.Program prg in slot order, the order its kernel takes them, as parameters or packed into one: those of Ops.program_info.params its linear order holds, then each other one whose slot none of those takes.val · tolk
- Sowilo.kernel_shapeThe type for structuring element shapes.type · sowilo
- Rig_amd_abi.Code_object.kernelskernels co is the names of co's kernels, in increasing order: each name such that co defines the symbol name ^ ".kd" in its image.val · rig
- Rig_nv_abi.Cubin.kernelskernels c is the names of the kernels of c, in the order of their code sections.val · rig
- Tolk.Realize.kernelskernels f is f () and each kernel of each schedule that lower_and_compile compiles on the calling domain while f runs, in order, whether its program was made or kept.val · tolk
- Hugin_font.Font.kerningkerning f g g' is the change to advance f g when g' follows g, usually negative.val · hugin
- Hugin.View.keyThe type for keys of values of type 'a.type · hugin
- Munin.Session.keykey m is m's key.val · munin
- Nx.Rng.keyThe type for keys, of shape [|2|], and batches of keys, of shape [|…; 2|]: two 32-bit words per key.type · nx
- Nx.Rng.keykey seed is the host key of seed.val · nx
- Rig.Local.keyThe type for keys of values of type 'a kept per device.type · rig
- Rig_amd.keykey identifies the driver.val · rig
- Rig_amd_abi.keykey is the key of an AMD device's record.val · rig
- Rig_cuda.keykey identifies the driver.val · rig
- Rig_cuda_abi.keykey is the key of a CUDA device's record.val · rig
- Rig_edge.Driver.keykey identifies the driver.val · rig
- Rig_metal.keykey identifies the driver.val · rig
- Rig_metal_abi.keykey is the key of a Metal device's record.val · rig
- Rig_nv.keykey identifies the driver.val · rig
- Rig_nv_abi.keykey is the key of an NVIDIA GPU's record.val · rig
- Rig_remote.keyThe type for a job's keys: 16 to 4096 bytes.type · rig
- Rig_remote.keykey s is s as a job's key, or Error why if it has fewer than 16 or more than 4096 bytes; why gives its length, never its bytes.val · rig
- Rig_remote_abi.keykey is the key of a proxy's record.val · rig
- Rig_remote_proxy.keykey identifies the driver.val · rig
- Tolk.Ops.keykey u is a digest of u's operation, type, argument and sources, recursively: equal for equal graphs, whenever and wherever they are built, and different for NaN constants of different bits.val · tolk
- Tolk.Setting.keykey v is the name of v's environment variable.val · tolk
- Munin.History.key_statsThe type for a key's statistics.type · munin
- Tolk.Ops.key_withkey_with ~arg u is the key of the graph u would be with each node n for which arg n is Some a taking the argument a, computed without building that graph.val · tolk
- Talon.Join.keyskeys ns matches the rows whose columns ns are the same keys on both sides.val · talon
- Ymir.Wcs.keywordsThe type for a header's keywords, as decoded values.type · ymir
- Rig_bin.Proc.killkill pid is signal pid Sys.sigkill.val · rig
- Ymir_units.Unit.kilokilo u is 103 u.val · ymir
- Ymir_units.Unit.kilogramkilogram is the symbol kg.val · ymir
- Hugin.kindkind c is the kind of the scale that reads the data of c, or None if c is a constant.val · hugin
- Hugin_kit.Scale.kindThe type for kinds of scales.type · hugin
- Hugin_kit.Scale.kindkind s is the kind of s.val · hugin
- Norn.Dist.kindThe type for the values of a distribution: a witness of 'x.type · norn
- Norn.Dist.kindkind d is the type of d's values.val · norn
- Nx_array.Dtype.kindThe type for kinds of number.type · nx
- Nx_array.Dtype.kindkind dt is the kind of number dt holds.val · nx
- Rig.Buffer.kindThe type for the kinds of a device's memory that create allocates.type · rig
- Rig_edge.kindThe type for the work of a part, as rig_edge.h tags it.type · rig
- Rig_pci.Memory.kindThe type for where new memory is and who reaches it.type · rig
- Talon.Join.kindThe type for join kinds, which say which rows a join keeps.type · talon
- Talon.KindKinds: the OCaml types that cells read as.module · talon
- Talon.Record.kindkind reads record columns of every field list.val · talon
- Talon.Type.kindkind t is the kind that t's cells read as: Kind.bool, Kind.int for the integer types, Kind.float for the float types, Kind.string for String and Categorical, Kind.binary, Kind.date, Kind.span for Clock and Duration, Kind.instant and Kind.wall for Datetime by its clock, Kind.list (kind e) for List e, Record.kind for every Record, and Kind.tensor dt for Tensor (dt, _).val · talon
- Ymir.Grid.kindThe type for a grid's kind.type · ymir
- Ymir.Grid.kindkind g is g's kind and parts.val · ymir
- Talon.KitCompositions of the verbs and expressions.module · talon
- Tolk.Ops.known_at_launchknown_at_launch r is true iff the range r has a bound that, its ordering after effects (Op.t.After) aside, reads no range and no memory (bound_loads), as j < n for a variable n and unlike j < x[i] or j < i + 1: a launch computes it from the run's variables alone, before the kernel runs, as it does a count of workgroups or of blocks.val · tolk
- Tolk.Shape.known_sizeknown_size s is s's value when it is an integer: an Ops.sint.Int or a literal constant, which may lie past an int.val · tolk
- Nx.kronkron a b is the Kronecker product.val · nx
- Jera.Quad.Rule.kronrodkronrod n is the (2n + 1)-point Gauss–Kronrod rule that extends gauss n (Piessens et al., QUADPACK, 1983): exact for polynomials of degree 3n + 1, with the n Gauss nodes among its own.val · jera
- Nx.kvKeys and values, by key head hk and key tk.type · nx
- Nx_kernel.kvThe type for keys and values.type · nx
- Jera.Ode.kvaerno5kvaerno5 ~mass ~linear derivative is Kværnø's (2004) singly diagonally implicit method of order 5 with an embedded order 4, for stiff fields: derivative t y dy is the field's Jacobian at (t, y) applied to dy, such as snd (Rune.jvp' (f t) y dy).val · jera
l
- Norn.Summary.labelslabels s is s's row labels: an element's path, then its index in brackets for a tensor that is not a scalar, as theta[3] or b[1,0].val · norn
- Vega.lamb_initlamb_init is adam_init: LAMB shares Adam's state.val · vega
- Vega.lamb_steplamb_step p ~lr st ~params ~grads is (params', st') after one LAMB step (You et al., 2020), for large-batch training: adamw_step's update, scaled per leaf by LARS's trust ratio (lars_step).val · vega
- Rune.lane_countlane_count a is Some n inside the map named a of n lanes, and None with no map named a around the call, as where lanes raises.val · rune
- Rune.lane_indexlane_index a x, inside the map named a, is the calling lane's index in it, a 0-d int32 from 0 to the map's number of lanes minus one, whole on each of x's devices.val · rune
- Jera.System.laneslanes ~tol ~budget ~jacobian f guess is a zero of f in each lane: one problem per lane, the last axis holding its k coordinates, as Root.newton is elementwise for one.val · jera
- Rune.laneslanes a x, inside the map named a of n lanes, is every lane's x stacked on a new leading axis of length n.val · rune
- Quill.Kernel.languagelanguage k is the language k runs, such as "ocaml".val · quill
- Norn.Dist.laplacelaplace ~loc ~scale has density exp (-|z|) / (2 scale).val · norn
- Sowilo.laplacianlaplacian img computes the Laplacian (sum of second spatial derivatives): the sum of the second x and y derivatives, each from sobel's kernel of aperture ksize with the binomial difference taken twice (for ksize 1 and 3, [1 -2 1] in place of [-1 0 1]).val · sowilo
- Vega.lars_initlars_init is sgd_init: LARS keeps SGD's momentum velocity.val · vega
- Vega.lars_steplars_step p ~lr st ~params ~grads is (params', st') after one LARS step (You, Gitman and Ginsburg, 2017), for large-batch SGD: each leaf is a layer whose step is scaled by the ratio of its weights' norm to its update's norm.val · vega
- Talon.Expr.lastlast a is a's last non-null value in frame order.val · talon
- Ymir.Direction.latlat d is the latitude in [−π/2, π/2] radians, atan2 z ρ with ρ = √(x² + y²).val · ymir
- Tolk.Decomp_op.late_patternslate_patterns ~disable_fast_idiv ops rewrites, for a target with the operations ops and rendered by the context:val · tolk
- Rig.Submission.launchlaunch ~refs fn ~params runs fn once, over the grid of groups of threads that its block in the run sets (block), with the params bytes of parameters of that block and each of refs (defaults to [||]) placed (ref), which launch reads once.val · rig
- Rig_nv_abi.LaunchKernels set up for launch on a GPU.module · rig
- Tolk.Shape.launchlaunch calls is the call of the launch of the kernel calls calls: its arguments are those of calls, each once, storage first, and each member calls its kernel, compiled on its own, on the launch's parameters at those arguments, a variable its kernel holds in an argument's slot becoming a nameless parameter of that slot.val · tolk
- Tolk.Shape.launch_dimslaunch_dims p vars is p's global and local sizes with each variable named in vars replaced by its value (sym_infer).val · tolk
- Tolk.Shape.launch_memberslaunch_members c is the calls of the members of the launch c calls, each on c's arguments: launch_members (launch cs) is cs.val · tolk
- Rig_remote.launchedlaunched () starts the job that a launcher set up for this process, or is None if RIG_REMOTE_REPORT is not in the environment.val · rig
- Hugin.layerlayer fs draws the figures of fs over one another, the first at the bottom, as one figure whose children share every scale (Scopes).val · hugin
- Kaun.Layer_normLayer normalization over the feature axis.module · kaun
- Layers and ModelsKaun has no layer abstraction.guide · kaun
- Hugin.layoutlayout ~prev ~theme size r lays r out at size in theme, Theme.default by default: it measures text in the theme's faces, chooses and freezes the ticks of each guide, solves the grid, places the guides and builds each panel's projection.val · hugin
- Hugin.LayoutLaid-out figures.module · hugin
- Hugin_text.Text.LayoutA layout is a text set in a list of faces at a size: lines of glyph runs placed about an origin, the point its alignment names.module · hugin
- Nx_array.layoutlayout a is a's layout.val · nx
- Nx_array.LayoutLayouts.module · nx
- Rig_amd_abi.Counter.layoutThe type for a run's samples: each counter's values, one after another.type · rig
- Rig_amd_abi.Counter.layoutlayout g names is the samples of a run of g counting names, in order.val · rig
- Talon.Column.layoutThe type for layouts.type · talon
- Talon.Column.layoutlayout c is c's layout, in O(1).val · talon
- Tolk.Device.Tiny_elf.layoutlayout params is iter_sig of the parameters params, in their order: how a kernel's arguments lie when packed, on every target.val · tolk
- Ymir_fits.Fits.Table.layoutThe type for how a column reads: read gives the data case of the same name.type · ymir
- Nx.lbetalbeta a b is log B(a, b) = lgamma a + lgamma b - lgamma (a + b) for a, b ≥ 0, computed without that sum's cancellation: lbeta 1e10 1.val · nx
- Jera.Minimize.lbfgslbfgs ~memory ~linear is limited-memory BFGS (Liu and Nocedal, 1989), for a smooth objective of many unknowns.val · jera
- Vega.lbfgs_initlbfgs_init p params is the initial state at params, with an empty history of history pairs (default 10).val · vega
- Vega.lbfgs_ptreelbfgs_ptree p is the structure of an L-BFGS state over parameters of structure p.val · vega
- Vega.lbfgs_statetype · vega
- Vega.lbfgs_steplbfgs_step p ~lr grads st is the state after one step from st.params, where the objective's gradient is grads.val · vega
- Talon.Join.lele l r is lt, as Expr.(<=) compares.val · talon
- Tolk.Ops.Elementwise.lele x y is logical_not (gt x y).val · tolk
- Tolk.Ops.lele x y is logical_not (gt x y).val · tolk
- Tolk.Ops.Upat.lele x y is logical_not (gt x y).val · tolk
- Nx.Ptree.Walk.leafleaf c x walks x, a position of the parameter, at c's path.val · nx
- Rig_program.leafThe type for what only a loaded program knows: a word of its memory or its images, filled once at load.type · rig
- Kaun.Fn.leaky_reluleaky_relu x is x where x > 0 and negative_slope * x elsewhere.val · kaun
- Ymir.Time.leap_seconds_known_untilleap_seconds_known_until is 0h UTC of the day IERS Bulletin C says the table of TAI − UTC that ymir carries expires.val · ymir
- Jera.Split.leapfrogleapfrog is K(h/2) D(h) K(h/2), Störmer–Verlet: order 2.val · jera
- Learning Rate SchedulesA learning rate schedule controls how the learning rate changes over the course of training.guide · vega
- Tolk.Dtype.least_upperleast_upper dts is the least data type by compare that every data type of dts promotes to.val · tolk
- Tolk.Dtype.least_upper_floatleast_upper_float dt is the float that dt promotes to: Weak_float for Weak_int, dt itself for a float, and least_upper [dt; default_float ()] otherwise.val · tolk
- Kaun.Init.lecun_normallecun_normal samples a truncated normal with target variance 1 / fan_in.val · kaun
- Kaun.Init.lecun_uniformlecun_uniform samples from U(-limit, limit) with limit = sqrt (3 / fan_in).val · kaun
- Hugin.legendlegend ~side ~ncol ~show ~title name is the legend of the scales named name in the scope of the figures it is layered with, in place of the one generated for each (Guides), placed on side of the panels that read the scales, `Right by default, drawn iff show, true by default, and titled title in place of the titles of the scales' channels, by default none: legend ~title:"seed" "color".val · hugin
- Brot.Encoding.lengthlength enc is the number of tokens in enc.val · brot
- Fehu.Collect.lengthlength traj is the number of transitions in traj.val · fehu
- Hugin.Mark.lengthlength rows is the number of rows of rows.val · hugin
- Hugin_font.Run.lengthlength r is the number of glyphs of r.val · hugin
- Hugin_gg_kit.Ring2.lengthlength r is the number of points of r.val · hugin
- Hugin_kit.Scale.lengthlength s is the length of the domain of s in the units its normalisation divides: |T b - T a| for a continuous domain [a;b] and the transform T (Continuous scales), in seconds on a temporal scale, and n + p on a band scale of n categories and padding p, the steps that [0;1] holds (Categorical scales), or 0.val · hugin
- Munin.Metric.lengthlength s is the number of samples of s.val · munin
- Nx_ragged.lengthlength r is the number of rows of r.val · nx
- Quill.Doc.lengthlength d is the number of cells in d.val · quill
- Rig.Buffer.lengthlength b is the number of b's bytes.val · rig
- Rig_amd_abi.Thread_trace.lengthlength g ~buffer w is the bytes a shader engine of g wrote into its buffer at address buffer, from the word w stop stored for it.val · rig
- Rig_bnxt.Region.lengthlength r is the number of bytes of r.val · rig
- Rig_pci.Space.lengthlength s is s's number of addresses.val · rig
- Rig_pci.Window.lengthlength w is w's size in bytes.val · rig
- Talon.Column.lengthlength c is the number of rows of c.val · talon
- Talon.Expr.Str.lengthlength a is the number of Unicode scalar values of a, as int64.val · talon
- Hugin_kit.Dash.lengthslengths d is the lengths of d as given to v, [] if solid.val · hugin
- Nx_ragged.lengthslengths r is the length of each row of r.val · nx
- Tolk.Bigint.leqleq n0 n1 is n0 <= n1.val · tolk
- Nx.lerplerp a b w is a + w (b - a), by one fma: a + w (b - a) where w < 0.5, else b + (w - 1) (b - a), b - a rounded once first.val · nx
- Norn.Bij.lessless ~high maps u to high - exp u, onto (-inf, high).val · norn
- Nx.lessless a b is a < b.val · nx
- Nx_kernel.Prog.E.lessval · nx
- Nx_wide.lessless a b is a < b, exactly.val · nx
- Nx.less_equalless_equal a b is a <= b.val · nx
- Nx_kernel.Prog.E.less_equalval · nx
- Nx.less_equal_sless_equal_s x c is less_equal x c.val · nx
- Nx.less_sless_s x c is less x c.val · nx
- Tolk.Ops.Axis_type.letterletter a is the letter that names a in kernel names: "g", "l", "w", "L" for Weak and Loop, and "u".val · tolk
- Jera.Minimize.levenberg_marquardtlevenberg_marquardt r ~linear is for |r x|² / 2 given by its residual r, a value of structure r whose float tensors are one vector.val · jera
- Nx.lexsortlexsort keys is the stable permutation that sorts the keys of keys: n elements of a 1-d keys, or n rows of a 2-d one, compared column 0 first, each in the sort order.val · nx
- Nx.lgammalgamma x is log |Γ(x)|, on every float: inf at 0., the negative integers and inf, NaN at -inf.val · nx
- Talon.limitsThe type for display limits.type · talon
- Talon.limitslimits is { head = 5; tail = 5; columns = 12; width = 32 }, the limits of pp.val · talon
- Hugin.lineline ~y () draws a curve through the points of each series of its rows, in order along the last axis of the mark's shape.val · hugin
- Jera.Quad.Range.lineline c is the whole line, around c at unit scale.val · jera
- Rig_bin.LineThe lines a launched process and a machine's half write to their launcher.module · rig
- Hugin_font.Font.line_gapline_gap f is the extra space f puts between the descent of a line and the ascent of the next, from the table ascent reads.val · hugin
- Tolk.Codegen.line_rewriteline_rewrite l m ctx rewrites the instruction list l, each node after its sources: each node is rebuilt on the results of its sources, and a rule of m that matches the rebuilt node gives its result, the node its users see, and the instructions that replace it; a node no rule matches stays.val · tolk
- Hugin_gg.Path.line_toline_to pt p is p with a line from the current point to pt.val · hugin
- Hugin_gg.Affine.linearlinear m is m without its translation: the map of vectors that m gives, m with x0 and y0 zero.val · hugin
- Hugin_kit.Curve.linearlinear joins consecutive points with straight segments.val · hugin
- Hugin_kit.Scale.linearlinear () is a linear scale.val · hugin
- Jera.Grid.linearlinear ~axes values is the multilinear interpolant: degree 1 along each axis.val · jera
- Jera.LinearLinear systems.module · jera
- Jera.Piecewise.linearlinear x y is the broken line through the samples: degree 1.val · jera
- Kaun.LinearDense (fully connected) layers, and the model-as-record pattern.module · kaun
- Vega.Schedule.linearlinear ~init_value ~end_value ~steps interpolates linearly from init_value to end_value over steps.val · vega
- Ymir.Transform.linearlinear m maps x, in Unit.one, to m · x in m's unit: CD, or PC in Unit.one.val · ymir
- Linear AlgebraNx provides linear algebra over batches of matrices and Fourier transforms.guide · nx
- Rune.linear_transposelinear_transpose p q f params is the transpose of f, a function linear in its argument: the function that maps a cotangent w of the result to u such that <u, v> = <w, f v> for every v (Derivatives).val · rune
- Rune.linearizelinearize p q f params is (f params, jvp_at), jvp_at mapping a tangent of params to the result's tangent, as jvp computes it at params: f runs once, in forward mode, recording the operations on tangents, and jvp_at evaluates that record, as often as it is applied, without running f again.val · rune
- Tolk.Codegen.linearizelinearize ast ren is the program of the kernel's sink ast for ren up to its instructions: an Op.t.Program whose sources are the first two of to_program's, the lowered sink and the Op.t.Linear of its instructions, and whose argument is its Ops.program_info.val · tolk
- Tolk.Linearizer.linearizelinearize sink is the nodes sink reaches outside the bodies of its calls (Ops.toposort ~calls:Skip), each after its sources, sink last.val · tolk
- Tolk.LinearizerOrdering a kernel into a list of instructions.module · tolk
- Rig_mlx5.linklink nic is the link layer of nic's port.val · rig
- Rig_pci.Memory.linkThe type for a GPU's place in a fabric of GPUs, which reach each other's memory over direct links.type · rig
- Rig_remote_proxy.LinkThis process's ends of its job's connections.module · rig
- Tolk_engine.linklink ~devices ~bound ~storage linear is the compiled schedule linear (Tolk.Realize.compile_linear, compiled for the target of each device) loaded on devices, which maps each device name of linear to its device.val · tolk
- Tolk_engine.link_calllink_call ~devices call is the call call of a compiled program (Tolk.Op.t.Program) linked as a schedule of one call, whose kernel is a submission of its own.val · tolk
- Nx.linspacelinspace dt ~endpoint start stop n is the host value of n values evenly spaced from start to stop, stop included where endpoint (default true).val · nx
- Nx.linspace_atlinspace_at p dt ~endpoint start stop n is linspace dt ~endpoint start stop n at p.val · nx
- Vega.lion_initlion_init p params is the initial state for optimizing params: an all-zero average and step = 0.val · vega
- Vega.lion_ptreelion_ptree p is Nx.Ptree.nest (module Lion_state) p.val · vega
- Vega.lion_stateThe state for lion_step: the moving average of the gradients, with the shape of the parameters, and the number of completed steps as a scalar tensor.type · vega
- Vega.Lion_stateThe structure of Lion states.module · vega
- Vega.lion_steplion_step p ~lr st ~params ~grads is (params', st') after one Lion step (Chen et al., 2023).val · vega
- Nx.Ptree.listlist a reports the length as Walk.list does and walks the i-th element with a at Index i.val · nx
- Nx.Ptree.Walk.listlist walk c l reports l's length at c's path, then walks its i-th element with walk at c's path then Index i, from the first.val · nx
- Talon.Expr.listlist es is the list of es' values at each row, in order, as a list of the type where the types of es meet (Type.common), with a null element for each null value.val · talon
- Talon.Kind.listlist k reads list columns whose elements k reads.val · talon
- Talon.Type.listlist t is List of t.val · talon
- Rig_remote.listenlisten ~key host port listens at host and port for the processes of one job that prove key.val · rig
- Hugin_literal.LiteralString literals for debugging printers.module · hugin
- Talon.Expr.Str.literalliteral s matches s anywhere in the text.val · talon
- Ymir_units.Unit.litrelitre is 10-3 m3.val · ymir
- Norn.Dist.lkj_choleskylkj_cholesky n ~concentration is the distribution of the Cholesky factor L of an n × n correlation matrix C = L Lᵀ whose density is proportional to det C^(η - 1), η the concentration in (0, inf) (Lewandowski, Kurowicka and Joe, Generating random correlation matrices based on vines and extended onion method, 2009): uniform over correlation matrices at η = 1, concentrated about the identity above.val · norn
- Kaun.Rope.llama3llama3 ...val · kaun
- Tolk.Ops.Elementwise.O.lnotval · tolk
- Tolk.Ops.O.lnotval · tolk
- Tolk.Ops.Upat.lnotval · tolk
- Tolk.Ops.Upat.O.lnotval · tolk
- Nx_wide.lolo w is w - hi w, exactly.val · nx
- Tolk.Helpers.lo32lo32 x is the low 32 bits of x.val · tolk
- Nx_kernel.loadThe type for loads.type · nx
- Quill_markdown.loadload path is the notebook in the file path, read as of_string does, and Files if its first line is <!-- quill:figures files -->, Inline otherwise.val · quill
- Rig.Image.loadload d binary loads binary on d, such as a code object for AMD, a cubin for NV, a CUDA module for CUDA, a metallib for Metal or a relocatable ELF object for host (Linking).val · rig
- Rig_program.loadload ~rails t devices is t loaded on devices, all of one machine (Rig.host_of).val · rig
- Tolk.Ops.loadload p rest is the Op.t.Load through p.val · tolk
- Tolk.Ops.Upat.loadval · tolk
- Tolk.Simplify.load_collapseload_collapse red is the sum red over one range (Op.t.Reduce of Op.t.Add) computed without the range, by pm_reduce_collapse's closed forms and by taking a value selected where an integer index that does not depend on the range equals the range, cast to a type that holds its values or not, at that index, as when indexing a tensor with another one: the value there where the index is within the range, and 0 elsewhere.val · tolk
- Kaun_hf.load_configload_config repo_id is the parsed config.json of repo_id.val · kaun
- Quill_top.load_errorWhy load_package did not load a package.type · quill
- Nx_io.load_ggufload_gguf path is the contents of the GGUF file path.val · nx
- Nx_io.load_imageload_image path is the image of the PNG or JPEG file path, recognised by its contents, with the channels the file holds.val · nx
- Nx_io.load_npyload_npy path is the array of the npy file path.val · nx
- Nx_io.load_npzload_npz path is the entries of the npz file path.val · nx
- Quill_top.load_packageload_package pkg resolves the findlib package pkg and all its transitive dependencies, adds their directories, and dynamically loads their bytecode archives.val · quill
- Rig_program.load_receivedload_received ~rails b ~device is load ~rails of the description in b, the binary another process's load sent to this machine's host, on the devices that device gives for the ids b names.val · rig
- Kaun_hf.load_safetensorsload_safetensors repo_id is the tensors of repo_id's SafeTensors weights.val · kaun
- Nx_io.load_safetensorsload_safetensors path is the entries of the SafeTensors file path.val · nx
- Nx_io.load_txtload_txt dt path is the array of the text file path at dt.val · nx
- Rig_program.loadedThe type for loaded programs.type · rig
- Rig.Localmodule · rig
- Rig_mlx5.Region.locallocal r at n is the n bytes at offset at of r, as this process's entries name them.val · rig
- Rig_mlx5_abi.Entry.localThe type for memory this process registered with its NIC.type · rig
- Talon.Time.localThe index of a wall clock.type · talon
- Talon.Tz.localThe type for the answers of local: the UTC offsets at which a zone's wall clock reads a local time t.type · talon
- Talon.Tz.locallocal z t is the offsets o at which the wall clock of z reads t, that is, those with offset_s z (t - o) = o.val · talon
- Rig_nv_abi.Launch.local_byteslocal_bytes l is the local memory each thread of a launch needs: the kernel's stack and the 576 bytes the driver reserves.val · rig
- Rig_nv_abi.Local_memoryKernels' local memory.module · rig
- Rig_nv_abi.Method.local_memorylocal_memory addr ~per_tpc gives kernels the local memory at addr, per_tpc bytes for each texture processing cluster (Local_memory.t.per_tpc), and lets every streaming multiprocessor use it.val · rig
- Rig_nv_abi.Method.local_memory_windowlocal_memory_window addr makes kernels' local memory appear at the address addr.val · rig
- Hugin.Theme.localelocale th is the locale of th.val · hugin
- Hugin_kit.LocaleLocales.module · hugin
- Talon.Expr.Temporal.localizelocalize z ~ambiguous ~gap a is the instant at which z's wall clock reads a, in a's unit: a - o for the offset o of Tz.local's Unique o.val · talon
- Talon.Expr.Temporal.localize_orderedlocalize_ordered z ~ambiguous ~gap a is localize where the order of each frame's rows (Frames), which must be time order, decides the pass of a reading that z repeats.val · talon
- Rig_amd.locatelocate r is where r lies.val · rig
- Rig_cuda.locatelocate r is where r lies.val · rig
- Rig_edge.Driver.locatelocate r is where r lies.val · rig
- Rig_metal.locatelocate r is where r lies.val · rig
- Rig_nv.locatelocate r is where r lies.val · rig
- Rig_remote_proxy.locatelocate r is where r lies.val · rig
- Rig_edge.locationThe type for where a region lies.type · rig
- Rig_amd.Lockmodule · rig
- Rig_nv.Lockmodule · rig
- Hugin_kit.Scale.loglog ~base () is a logarithmic scale in base base, which defaults to 10.val · hugin
- Munin.Session.loglog m ~step v is log_metrics s ~step ?time [ (m, v) ] for m's session s.val · munin
- Nx.loglog x is the natural logarithm: -inf at 0., NaN below.val · nx
- Nx_kernel.Prog.E.logval · nx
- Nx.log1plog1p x is log (1 + x), accurate near 0: -0.val · nx
- Nx_kernel.Prog.E.log1pval · nx
- Nx.log2log2 x is the base-2 logarithm.val · nx
- Nx_kernel.Prog.E.log2val · nx
- Tolk.Ops.Elementwise.log2val · tolk
- Tolk.Ops.log2val · tolk
- Tolk.Ops.Upat.log2val · tolk
- Munin.Session.log_artifactlog_artifact s ~name path publishes the file or directory path as a new version of the artifact name, produced by s's run at step, and is that version.val · munin
- Nx.log_betainclog_betainc a b x is log I_x(a, b) (betainc), finite for finite a, b > 0 and x in (0, 1], where betainc underflows included.val · nx
- Nx.log_betaincclog_betaincc a b x is log (1 - I_x(a, b)) (betaincc), finite for finite a, b > 0 and x in [0, 1).val · nx
- Norn.Dist.log_densitylog_density d x is Nx.sum (factors d x), a scalar.val · norn
- Norn.Gaussian.log_densitylog_density u g is g's log density at a position: one value per row of the leading axis.val · norn
- Norn_model.log_densitylog_density m y is the log density of the coordinates conditioned on y: the prior of the latent values, each site's log-determinant and the likelihood of y, factors included.val · norn
- Norn.Evidence.log_evidencelog_evidence z is the estimate of ln Z, a scalar.val · norn
- Munin.Session.log_filelog_file s ~step ~key path copies the file path into s's run and appends a file record that names the copy at step (Files a run logged).val · munin
- Nx.log_gammainclog_gammainc a x is log P(a, x) (gammainc), finite for a, x > 0 where P underflows: log_gammainc 1000.val · nx
- Nx.log_gammaincclog_gammaincc a x is log Q(a, x) (gammaincc), finite for a > 0 wherever x is.val · nx
- Norn_model.log_jointlog_joint m y p is the log density of the values p and the observations y, with no log-determinant: the argument's type says which.val · norn
- Norn_model.log_likelihoodlog_likelihood m y is the log likelihood of y, factors included: log_density m y c = log_prior m c + log_likelihood m y c.val · norn
- Munin.Session.log_metricslog_metrics s ~step pairs appends one metrics record that holds a sample of each metric of pairs, all at step and time.val · munin
- Nx.log_ndtrlog_ndtr x is log (ndtr x), finite where ndtr x underflows while x²/2 is finite: down to x about -1.9e154 at float64 and -2.6e19 at float32, and -inf below.val · nx
- Norn_model.log_priorlog_prior m is the prior density of the coordinates: the latent sites' terms and log-determinants.val · norn
- Nx.log_softmaxlog_softmax ~axes ~scale ~where x is the logarithm of softmax, computed as (scale x - m) - log Σ exp (scale x - m) for the lane's greatest m: an entry far below m keeps its digits where log (softmax x) would be -∞.val · nx
- Nx_kernel.Prog.E.logandval · nx
- Tolk.Bigint.logandlogand n0 n1 is the bitwise and of n0 and n1.val · tolk
- Nx.logcumsumexplogcumsumexp ~axis x is, at each index i along axis, log Σ_{k ≤ i} exp x_k; without axis, over its elements in C order up to that one, in x's shape.val · nx
- Nx.logical_andlogical_and a b is one where a and b are both not zero, zero elsewhere, in their dtype.val · nx
- Nx.logical_notlogical_not x is one where x is zero, zero elsewhere, as logical_and.val · nx
- Tolk.Ops.Elementwise.logical_notlogical_not x is eq (cast x Bool) false.val · tolk
- Tolk.Ops.logical_notlogical_not x is eq (cast x Bool) false.val · tolk
- Tolk.Ops.Upat.logical_notlogical_not x is eq (cast x Bool) false.val · tolk
- Nx.logical_orlogical_or a b is one where a or b is not zero, as logical_and.val · nx
- Nx.logical_xorlogical_xor a b is one where exactly one of a and b is not zero, as logical_and.val · nx
- Norn.Dist.logisticlogistic ~loc ~scale has density exp (-z) / (scale (1 + exp (-z))²).val · norn
- Nx.logmeanexplogmeanexp ~axes x is log (Σ exp x / n): logsumexp less log n, taken before the one rounding to the dtype.val · nx
- Norn.Dist.lognormallognormal ~loc ~scale is the distribution of exp x, x normal.val · norn
- Tolk.Bigint.lognotlognot n is the bitwise complement of n, -n - 1.val · tolk
- Nx_kernel.Prog.E.logorval · nx
- Tolk.Bigint.logorlogor n0 n1 is the bitwise or of n0 and n1.val · tolk
- Nx.logspacelogspace dt ~endpoint ~base start stop n is the host value of base (default 10.) to the powers linspace float64 ~endpoint start stop n, each power computed at float64 and stored once in dt: logspace float32 0.val · nx
- Nx.logspace_atlogspace_at p dt ~endpoint ~base start stop n is logspace dt ~endpoint ~base start stop n at p.val · nx
- Nx.logsumexplogsumexp ~axes x is log Σ exp x, computed as m + log Σ exp (x - m) for the elements' greatest m, so no term overflows: -∞ of no element or of elements all -∞, +∞ where an element is +∞ and none is NaN.val · nx
- Nx_kernel.Prog.E.logxorval · nx
- Tolk.Bigint.logxorlogxor n0 n1 is the bitwise exclusive or of n0 and n1.val · tolk
- Ymir.Direction.lonlon d is the longitude in [+0, 2π) radians, atan2 y x.val · ymir
- Talon.Kit.longerlonger ~name ~value sel q is q with the columns of sel replaced by two columns: name, the categorical whose dictionary is the selected names in order, and value, their cells at the type where their types meet (Type.common), one row per cell, nulls included.val · talon
- Ymir.Direction.lonlatlonlat f ~lon ~lat is the unit vectors (cos b cos l, cos b sin l, sin b) at longitude l and latitude b, their batch axes broadcast.val · ymir
- Norn.Diag.looThe type for leave-one-out cross-validations.type · norn
- Norn.Diag.looloo ll is the Pareto-smoothed leave-one-out cross-validation (Vehtari, Gelman and Gabry 2017) of the log likelihoods ll, of shape [chain; draw; ...]: log p(y_i | θ_s) of each point y_i at each posterior draw θ_s, as Draws.map of Norn_model.pointwise gives them.val · norn
- Ymir.Cosmology.lookback_timelookback_time c z is the time light from z has travelled.val · ymir
- Ymir_units.Vocabulary.lookuplookup voc s is the unit s names in voc: the unit of the symbol s, otherwise an SI prefix followed by a prefixable symbol, the prefix's power of ten times the symbol's unit.val · ymir
- Tolk.Ops.looploop axis is the unbounded loop of identity [axis], which a backedge (Op.t.Backedge) repeats.val · tolk
- LoopsA loop applies a step to a carry.guide · rune
- Kaun.LossScalar training objectives: regression and classification losses.module · kaun
- LossThis page shows what a program sees when a device fails.guide · rig
- Vega.Loss_scaleLoss scales for float16 training, after JAX's jmp.module · vega
- Rig.LostLost (d, why) is raised by a use that the loss of d makes invalid (Memory after a loss).exception · rig
- Rig.lostlost d is Some why if d is lost, where why is the reason its Lost carries, and None otherwise.val · rig
- Norn.Gaussian.low_ranklow_rank u ~mean ~scale ~directions ~variances is the Gaussian whose directions are directions, each tensor with a leading axis of k, from 0 to the number of elements, orthonormalised, and whose variances along them are variances, of shape [k].val · norn
- Talon.Expr.Str.lowerlower a is a in lowercase by Unicode's Default Case Conversion (Unicode 18.0.0, chapter 3, 3.13): each scalar value maps to its full lowercase mapping, from SpecialCasing.txt where it has an unconditional entry and from UnicodeData.txt otherwise, so the text can change byte length (U+0130 lowers to i and U+0307, three bytes for two).val · talon
- Tolk.Realize.lower_and_compilelower_and_compile ~search ~targets linear is linear with the body of each call that is a kernel (Op.t.Sink with Ops.kernel_info), outside the bodies of the calls, replaced by its compiled program (Codegen.to_program), for the renderer Device.renderer picks for the target targets d of the device d of the call's first storage argument (get_call_arg_uops): a kernel's parameters name no device.val · tolk
- Brot.Normalizer.lowercaselowercase is the full Unicode lowercase mapping.val · brot
- Vega.lrlr v is the constant rate v, Schedule.constant v, which serves any parameter dtype and set: Vega.adam_step p ~lr:(Vega.lr 1e-3).val · vega
- Nx.lshiftlshift x n is x shifted n bits toward its high end: x 2{^n} modulo 2{^w} for a dtype of w bits, 0 once n reaches w.val · nx
- Nx.lstsqlstsq dt a b is (x, residuals, rank, s): x the least-squares solution of matmul a x ≈ b of least norm, whatever a's shape and rank, from s, svd's singular values of a in dt.val · nx
- Talon.Join.ltlt l r matches the rows whose left column l comes before the right column r, as Expr.(<) compares.val · talon
- Tolk.Bigint.ltlt n0 n1 is n0 < n1.val · tolk
- Tolk.Ops.Elementwise.ltval · tolk
- Tolk.Ops.ltval · tolk
- Tolk.Ops.Upat.ltval · tolk
- Nx.lulu a is (perm, l, u) where row i of matmul l u is row perm.{i} of a, l is […; m; k], lower triangular with ones on its diagonal, and u is […; k; n], upper triangular; perm is […; m].val · nx
- Nx_cpu.lulu ~lu ~pivots ~perm a stores into lu the unit lower triangular L below its diagonal, its ones unstored, and the upper triangular U on and above it.val · nx
- Nx_kernel.S.lulu ~lu ~pivots ~perm a stores into lu the unit lower triangular L below its diagonal, its ones unstored, and the upper triangular U on and above it.val · nx
- Nx_kernel.Shape.lulu a is the shapes of lu, pivots and perm: a, […; k] and […; m].val · nx
- Ymir_units.Unit.lumenlumen is cd sr.val · ymir
- Ymir.Cosmology.luminosity_distanceluminosity_distance c ~observed z is (1 + observed) DM(z): z places the source, observed is the redshift the observer measures, z by default.val · ymir
- Ymir_units.Unit.luminous_efficacyluminous_efficacy is Kcd, 683 cd sr W-1.val · ymir
- Ymir_units.Unit.luxlux is cd sr m-2.val · ymir
m
- Hugin.Mark.m4m4 keeps, of the series (series) whose channels other than their positions and facets are constant along them, when they have more than four rows per device-pixel column of the panel and per run, the first, last, lowest and highest rows of each run of consecutive rows whose x lies in one column, and the first row of each run of dropped rows.val · hugin
- Rig_bin.Address.machinemachine s is the host s names, its brackets taken off.val · rig
- Rig_pci.Function.machinemachine f is the machine f is on.val · rig
- Rig_pci.MachineMachines, their PCI functions, and transports to other machines.module · rig
- MachinesA program uses the devices of other computers through Rig as it uses its own.guide · rig
- Kaun.Loss.maemae predictions targets is the reduction over all elements of the absolute error |predictions - targets|.val · kaun
- Hugin_kit.Scheme.magmamagma runs from black through purple and orange to light yellow.val · hugin
- Nx.magnitudemagnitude dt z is z's modulus, hypot of its parts computed in the wider of dt and the parts' format, then stored in dt.val · nx
- Rig_pci.Page_table.main_poolmain_pool t is the number of bytes of the main pool, memory less the boot pool and the tables' pool.val · rig
- Fehu_envs.Cartpole.makemake () is a cart-pole environment.val · fehu
- Fehu_envs.Grid_world.makemake () is a grid world environment.val · fehu
- Fehu_envs.Mountain_car.makemake () is a mountain car environment.val · fehu
- Fehu_envs.Random_walk.makemake () is a random walk environment.val · fehu
- Kaun.Attention.Cache.makemake ~slots ~kv_heads ~head_dim dtype at is a cache of slots slots holding zeros, made at at (a cache is made from no value, and its pools are the largest arrays a decoder keeps, so they are made where they will lie).val · kaun
- Kaun.Attention.makemake ~key ~embed_dim dtype is a fresh layer attending over embed_dim features, drawn from key and lying where it lies: q maps embed_dim to q_dim, k and v map embed_dim to kv_dim, and out maps q_dim back to embed_dim.val · kaun
- Kaun.Cache_index.makemake ~pos ~table () is the index of tokens at positions pos in sequences held at the slots table, with:val · kaun
- Kaun.Conv.makemake ~key ~in_channels ~out_channels ~kernel_size:(kh, kw) dtype is a fresh layer of out_channels filters of shape in_channels × kh × kw, drawn from key and lying where it lies, the filters and the bias from two keys Nx.Rng.split makes of it, with:val · kaun
- Kaun.Embedding.makemake ~key ~vocab ~dim dtype is a fresh table of vocab rows of dim features, drawn from key and lying where it lies.val · kaun
- Kaun.Layer_norm.makemake ~dim dtype at is a fresh identity normalization over dim features at at: gamma all ones, beta all zeros.val · kaun
- Kaun.Linear.makemake ~key ~inputs ~outputs dtype is a fresh layer mapping inputs features to outputs features, drawn from key and lying where it lies, with:val · kaun
- Kaun.Rms_norm.makemake ~dim dtype at is a fresh normalization over dim features at at, with gamma all ones.val · kaun
- Kaun.Rope.makemake ~head_dim ~context () is the standard schedule: pair i has frequency theta ** (-2 i / head_dim).val · kaun
- Quill.Kernel.makemake ~language ~on_event ~execute ~interrupt ~complete ~status ~shutdown () is the kernel that runs the cells of language with these functions of its backend:val · quill
- Rig.Submission.makemake ~keep ~fixed d parts is a submission of parts on d.val · rig
- Rig_amd.makemake ~word p is a device of the GPU p reaches, whose timeline word is the 8 bytes at the host address word (Rig_edge.opener).val · rig
- Rig_bnxt.Qp.makemake nic is an unconnected queue pair of nic, whose sends start at a random packet sequence number.val · rig
- Rig_mlx5.Cq.makemake nic n is a completion queue of nic with room for n completions, n rounded up to a power of two no smaller than 2.val · rig
- Rig_mlx5.Qp.makemake cq ~receives n is an unconnected queue pair of cq's NIC with room for n entries, n rounded up to a power of two, and for receives receives, rounded up to a power of two no smaller than 4, or no receive ring if receives is 0.val · rig
- Rig_nv.makemake ~word p is the device of the GPU p reaches, whose timeline word is the 8 bytes at the host address word (Rig_edge.opener).val · rig
- Rig_nv_abi.Launch.makemake g k is Ok l, k set up for launch on g, or Error msg if k takes more than a launch can, msg saying which:val · rig
- Rig_nv_abi.Local_memory.makemake g n is the local memory of g for kernels whose threads need n bytes each (Launch.local_bytes):val · rig
- Rig_nv_abi.Qmd.makemake l d is the descriptor of a launch of l with the values d: 256 bytes for version 3 and 384 for version 5.val · rig
- Rig_pci.Driver.makemake ~name ~memory_bar ~siblings ~nodes ~unreleased ~teardown_ms ~reset ~ready matches is the driver of the functions f with matches f, named after name (name).val · rig
- Rig_pci.Machine.makemake ~name ops is the machine named name that ops reaches.val · rig
- Rig_remote_proxy.Link.makemake j fd ~name ~peer is a link of j over the connected socket fd, whose handshake has admitted it, to the process peer named name: its machine's name, which starts its reasons.val · rig
- Rune.Probe.makemake s is a fresh probe of values of s's structure, distinct from every other.val · rune
- Rune.Total.makemake () is a fresh total, distinct from every other.val · rune
- Norn.Draws.mapmap a b f d is f applied to every draw of d, a value of 'a without the two leading axes, the results stacked on them.val · norn
- Nx.Ptree.mapmap s f x is x with each array a at path p replaced by f p a.val · nx
- Nx.Ptree.Payload.mapmap (module U) f x is x with each payload a at path p replaced by f p a, in walk order.val · nx
- Nx_cpu.mapmap p ~dsts loads stores p's output k at each index into dsts.(k), of that output's dtype.val · nx
- Nx_kernel.S.mapmap p ~dsts loads stores p's output k at each index into dsts.(k), of that output's dtype.val · nx
- Nx_ragged.mapmap f r is r with values f (values r) and r's offsets.val · nx
- Rig_pci.Function.mapmap f i is a window on length bytes (defaults to the rest of the BAR) of f's BAR i from byte off (defaults to 0), until unmap or release.val · rig
- Rig_pci.Page_table.mapmap t ~va tg ranges maps the physical ranges of tg, in order, from va on, creating the tables it needs, and flushes, whatever the flush returns.val · rig
- Tolk.Worker.mapmap f l is List.map f l, computed on the calling domain and on up to p - 1 other domains, where p is the value of Setting.parallel.val · tolk
- Ymir_fits.Fits.Value.mapmap f g v reads with v then f, and prints with g then v.val · ymir
- Ymir_units.Quantity.mapmap f q is f of q's payload, in q's unit.val · ymir
- Nx.Ptree.map2map2 s f x y is x with each array a at path p replaced by f p a b, b y's array at p.val · nx
- Nx.Ptree.Payload.map2map2 (module U) f x y is x with each payload a at path p replaced by f p a b, b y's payload at p.val · nx
- Ymir_units.Quantity.map2map2 f a b is f of a's payload and value (unit a) b, in a's unit.val · ymir
- Fehu.Env.map_actionmap_action ~action_space ~f env transforms actions before passing them to the inner environment.val · fehu
- Rig_amd.map_hostmap_host d p n maps for d the n bytes of host memory at p.val · rig
- Rig_cuda.map_hostmap_host d p n maps for d the n bytes of host memory at p.val · rig
- Rig_edge.Driver.map_hostmap_host d p n maps for d the n bytes of host memory at p.val · rig
- Rig_metal.map_hostmap_host d p n maps for d the n bytes of host memory at p.val · rig
- Rig_nv.map_hostmap_host d p n maps for d the n bytes of host memory at p.val · rig
- Rig_pci.Memory.map_hostmap_host m a n maps for the GPU the pages of its machine that hold the n bytes at a.val · rig
- Rig_remote_proxy.map_hostmap_host d p n maps for d the n bytes of host memory at p.val · rig
- Nx.map_itemmap_item f x is the host value of x's shape whose element at each index is f of x's element there, a float stored as Dtype.of_float says.val · nx
- Fehu.Env.map_observationmap_observation ~observation_space ~f env transforms observations.val · fehu
- Rig_amd.map_peermap_peer d d' r maps for d the memory of r, any memory of d'.val · rig
- Rig_cuda.map_peermap_peer d d' r maps for d the memory of r, any memory of d'.val · rig
- Rig_edge.Driver.map_peermap_peer d d' r maps for d the memory of r, any memory of d'.val · rig
- Rig_metal.map_peermap_peer d d' r maps for d the memory of r, any memory of d'.val · rig
- Rig_nv.map_peermap_peer d d' r maps for d the memory of r, any memory of d'.val · rig
- Rig_pci.Memory.map_peermap_peer m r maps r, a region of a GPU, m's own included, for the GPU of m, at the address r has there.val · rig
- Rig_remote_proxy.map_peermap_peer d d' r maps for d the memory of r, any memory of d'.val · rig
- Hugin.map_rangemap_range f c is c with f applied to each of its values after the role maps it into the range, and to the value of a constant: ~fill:(map_range Color.contrast recall) paints text that reads on the colour recall gives its cell.val · hugin
- Nx_kernel.map_refusalmap_refusal p ~rank ~dsts:dv dsts ~loads:lv loads is Some (r, why) where S.map's program p over loads into dsts, read by lv and dv, on a loop of rank rank, does not fit: r the refusal that names the misfit and why what does not fit, in words; None where all fit.val · nx
- Fehu.Env.map_rewardmap_reward ~f env transforms rewards after each step.val · fehu
- Nx.map_segmentsmap_segments ~segments ids f x is f applied to each position alone, the per-position map: at each position p of ids, the row f gives p's row of x with owner ids.{p} where ids.{p} is in [0, segments), and zeros otherwise, whatever x holds at p.val · nx
- Ymir.Grid.map_worldmap_world t g is g's cells seen in the world t maps g's world to: its transform followed by t.val · ymir
- Ymir.Observation.map_worldmap_world t o is o on Grid.map_world t (grid o): its data, variance, validity and area are o's.val · ymir
- Rig_pci.Window.mappedmapped w is true iff w is mapped into the process, so that address is an address of the process.val · rig
- Rig_edge.mappingThe type for a driver's answer to a request to map memory (Driver.map_host, Driver.map_peer).type · rig
- Rig_mlx5_abi.Uar.mappingmapping ~page i is the offset in the NIC's device file of system page i of the process's access regions, mapped uncached, on a host whose pages are page bytes.val · rig
- Rig_pci.Page_table.mappingThe type for mapped virtual ranges.type · rig
- MapsA map applies a function once per lane: lane k receives row k of each array of the arguments, and the lanes' results are stacked along a new axis 0.guide · rune
- Jera.Ode.marchmarch y m ~steps f ~at y0 is the state at each time of at, stacked on a new leading axis of each leaf, y0 first at at.(0).val · jera
- Jera.Sde.marchmarch y m ~steps ~drift ~diffusion w ~at y0 is the state at each time of at, stacked on a new leading axis of each leaf, y0 first, along the path w.val · jera
- Jera.Split.marchmarch s m ~steps ~kick ~drift ~at s0 is the state at each time of at, stacked on a new leading axis of each leaf, s0 first at at.(0).val · jera
- Tolk.Shape.margmarg u is the argument of the movement u.val · tolk
- Hugin.MarkMarks.module · hugin
- Quill.Session.mark_idlemark_idle id s marks cell id as idle.val · quill
- Quill.Session.mark_queuedmark_queued id s marks cell id as queued.val · quill
- Quill.Session.mark_runningmark_running id s marks cell id as running.val · quill
- MarksA mark draws rows of data.guide · hugin
- Kaun.Cache_index.maskmask index, of shape [| batch; seq; context |], is which columns of extend's seen each token sees: those that stand at or before its position, and under the index's window only the last of them.val · kaun
- Rig_edge_primitives.maskmask () blocks on the calling thread every signal but those a fault raises (rig_mask).val · rig
- Tolk.Ops.Upat.match_match_ p u is the naming of each way p matches u, in order: the alternatives of any in turn, the orderings of a commutative operation's operands in the order of their permutations, the first operand varying slowest, and each source's namings before the next source's; [] if p does not match.val · tolk
- Talon.Expr.Str.matchesmatches p a is true iff p matches a.val · talon
- Nx.matmulmatmul a b is the matrix product over the last two axes, leading axes broadcast, at a's dtype and accumulated as contract does.val · nx
- Nx.matmul_segmentsmatmul_segments ids x w is each position's row of x times the matrix of w its id names: at each position p of ids, p's row of x times w's matrix ids.{p}, or zeros where ids.{p} is not in [0, e), whatever that row holds.val · nx
- Ymir.Frame.matrixmatrix a b is M, [3; 3] on the host, with v_b = M · v_a: for quantities that are not directions, such as a velocity (M v) or a covariance (M Σ Mᵀ).val · ymir
- Ymir.Frame.matrix_atmatrix_at e a b is M, [...; 3; 3] beside and across e's instants, with v_b = M · v_a at each instant, for any two frames, as matrix is for fixed ones: R_b · R_aᵀ, with GCRS's axes ICRS's.val · ymir
- Nx.matrix_normmatrix_norm dt x is, in dt, the norm of order ord (default `Fro) of each matrix of x, dropping its last two axes unless keepdims.val · nx
- Nx.matrix_powermatrix_power a n is each square matrix of the batch to the power n: the identity for n = 0, and inv a to the power -n for a negative n.val · nx
- Nx.matrix_rankmatrix_rank a is the rank of each matrix of a, of a's batch shape: the number of its singular values above rtol times the greatest, as lstsq counts them, rtol defaulting to max m n times the dtype's machine epsilon, float32's for a narrower float.val · nx
- Nx.matrix_transposematrix_transpose x is x with its last two axes exchanged; x itself below two axes.val · nx
- Nx.maxmax ~axes x is the greatest element, as maximum orders them: -0.val · nx
- Talon.Expr.maxmax a is the greatest of a's values, like min.val · talon
- Tolk.Bigint.maxmax n0 n1 is the greatest of n0 and n1.val · tolk
- Tolk.Dtype.maxmax dt is the greatest value of dt: infinity for the floats that have infinities, the greatest finite value for the other floats, the greatest integer of dt's bitsize for the integers, weak included, and true for Bool and Void.val · tolk
- Tolk.Dtype.Value.maxmax v0 v1 is v1 if v1 > v0, and v0 otherwise.val · tolk
- Rig_nv_abi.Launch.max_bank_bytesmax_bank_bytes is the most bytes of a constant bank a launch reads, 64 KiB: make refuses a kernel's bank of more.val · rig
- Rig_nv_abi.Method.max_copymax_copy is the most bytes one copy moves, of one row of a box, and of one fill, 2{^31}.val · rig
- Tolk.Ops.max_exact_shiftmax_exact_shift is the most bits a left shift that nothing truncates computes exactly in exec_alu.val · tolk
- Rig_mlx5_abi.Entry.max_inlinemax_inline is 28, the most bytes a Write_inline carries in one basic block.val · rig
- Tolk.Setting.max_kernel_buffersmax_kernel_buffers is the number of buffers one kernel may access, from MAX_KERNEL_BUFFERS.val · tolk
- Rig_remote_proxy.Wire.max_keymax_key is the most bytes of a job's key: 4096.val · rig
- Rig_mlx5.max_messagemax_message nic is the most bytes one message of nic's port carries, as the kernel reported at open_: the most one write, read or send moves.val · rig
- Nx_array.Layout.max_numelmax_numel is 2{^53}.val · nx
- Tolk.Shape.max_numelmax_numel u is the product of max_shape u.val · tolk
- Nx_array.max_operandsmax_operands is 18, nx_array.h's NX_MAX_OPERANDS: the length of the runs in which the door claims and releases arrays.val · nx
- Kaun.Pool.max_pool2dmax_pool2d ~kernel_size:(kh, kw) x is the maximum of each kh × kw window over x's last two axes.val · kaun
- Nx_array.Layout.max_rankmax_rank is 32, the most axes a layout has.val · nx
- Tolk.Shape.max_shapemax_shape u is u's shape with each symbolic size replaced by its greatest value (Ops.to_max_shape).val · tolk
- Rig_nv_abi.Qmd.max_sizemax_size d is the largest size d takes: 2{^31}-1 blocks along a grid's X, 65535 along Y and Z; 1024 threads along a block's X and Y, 64 along Z.val · rig
- Rig_nv_abi.Launch.max_threadsmax_threads l is the most threads a block of l may have: 1024, or fewer for the registers its threads use.val · rig
- Nx_array.Dtype.max_valuemax_value dt is dt's greatest value: infinity for a float format with infinities, 448.val · nx
- Rig_nv_abi.Gpfifo.max_wordsmax_words is the most words a segment holds, 2{^21} - 1.val · rig
- Nx.maximummaximum a b is the greater element by the dtype's order (less).val · nx
- Nx_kernel.Prog.E.maximumval · nx
- Tolk.Ops.Elementwise.maximumval · tolk
- Tolk.Ops.maximumval · tolk
- Tolk.Ops.Upat.maximumval · tolk
- Nx.maximum_smaximum_s x c is maximum x c.val · nx
- Hugin_gg.Box2.maxxmaxx b is the largest x of b.val · hugin
- Hugin_gg.Box2.maxymaxy b is the largest y of b, its bottom edge on screen.val · hugin
- Jera.Split.mclachlanmclachlan is K(λh) D(h/2) K((1 − 2λ)h) D(h/2) K(λh) with λ = 0.1931833275037836, the two-stage scheme of least error (Omelyan, Mryglod and Folk, 2002): order 2.val · jera
- Norn.Diag.mcse_meanmcse_mean u d is the Monte Carlo standard error of each element's mean: its standard deviation over the square root of the effective sample size of the split chains' draws.val · norn
- Norn.Gaussian.meanmean u g is g's mean.val · norn
- Nx.meanmean ~axes x is the sum divided by the number of elements.val · nx
- Talon.Expr.meanmean a is the arithmetic mean of a's values, as float64: their sum divided by their count.val · talon
- Rig_amd.measureThe type for the counters a device measures, by their names in Rig_amd_abi.Counter.names.type · rig
- Ymir.Grid.measuremeasure g is each cell's measure, batch @ shape g: in steradians on the sky, by Van Oosterom and Strackee's formula on corner differences, within about 1e-9 of a 0.031″ TAN pixel's solid angle; in the plane's unit squared in a plane, by the shoelace formula.val · ymir
- Talon.Expr.medianmedian a is quantile 0.5 a.val · talon
- Sowilo.median_blurmedian_blur ~ksize img applies a median filter.val · sowilo
- Hugin_vg_instances.Instances.meetsmeets a b is the box of the points in both a and b, or None if they have none or only touch, to within rounding, each extending past the other along an axis: what a leaf whose box touches a clip's box paints is nothing, however rounding has placed the two.val · hugin
- Ymir_units.Unit.megamega u is 106 u.val · ymir
- Tolk.Dtype.memmem dt dts is true iff dt is in dts.val · tolk
- Tolk.Op.Set.memmem o s is true iff o is in s.val · tolk
- Tolk.Ops.Nodes.memmem u s is true iff u is in s.val · tolk
- Norn.Dist.membershipmembership d x is the posterior probability of each value of d's discrete latent variables given x: of shape [k], each component's, for a mixture of k components; of shape [n; k], each step's state's, for an hmm of n steps over k states, from the forward and backward variables (γ_t ∝ α_t β_t, Rabiner 1989, eq.val · norn
- MemoryThis page traces memory through its life, from a buffer's creation, through its use on several devices and the host, to its return.guide · rig
- Rig_amd.memoryThe type for memory a path gives a device.type · rig
- Rig_edge.memoryThe type for the kinds of a device's memory.type · rig
- Rig_mlx5.Region.memoryThe type for memory to register.type · rig
- Rig_nv.memoryThe type for memory a path gives a device.type · rig
- Rig_pci.MemoryThe memory a GPU addresses, placed by kind.module · rig
- Rig_pci.Page_table.memorymemory t is the number of bytes of the GPU's physical memory t manages: memory of create.val · rig
- Rig_program.memoryThe type for a program's memory.type · rig
- Tolk.MemoryMemory planning.module · tolk
- Tolk.Coalesce.memory_coalescingmemory_coalescing sink r merges the loads, and the stores, of sink that access consecutive elements of one buffer, under one gate, with one argument and at one depth, into vector loads and stores, if r supports them (supports_float4):val · tolk
- Rig.memory_devicememory_device name is open_ of the memory device named name on this machine.val · rig
- Tolk.Memory.memory_plan_rewritememory_plan_rewrite ~held_bufs linear is the schedule linear, an Op.t.Linear of calls run in order, with its buffers placed in arenas.val · tolk
- Fehu.Info.mergemerge a b is the union of a and b.val · fehu
- Hugin_kit.Scale.mergemerge s s' is Ok m if no property is set to different values in s and s', where m sets each property that s or s' sets to its value there, and Error p otherwise, for the first such property p in the order of property.val · hugin
- Kaun.Attention.mergemerge l y concatenates the heads of y, of shape [| batch; heads; seq; d |], and projects them through l, with sum as Linear.apply takes it: shape [| batch; seq; embed |].val · kaun
- Talon.Kit.merge_intervalsmerge_intervals ~by (s, e) q is q's sets in normal form (Normal form), so periods that overlap merge, and so do periods that touch, where one's stop is the next one's start.val · talon
- Nx.MeshNamed axes over a set's devices.module · nx
- Nx.Placement.meshmesh m cuts cuts each axis a of (a, names) over the mesh axes names, major first; a mesh axis no cut names holds copies.val · nx
- Nx.meshgridmeshgrid ~indexing [x1; …; xn] is the coordinate grids of the 1-d x1 to xn: n views of one shape, the ith holding xi's element j wherever its axis for xi is at j.val · nx
- Rig_mlx5_abi.Completion.messagemessage c describes c for a person, such as "queue pair 0x1c3, entry 7: retry counter exceeded (vendor syndrome 0x81)".val · rig
- Munin.Artifact.metadatametadata a is what the producing run attached to a, in the order it gave.val · munin
- Nx_io.Gguf.metadatametadata g is g's key-values, in file order.val · nx
- Quill.Doc.metadatametadata d is the document-level metadata of d.val · quill
- Talon_arrow.metadatametadata ?column f is f's schema-level custom metadata, or that of the field column, in file order, without the extension keys its type holds.val · talon
- Tolk.Cstyle.metalmetal target renders Metal Shading Language for an Apple GPU.val · tolk
- Tolk.Tc.metalmetal is Apple's 8 x 8 x 8 tensor cores.val · tolk
- Brot.Decoder.metaspacemetaspace ~replacement ~prepend_scheme () converts metaspace markers back to spaces.val · brot
- Brot.Pre_tokenizer.metaspacemetaspace () replaces spaces with a visible marker.val · brot
- Rig_nv_abi.MethodChannel methods.module · rig
- Ymir_units.Unit.metremetre is the symbol m.val · ymir
- Kaun.MetricEvaluation metrics: accuracy, precision/recall/F1, AUC-ROC.module · kaun
- Munin.MetricSamples, series and their statistics.module · munin
- Munin.Session.metricThe type for handles.type · munin
- Munin.Session.metricmetric s key declares the metric key in s's run and is the handle that logs its samples.val · munin
- Munin.History.metricsmetrics h is the keys of the run's metrics that have a sample that is not an orphan, sorted: the metrics whose stats is Some.val · munin
- Ymir_units.Unit.micromicro u is 10-6 u.val · ymir
- Hugin_kit.Time.microsecondsmicroseconds k is nanoseconds (1_000 * k).val · hugin
- Hugin_gg.Box2.midmid b is the centre of b.val · hugin
- Ymir_units.Unit.millimilli u is 10-3 u.val · ymir
- Hugin_kit.Time.millisecondsmilliseconds k is nanoseconds (1_000_000 * k).val · hugin
- Jera.Sde.milsteinmilstein is the derivative-free Milstein method (Kloeden and Platen, 1992, §11.1), Itô: strong order 1 for diagonal noise, and 1/2 otherwise.val · jera
- Nx.minmin ~axes x is the least element, as max; all on booleans.val · nx
- Talon.Expr.minmin a is the least of a's values in talon's total order, of a's type.val · talon
- Tolk.Bigint.minmin n0 n1 is the least of n0 and n1.val · tolk
- Tolk.Dtype.minmin dt is the least value of dt: -infinity for the floats that have infinities, the negated max for the other floats, the least integer of dt's bitsize for the integers, weak included, and false for Bool and Void.val · tolk
- Tolk.Dtype.Value.minmin v0 v1 is v1 if v1 < v0, and v0 otherwise.val · tolk
- Rig_remote_proxy.Wire.min_keymin_key is the fewest bytes of a job's key: 16.val · rig
- Nx_array.Dtype.min_valuemin_value dt is dt's least value: neg_infinity for a float format with infinities, -448.val · nx
- Jera.MinimizeMinima.module · jera
- Nx.minimumminimum a b is the lesser element, as maximum.val · nx
- Nx_kernel.Prog.E.minimumval · nx
- Tolk.Ops.Elementwise.minimumminimum x y is neg (maximum (neg x) (neg y)) for floats; for integers, the same through an order-reversing xor.val · tolk
- Tolk.Ops.minimumminimum x y is neg (maximum (neg x) (neg y)) for floats; for integers, the same through an order-reversing xor.val · tolk
- Tolk.Ops.Upat.minimumminimum x y is neg (maximum (neg x) (neg y)) for floats; for integers, the same through an order-reversing xor.val · tolk
- Nx.minimum_sminimum_s x c is minimum x c.val · nx
- Hugin_kit.Locale.minusminus l is the minus sign of l.val · hugin
- Tolk.Bigint.minus_oneminus_one is -1.val · tolk
- Ymir_units.Unit.minuteminute is 60 s.val · ymir
- Hugin_kit.Time.minutesminutes k is seconds (60 * k).val · hugin
- Talon.Time.Span.minutesminutes n is like us in minutes.val · talon
- Hugin_gg.Box2.minxminx b is the smallest x of b.val · hugin
- Hugin_gg.Box2.minyminy b is the smallest y of b, its top edge on screen.val · hugin
- Tolk.Postrange.missesmisses r e is true iff the effect e does not depend on the range r: r is none of the ranges (Ops.ranges) of e's index, its validity included, if e is a store, or of e's sources, but a call's body, if e is a call (Ops.call).val · tolk
- Hugin_kit.Scale.missingmissing s x is true where an element of x is missing for s, and has the shape and placement of x.val · hugin
- Hugin_text.Text.Layout.missingmissing l is the characters of l's text that no face has, each once, in order of first occurrence (Faces).val · hugin
- Hugin_gg.Stroke.miter_limitmiter_limit s is the miter limit of s.val · hugin
- Hugin_gg.Color.mixmix t c c' is the colour a fraction t of the way from c to c' in Oklab, interpolated with premultiplied alpha as CSS Color 4 does: the opacity is (1 - t) αc + t αc', and each Oklab coordinate is the average of those of c and c' weighted by (1 - t) αc and t αc', so that a transparent end adds opacity and no colour.val · hugin
- Norn.Dist.mixturemixture ~logits d draws one component k with probability softmax logits for the whole value, then the value from d_k, the slice of d at k on its leading axis: the log density is logsumexp_k (log_softmax logits_k + log_density d_k x).val · norn
- Hugin.Size.mmmm l is l millimetres in points, l *.val · hugin
- Kaun_datasets.mnistmnist () is (train_x, train_y, test_x, test_y): 60,000 training and 10,000 test examples.val · kaun
- Nx.mod_mod_ a b is the remainder of a / b, of a's sign: fmod on floats, and a for a divisor of 0 on integers, so that a = add (mul b (div a b)) (mod_ a b) on integers.val · nx
- Tolk.Ops.Elementwise.mod_mod_ x y is the remainder of div ~rounding:`Floor x y.val · tolk
- Tolk.Ops.mod_mod_ x y is the remainder of div ~rounding:`Floor x y.val · tolk
- Tolk.Ops.Upat.mod_mod_ x y is the remainder of div ~rounding:`Floor x y.val · tolk
- Nx.mod_smod_s x c is mod_ x c.val · nx
- Kaun.Init.modeThe type for variance-scaling fan modes.type · kaun
- Ymir_units.Unit.molemole is the symbol mol.val · ymir
- Nx.moments_rangesmoments_ranges ~lo ~hi x is (mean, var), one row of each per range.val · nx
- Nx_kernel.monoidtype · nx
- Hugin_kit.Curve.monotone_xmonotone_x is Steffen's interpolant of y as a function of x.val · hugin
- Hugin_kit.Curve.monotone_ymonotone_y is monotone_x with the roles of x and y exchanged: the interpolant of x as a function of y.val · hugin
- Hugin_kit.Locale.monthmonth l m is the short name of month m in l, 1 for January.val · hugin
- Hugin_kit.Time.monthsmonths k is the interval of k calendar months.val · hugin
- Tolk.Shape.mopmop u m applies the movement m to u as one node, its shape arguments stored as simplified sources.val · tolk
- Tolk.Shape.mop_cleanupmop_cleanup merges and removes movements and indexing:val · tolk
- Sowilo.morphological_gradientmorphological_gradient ~kernel img is dilate ~kernel img - erode ~kernel img, as Nx.sub computes it, and on a boolean image true where the dilation is true and the erosion false.val · sowilo
- Fehu_envs.Mountain_carMountain car with sparse reward (MountainCar-v0).module · fehu
- Norn.Smc.moveThe type for the kernels that move the particles.type · norn
- Nx_array.Layout.movemove m l is the layout of m's result over l's positions: its index i reaches the position l reaches at m's map of i.val · nx
- Nx_array.movemove m a is the array over a's buffer laid out by Layout.move m (layout a), or None where that is None.val · nx
- Nx_array.MoveMovements.module · nx
- Quill.Doc.movemove id ~pos d moves the cell id to position pos.val · quill
- Quill.Session.move_cellmove_cell id ~pos s moves cell id to position pos.val · quill
- Hugin_gg.Path.move_tomove_to pt p is p with a new subpath started at pt.val · hugin
- Nx.moveaxismoveaxis a b x is x with its axis a moved to position b, the other axes in their order.val · nx
- Tolk.Op.Set.movementmovement is the operations that move elements without computing: Reshape, Expand, Permute, Pad and Shrink.val · tolk
- Tolk.Ops.movementThe type for the arguments of movements.type · tolk
- Talon.Time.Span.msms n is like us in milliseconds.val · talon
- Kaun.Loss.msemse predictions targets is the reduction over all elements of the squared error (predictions - targets)².val · kaun
- Tolk.Mtlcompiler_workermodule · tolk
- Nx.mulmul a b is the elementwise product, as add.val · nx
- Nx_kernel.Prog.E.mulval · nx
- Nx_wide.mulmul a b is a * b within 4u² relative, exactly when both have a zero lo.val · nx
- Tolk.Bigint.mulmul n0 n1 is n0 * n1.val · tolk
- Tolk.Ops.Elementwise.mulval · tolk
- Tolk.Ops.mulval · tolk
- Tolk.Ops.Upat.mulval · tolk
- Ymir_units.Quantity.mulmul a b is the product of the payloads in Unit.(unit a * unit b).val · ymir
- Nx.mul_smul_s x c is mul x c.val · nx
- Fehu.Space.Multi_binarymodule · fehu
- Fehu.Space.Multi_discretemodule · fehu
- Nx.multi_dotmulti_dot [x1; …; xn] is the matrix product x1 x2 ⋯ xn of n >= 2 operands.val · nx
- Tolk.C.multiarchmultiarch is the name Linux distributions give the host's machine in library directories, as in "x86_64-linux-gnu".val · tolk
- Tolk.Ops.multirange_strmultirange_str ~color ~pad rs is the range_str of each of rs, sorted by argument, separated by commas and padded with spaces to pad printed columns.val · tolk
- muninmunin records what a program does as it runs: its params, its metrics step by step, readings of its machine and devices, the files it logs, and versioned artifacts such as checkpoints.library · munin
- MuninRecords of runs and artifacts, in plain files.module · munin
- Munin_float_textmodule · munin
- Munin_sysReadings of the machine, of this process and of devices.module · munin
- Munin_tuiA terminal dashboard of a munin store.module · munin
- Ymir_units.Codata.muon_massmuon_mass r is mμ in kg.val · ymir
- Tolk.Setting.mvmv lets hand-coded optimizations lay out matrix-vector products (Heuristic.hand_coded_optimizations), from MV.val · tolk
- Norn.Dist.mvnmvn ~loc ~scale_tril is the multivariate normal of vectors along the last axis, with mean loc and covariance L Lᵀ, L the lower triangle of scale_tril.val · norn
- Nx_quant.mxfp4mxfp4 ~scales codes is the MXFP4 weight of codes, two per byte, and scales: Mxfp4 { codes; scales }.val · nx
- Nx_quant.mxfp4_blocksmxfp4_blocks blocks is the MXFP4 weight stored as GGUF's blocks.val · nx
n
- Fehu.Space.Discrete.nn s is the number of choices in s.val · fehu
- Talon.Expr.n_uniquen_unique a is the number of distinct keys of a, null being one key, as int64.val · talon
- Hugin.namename s f is f with the id segment Field s in place of its index (Ids), so that the ids in f stay the same when the siblings before it change.val · hugin
- Hugin_kit.Scale.namename s is the name of s, if set.val · hugin
- Munin.Artifact.namename a is a's artifact name.val · munin
- Munin.Run.namename r is r's name, if it was given one.val · munin
- Nx_array.Dtype.namename dt is dt's constructor name in lower case, as "float32" or "float4_e2m1fn".val · nx
- Nx_cpu.namename names the kernels in messages, as "nx.cpu".val · nx
- Nx_kernel.S.namename names the kernels in messages, as "nx.cpu".val · nx
- Rig.Image.namename fn is fn's name in its image.val · rig
- Rig.namename d is d's name for people: its opener's name, with "@MACHINE" appended for a device of another machine.val · rig
- Rig_bnxt.namename nic is nic's name, such as "BNXT:1".val · rig
- Rig_mlx5.namename nic is the kernel's name of nic, such as "mlx5_0".val · rig
- Rig_pci.Driver.namename d i is the name of function i.val · rig
- Rig_pci.Machine.namename m is the name of m as its transport gives it, such as "host:7000", or None if m is reached without a transport.val · rig
- Rig_remote_proxy.Link.namename l is l's name.val · rig
- Talon.Tz.namename z is the name of z: the name its database holds it under (find), such as "Europe/Paris", "UTC" for utc, and "+05:30" for fixed.val · talon
- Tolk.Dtype.namename dt is the short name of dt: "i8" to "i64" for the signed integers, "u8" to "u64" for the unsigned ones, "f16", "bf16", "f32" and "f64", "fp8e4m3", "fp8e5m2", "fp8e4m3fnuz" and "fp8e5m2fnuz" for the 8-bit floats, and "bool", "void", "weakint" and "weakfloat".val · tolk
- Tolk.Op.namename o is the name of o in upper case, words separated by underscores: "ADD", "CONTIGUOUS_BACKWARD".val · tolk
- Ymir.Frame.namename f is f's canonical name: "icrs", "fk5_j2000", "galactic", "ecliptic_j2000", "supergalactic", "gcrs", "cirs", "true_of_date", "tirs".val · ymir
- Ymir.Time.namename s is s's canonical name: "utc", "tai", "tt", "tdb".val · ymir
- Ymir_fits.Fits.namename hdu is the path given to read, of_bytes's name, or "" for a constructed HDU.val · ymir
- Ymir_units.Constant.namename k is k's name.val · ymir
- Tolk.Ops.Upat.namednamed name p is p naming the node it matches name.val · tolk
- Nx_io.Archive.namesnames a is the names of a's entries, sorted.val · nx
- Rig_amd_abi.Counter.namesnames g is the counters g counts, in increasing order.val · rig
- Rig_mlx5.namesnames () is the names of this machine's RDMA devices whose PCI function the mlx5_core driver holds, sorted as strings, so "mlx5_10" comes before "mlx5_2".val · rig
- Talon.Record.namesnames r is the names of r's fields, in order.val · talon
- Talon.Sel.namesnames ns selects the columns ns, in the order of ns.val · talon
- Tolk.Dtype.nannan is the positive quiet NaN whose payload is zero: the NaN a constant takes when an operation makes one from operands that are not NaNs.val · tolk
- Ymir_units.Unit.nanonano u is 10-9 u.val · ymir
- Hugin_kit.Time.nanosecondsnanoseconds k is the interval of k nanoseconds.val · hugin
- Tolk.Decomp_dtype.narrownarrow x dt is x cast to dt, but for a Dtype.t.Float32 from a Dtype.t.Float64 or an integer of 32 or 64 bits whose bounds (Ops.vmin, Ops.vmax) do not lie within 2{^24}, more precise than a float32: then it is x rounded to odd, towards zero with the last bit set if bits were dropped, so that a narrow float rounded to nearest from it is x rounded to nearest once.val · tolk
- Hugin_kit.Curve.naturalnatural is the natural cubic spline through the points, taken parametrically: x and y are each the cubic spline of the point's index that passes through every point, has continuous first and second derivatives, and has a zero second derivative at both ends.val · hugin
- Nx.nbytesnbytes x is the bytes x's elements fill when stored contiguously, (numel x * b + 7) / 8 for a dtype of b bits.val · nx
- Tolk.Shape.nbytesnbytes u is the number of bytes of u's elements.val · tolk
- Nx.ndimndim x is x's number of axes.val · nx
- Tolk.Shape.ndimndim u is the length of u's shape.val · tolk
- Nx.ndtrndtr x is the standard normal distribution function Φ(x) = erfc (-x/√2) / 2.val · nx
- Nx.ndtrindtri p is the x with ndtr x = p, on [0., 1.]: -inf at 0.val · nx
- Tolk.Ops.Elementwise.nene x y is logical_not (eq x y).val · tolk
- Tolk.Ops.nene x y is logical_not (eq x y).val · tolk
- Tolk.Ops.Upat.nene x y is logical_not (eq x y).val · tolk
- Norn_model.Init.nearnear p is the coordinates of the values p, each moved uniformly in (-0.1, 0.1).val · norn
- Tolk.Ops.needs_storageneeds_storage u is true iff realizing u allocates.val · tolk
- Nx.negneg x is -x.val · nx
- Nx_kernel.Prog.E.negval · nx
- Tolk.Bigint.negneg n is -n.val · tolk
- Tolk.Ops.Elementwise.negneg x is logical_not x for booleans and x * -1 otherwise.val · tolk
- Tolk.Ops.negneg x is logical_not x for booleans and x * -1 otherwise.val · tolk
- Tolk.Ops.Upat.negneg x is logical_not x for booleans and x * -1 otherwise.val · tolk
- Norn.Dist.neg_binomialneg_binomial ~mean ~dispersion is the count of mean mean and variance mean + mean² / dispersion, a Poisson whose rate is gamma with concentration dispersion.val · norn
- Ymir.Healpix.neighboursneighbours h p is each pixel's eight neighbours, p's shape @ [8]: south-west, west, north-west, north, north-east, east, south-east and south, as the HEALPix library orders them.val · ymir
- Jera.Minimize.nelder_meadnelder_mead is the Nelder–Mead simplex, for an objective with no useful gradient: non-smooth or piecewise constant.val · jera
- Nx.Ptree.nestnest (module U) s is U whose parameter's positions are values of s, walked by s with their paths after the position's.val · nx
- Norn.NestedNested sampling, by slice moves on the prior.module · norn
- Norn.Diag.nested_rhatnested_rhat u ~superchains d is the rank-normalised nested R-hat (Margossian, Hoffman, Sountsov, Riou-Durand, Vehtari and Gelman 2024) of each element, for many short chains: the chains form superchains consecutive groups of equal size, each started from one point, and the statistic compares the variance between groups with the variance within them, chains split and scored as for rhat.val · norn
- Ymir_units.Codata.neutron_massneutron_mass r is mn in kg.val · ymir
- Tolk.Ops.new_buffernew_buffer ~slot ~phase d size dt is a Op.t.Buffer of size elements of dt on d, in slot (default unique_num), whose first element lies phase bytes past a 16-byte boundary (default 0; see param_arg).val · tolk
- Rig.Local.new_keynew_key init is a key whose value on a device d is init d.val · rig
- Jera.Minimize.newtonnewton ~linear is Newton's method, for a smooth, ill-conditioned objective.val · jera
- Jera.Root.newtonnewton ~tol ~budget ~slope f x0 is a zero of f near x0 in each element, by Newton's method: slope x is f's derivative, elementwise.val · jera
- Jera.System.newtonnewton ~derivative is Newton's method: derivative x dx is f's Jacobian at x applied to dx, such as snd (Rune.jvp' f x dx).val · jera
- Ymir_units.Unit.newtonnewton is kg m s-2.val · ymir
- Ymir_units.Codata.newtonian_gravitationnewtonian_gravitation r is G, the Newtonian constant of gravitation, in m3 kg-1 s-2.val · ymir
- Rig_remote_proxy.Link.nextnext l is the next command l's peer sent, waiting for it.val · rig
- Nx.Rng.next_keynext_key () is the scope's next key, a host key computed at the call.val · nx
- Quill_project.next_notebooknext_notebook t nb is the notebook after nb, or None.val · quill
- Brot.Normalizer.nfcnfc is Unicode NFC normalization (canonical composition).val · brot
- Brot.Normalizer.nfdnfd is Unicode NFD normalization (canonical decomposition).val · brot
- Brot.Normalizer.nfkcnfkc is Unicode NFKC normalization (compatibility composition).val · brot
- Brot.Normalizer.nfkdnfkd is Unicode NFKD normalization (compatibility decomposition).val · brot
- Brot.Normalizer.nmtnmt is the control character cleanup of neural machine translation models.val · brot
- Tolk.Setting.no_colorno_color makes Helpers.colored leave text unchanged, from NO_COLOR.val · tolk
- Tolk.Setting.no_memory_plannerno_memory_planner keeps scheduling from reusing the memory of buffers that are no longer needed, from NO_MEMORY_PLANNER.val · tolk
- Nx_kernel.Prog.nodeThe type for a program's nodes.type · nx
- Jera.Quad.Rule.nodesnodes r dtype is (x, w), r's nodes in increasing order and their weights, 1-D tensors of dtype: Σ w f(x) approximates ∫₋₁¹ f.val · jera
- Tolk.Ops.NodesSets of nodes that remember the order in which nodes joined them.module · tolk
- Jera.Quad.Rule.nodes_atnodes_at p r dtype is nodes r dtype at p.val · jera
- Nx_top.nolocval · nx
- Norn_model.noncentrenoncentre sel m is reparam sel Norn.Dist.standardize m.val · norn
- Nx.nonzerononzero ~size t is the coordinates of t's elements other than zero, one 1-d value per axis: the jth is column j of argwhere ~size t, a view.val · nx
- Tolk.Setting.nooptnoopt disables kernel optimizations, from NOOPT.val · tolk
- Nx.Ptree.normnorm s dt x is x's Euclidean norm, sqrt (dot s dt x x) in exact arithmetic, computed so that no square overflows: each coordinate is divided by their greatest magnitude m, and m multiplies the root of the quotients' sum of squares.val · nx
- Kaun.Init.normalnormal ~stddev samples from N(0, stddev²).val · kaun
- Norn.Dist.normalnormal ~loc ~scale has density exp (-z²/2) / (scale sqrt (2π)), z = (x - loc) / scale.val · norn
- Nx.Rng.normalnormal dt shape is standard normal draws, by the Box-Muller transform: draws 2q and 2q + 1 in C order are the cosine's and the sine's of one transform, of the uniform draws 2q and 2q + 1 at the compute dtype.val · nx
- Nx.Rng.normal_atnormal_at p dt shape is normal dt shape at p.val · nx
- Fehu.Gae.normalizenormalize arr is a copy of arr with zero mean and unit variance.val · fehu
- Hugin_kit.Scale.normalizenormalize s x is the normalised value of x on s (Normalisation), nan iff x or an end of the domain of s is missing for s; only the default domain of a custom scale can have a missing end.val · hugin
- Sowilo.normalizenormalize ~mean ~std img is per-channel normalization: (img - mean) / std.val · sowilo
- Hugin_kit.Scale.normalize_indexnormalize_index s k is normalize s c for the category c at the index k of the domain of s, and nan if k is not an index of the domain.val · hugin
- Hugin.Mark.normalizednormalized rows r is Some us, the value of each row for the role r normalised by its scale, before the role's range and map_range, or None if r is unbound, bound to a constant or reads no scale.val · hugin
- Brot.normalizernormalizer t is t's normalizer, if any.val · brot
- Brot.NormalizerText normalization.module · brot
- nornProbabilistic inference for OCaml, built on Nx, Rune and Jera.library · norn
- NornProbabilistic inference.module · norn
- Norn_float_textmodule · norn
- Norn_modelModels as generative functions.module · norn
- Talon.Expr.notnot a is the negation of a, null where a is.val · talon
- Tolk.Shape.Not_a_numberThe exception number raises on `Invalid.exception · tolk
- Nx.not_equalnot_equal a b is a <> b.val · nx
- Nx_kernel.Prog.E.not_equalval · nx
- Nx.not_equal_snot_equal_s x c is not_equal x c.val · nx
- Hugin_kit.Number.notationThe type for notations.type · hugin
- Hugin_kit.Scale.notationnotation s is the notation of the tick labels of s, if set.val · hugin
- Quill_project.notebookA notebook reference.type · quill
- Notebook FormatA Quill notebook is a CommonMark markdown file.guide · quill
- Quill_book.Render.notebook_output_pathnotebook_output_path nb is the output HTML path for nb relative to the project root (e.g.val · quill
- Quill_project.notebooksnotebooks t is the flat, ordered list of all notebooks in t, excluding placeholders.val · quill
- Munin.Run.notesnotes r is r's notes as read, if it had any.val · munin
- Rig.Profile.nownow () is the host clock, in nanoseconds of the system's monotonic clock from an unspecified start.val · rig
- Rig_edge_primitives.nownow () is the host's clock (rig_now_ns): nanoseconds of the system's monotonic clock from an unspecified start, so only differences and comparisons of its values mean anything.val · rig
- Talon.Time.Span.nsns n is n nanoseconds.val · talon
- Ymir.Healpix.nsidenside h is h's number of divisions of a base pixel's side.val · ymir
- Quill.Doc.nthnth i d is the ith cell (zero-indexed), or None.val · quill
- Tolk.Ops.nthnth u i is the ith source of u, from 0.val · tolk
- Ymir_units.Codata.nuclear_magnetonnuclear_magneton r is μN in J T-1.val · ymir
- Talon.Expr.nullnull is null, typed by what it meets.val · talon
- Talon.Column.null_countnull_count c is the number of null rows of c.val · talon
- Talon.Kit.null_countnull_count q is one row with, for each column of q, in order and under its name, its number of nulls as int64:val · talon
- Talon.Order.nulls_firstnulls_first k is k with its nulls before every value.val · talon
- Hugin.numnum ~scale ~valid ~title x is the quantities of x, one datum per element, where:val · hugin
- Fehu.Vec_env.num_envsnum_envs t is the number of environments.val · fehu
- Hugin.View.numbernumber name ~init is the key name of a number, initially init.val · hugin
- Hugin_kit.NumberNumber formats.module · hugin
- Quill_project.numbernumber toc nb is the section number of nb in toc, derived from its position.val · quill
- Rig_mlx5.Qp.numbernumber qp is qp's number, which completions name.val · rig
- Tolk.Shape.numbernumber c is the value of c.val · tolk
- Munin_float_text.Float_read.number_endnumber_end s i stop is the index after the number in JSON's syntax that starts at i in s, read no further than stop, or -1 if none starts there: an optional minus, then a zero alone or digits not starting with one, an optional point and digits, and an optional e or E, sign and digits.val · munin
- Quill_project.number_stringnumber_string [1; 2] is "1.2".val · quill
- Tolk.Schedule.numberednumbered g numbers the ranges and the storage g reaches, that of its calls' bodies too, by their order in g.val · tolk
- Tolk.Schedule.numberingThe type for a graph numbered by its own order: a process names each range and each piece of storage from counters that depend on what it did before, so a key or a kept graph names them by their order instead.type · tolk
- Tolk.Bigint.numbitsnumbits n is the number of bits of abs n: 0 for zero, and otherwise the k for which 2{^k-1} <= abs n < 2{^k}.val · tolk
- Nx.numelnumel x is x's number of elements: the product of its extents, 1 for a 0-d value.val · nx
- Nx_array.Layout.numelnumel l is l's number of indices, the product of its extents.val · nx
- Tolk.Shape.numelnumel u is the product of u's shape.val · tolk
- NumPy ComparisonThis document compares the Nx library (OCaml) with NumPy (Python), highlighting similarities, differences, and providing equivalent code examples.guide · nx
- Norn.NutsThe No-U-Turn sampler over many chains.module · norn
- Rig_bin.Images.nvnv is rig.nv.pci's list.val · rig
- Tolk.Compiler_cuda.nvrtcnvrtc ~ptx ~cache_key arch is the compiler of CUDA C++ source for the GPU architecture arch, as in "sm_89": to PTX if ptx is true (the default), which the driver translates for the GPU it loads it on, and else to a cubin, the binary of arch itself.val · tolk
- Tolk.Nvrtc_workermodule · tolk
- nxNx provides n-dimensional arrays for OCaml.library · nx
- NxN-dimensional arrays as values on device sets.module · nx
- Talon.Expr.nxnx { f } a applies f to a's values with nx's semantics, IEEE's for floats: nx { f = Nx.exp } x.val · talon
- Talon.Expr.nx2nx2 { f2 } a b is like nx for two arguments, which meet at their common type: nx2 { f2 = Nx.atan2 } y x.val · talon
- Nx_arrayDtypes, layouts and arrays over device buffers.module · nx
- Nx_cpuKernels for host memory.module · nx
- Nx_ioArrays to and from files.module · nx
- Nx_kernelKernels: what they compute and the contract they implement.module · nx
- Nx_quantQuantised weights.module · nx
- Nx_raggedRagged arrays.module · nx
- Nx_topThe toplevel printer for arrays.module · nx
- Nx_wideDouble-word numbers.module · nx
o
- Tolk.Ops.Elementwise.OOperators, for local opening: O.(x + int 4 * y).module type · tolk
- Tolk.Ops.OOperators, for local opening: O.(x + int 4 * y).module · tolk
- Tolk.Ops.Upat.OOperators, for local opening: O.(x + int 4 * y).module · tolk
- Fehu_envs.Cartpole.obstype · fehu
- Fehu_envs.Grid_world.obstype · fehu
- Fehu_envs.Mountain_car.obstype · fehu
- Fehu_envs.Random_walk.obstype · fehu
- Ymir.ObservationMeasured samples on a grid.module · ymir
- Fehu.Env.observation_spaceobservation_space env is the space of valid observations.val · fehu
- Fehu.Vec_env.observation_spaceobservation_space t is the shared observation space.val · fehu
- Jera.OdeOrdinary differential equations.module · jera
- Nx.Repr.of_arrayof_array s a is the value whose bytes are a's.val · nx
- Nx_array.of_arrayof_array dt s xs is a fresh C-contiguous array of shape s on Rig.host holding xs in C order of indices, each stored as Dtype.of_float says.val · nx
- Talon.of_batchesof_batches ts is the rows of ts one after the other, without a copy, in O(number of batches).val · talon
- Nx.of_bigarrayof_bigarray b is a host value holding a copy of b's elements, of b's shape and the dtype that stores its kind.val · nx
- Nx_array.of_bigarrayof_bigarray dt b is the C-contiguous array of dt elements over b's bytes, of b's shape.val · nx
- Rig.Buffer.of_bigarrayof_bigarray ba is a buffer on host over ba's bytes, without a copy.val · rig
- Nx_array.Dtype.of_bigarray_kindof_bigarray_kind k is the dtype that stores k's elements, if one does: None for Char, Int and Nativeint.val · nx
- Ymir_fits.Fits.of_bytesof_bytes ~name b is the HDUs of the bytes b in C order, as read reads a file.val · ymir
- Quill.Doc.of_cellsof_cells ?metadata cs is a document containing cs in order with the given metadata (defaults to []).val · quill
- Talon.Time.Date.of_civilof_civil (y, m, d) is day d of month m of year y in the proleptic Gregorian calendar, or None if m is not in [1;12], d is not a day of that month, or the date is out of range.val · talon
- Nx_array.Dtype.of_codeof_code c is the dtype whose code is c, if any: None unless 0 <= c < List.length all.val · nx
- Tolk.Dtype.of_constof_const c is the data type of the literal c: Bool for `Bool and `Invalid, Weak_int for `Int, and Weak_float for `Float.val · tolk
- Tolk.Dtype.of_constsof_consts cs is the committed data type of the literals cs.val · tolk
- Hugin_kit.Time.of_dateof_date ~tz_offset_s d is of_date_time ~tz_offset_s (d, (0, 0, 0)), the start of day d at tz_offset_s.val · hugin
- Ymir.Time.of_dateof_date s (y, m, d) (h, mi, sec) is the instant at h:mi:sec of the date (y, m, d) of the proleptic Gregorian calendar on s, one instant on the host.val · ymir
- Hugin_kit.Time.of_date_timeof_date_time ~tz_offset_s (d, tm) is the instant whose civil date and time at tz_offset_s are d and tm, on a whole second.val · hugin
- Talon.Time.Date.of_daysof_days n is the date n days after 1970-01-01, or None if it is out of range.val · talon
- Talon.of_denseof_dense ~validity ~axes value x is the table of one row per element of x, in row-major order: an int64 key column per name of axes, then the elements in column value, null where validity is clear.val · talon
- Talon.Tz.of_dirof_dir dir is the database of the TZif files under dir, as a system's zoneinfo directory holds them.val · talon
- Nx_kernel.Prog.of_exprsof_exprs ~ins outs is the program over operands of dtypes ins whose results are the expressions outs.val · nx
- Hugin_font.Font.of_fileof_file path is of_string s for the contents s of the file path, or Error (Io msg) if the file cannot be read.val · hugin
- Rig_disk.of_fileof_file path is the bytes of the regular file path on device, for reading only (Rig.Buffer.access), as long as the file was when opened.val · rig
- Nx_array.Dtype.of_floatof_float dt x is the value a store of x into dt holds (Stores).val · nx
- Tolk.Bigint.of_floatof_float x is x rounded toward zero.val · tolk
- Kaun.Rope.of_frequenciesof_frequencies ~context f is the schedule whose pair i has inverse frequency f.(i), for a head of 2 * Array.length f features and positions below context.val · kaun
- Ymir_fits.Fits.Image.of_hduof_hdu hdu describes a primary array or an IMAGE extension.val · ymir
- Ymir_fits.Fits.Table.of_hduof_hdu hdu describes a BINTABLE or TABLE extension, a tile-compressed image included.val · ymir
- Hugin_gg.Color.of_hexof_hex s is Ok c if s is a CSS hex colour, # then 3, 4, 6 or 8 hexadecimal digits of either case (#rgb, #rgba, #rrggbb or #rrggbbaa), and Error msg otherwise, msg saying why.val · hugin
- Nx_ragged.of_idsof_ids ~segments ids x groups the rows of x by ids: row s of the result holds the rows i of x with ids.{i} = s, in their order in x, for each s in [0, segments).val · nx
- Fehu.Space.Discrete.of_intof_int v is a discrete element with value v.val · fehu
- Tolk.Bigint.of_intof_int n is n.val · tolk
- Tolk.Dtype.Value.of_intof_int n is `Int n.val · tolk
- Tolk.Bigint.of_int32of_int32 n is n.val · tolk
- Tolk.Bigint.of_int32_unsignedof_int32_unsigned n is n read as unsigned, in [0;2{^32}-1].val · tolk
- Tolk.Bigint.of_int64of_int64 n is n.val · tolk
- Tolk.Bigint.of_int64_unsignedof_int64_unsigned n is n read as unsigned, in [0;2{^64}-1].val · tolk
- Rig.Buffer.of_ioof_io d k r ~access n is a buffer over the n bytes of r, a region of the io device d.val · rig
- Ymir.Time.of_jdof_jd s j1 j2 is the instants of Julian date j1 + j2 on s, their shapes broadcast.val · ymir
- Brot.Decoder.of_jsonof_json json is a decoder from HuggingFace JSON format.val · brot
- Brot.Normalizer.of_jsonof_json json is a normalizer deserialized from HuggingFace JSON.val · brot
- Brot.of_jsonof_json json is a tokenizer deserialized from HuggingFace JSON format.val · brot
- Brot.Post_processor.of_jsonof_json json is a post-processor from HuggingFace tokenizer.json format.val · brot
- Brot.Pre_tokenizer.of_jsonof_json json is a pre-tokenizer from HuggingFace JSON format.val · brot
- Talon.Sel.of_kindof_kind k selects the columns that a handle of kind k binds (see Kind.provably_equal), in schema order.val · talon
- Talon.Column.of_layoutof_layout ty l is the column of type ty laid out as l, without a copy, or Error (row, reason) for the first row whose value ty does not hold: text that is not UTF-8, a code outside a categorical's dictionary, a time of day outside the day.val · talon
- Nx_ragged.of_lengthsof_lengths lengths values is the ragged array whose rows have lengths, one after the other from the first row of values: its offsets are 0 followed by the running sum of lengths.val · nx
- Fehu.Info.of_listof_list kvs is a dictionary from the given key-value pairs.val · fehu
- Nx_io.Archive.of_listof_list entries is the archive of entries.val · nx
- Tolk.Op.Set.of_listof_list ops is the set of the operations in ops.val · tolk
- Talon.Time.of_msof_ms n is like of_us in milliseconds.val · talon
- Talon.Time.Span.of_msof_ms n is like of_us in milliseconds.val · talon
- Nx_array.Dtype.of_nameof_name s is the dtype whose name is s, if any.val · nx
- Nx_kernel.Prog.of_nodeof_node ~ins n is the program of the one node n over operands of dtypes ins: v ~ins nodes ~outs:[| k |], where nodes is In 0 to In (k - 1) then n and k the length of ins.val · nx
- Talon.Time.of_nsof_ns n is the stamp n nanoseconds after 1970-01-01 00:00:00 of its clock.val · talon
- Talon.Time.Span.of_nsof_ns n is n nanoseconds.val · talon
- Hugin_gg.Color.of_oklabof_oklab ~alpha l a b is the colour at Oklab coordinates (l, a, b), gamut mapped.val · hugin
- Hugin_gg.Color.of_oklchof_oklch ~alpha l c h is of_oklab ~alpha l (c *.val · hugin
- Talon.Expr.of_optionof_option a is v where a is Some v and null where it is None.val · talon
- Talon.Column.of_optionsof_options ty vs is like v, with a null for each None.val · talon
- Norn.Gaussian.of_precisionof_precision u dtype ?low_rank ~mean p is the Gaussian of mean mean whose precision is the diagonal p plus, with ~low_rank:(w, l), the sum of l_j w_j w_jᵀ over the directions w_j, the leading axis of w's tensors.val · norn
- Hugin_gg.Box2.of_ptsof_pts p q is the smallest box containing p and q, which are any two opposite corners.val · hugin
- Talon.Column.of_raggedof_ragged ?validity r is the list column whose rows are r's rows, null where validity is false: a list of the element type of_tensor gives r's values, such as list[int32] for Nx.int32 values, and of a tensor type for values of more than one axis.val · talon
- Talon.Time.of_sof_s n is like of_us in seconds.val · talon
- Talon.Time.Span.of_sof_s n is like of_us in seconds.val · talon
- Nx.Repr.of_shardsof_shards p arrays is the value whose bytes are arrays, one per device of p in order, each that device's window of one shape: along an axis p cuts, the sum of its windows' extents, and along any other, each array's.val · nx
- Talon.Query.of_sourceof_source s is the query of the rows of the source s, which it reads when it runs.val · talon
- Hugin_font.Font.of_stringof_string s is Ok f if s is an OpenType font file with TrueType (glyf), CFF (CFF ) or CFF2 (CFF2) outlines, and Error e otherwise.val · hugin
- Munin_float_text.Float_read.of_stringof_string s is the float nearest the decimal number s writes, ties to even, the same on every system, or None if s is not one number in JSON's syntax (number_end).val · munin
- Quill_markdown.of_stringof_string s is the notebook s.val · quill
- Rig.Buffer.of_stringof_string s is a buffer on host holding s's bytes, placed as create places a host buffer.val · rig
- Rig_amd_abi.Code_object.of_stringof_string obj reads the code object obj.val · rig
- Rig_bin.Line.of_stringof_string s is the line s, without its newline, or None.val · rig
- Rig_elf.of_stringof_string ~align ~held obj is the object obj laid out in its image (Reading), which holds the sections s with held s.val · rig
- Rig_nv_abi.Cubin.of_stringof_string obj is the cubin obj, the sections the GPU's memory holds laid out at an alignment of 128 bytes (Rig_elf.of_string).val · rig
- Rig_program.of_stringof_string s is the description to_string wrote as s.val · rig
- Talon.Binary.of_stringof_string s is the bytes of s.val · talon
- Tolk.Bigint.of_stringof_string s is the integer s writes: an optional - or +, an optional 0x, 0o or 0b for base 16, 8 or 2 (base 10 otherwise), then digits of the base, where underscores may follow the first digit.val · tolk
- Tolk.Dtype.of_stringof_string s is the data type named s, in any case: its name (f32, i8), its constructor's name without underscores (float32, int8, bfloat16, weakint), a C-flavoured alias (half, float, double, char, uchar, short, ushort, int, uint, long, ulong), or default_float and default_int for default_float and default_int.val · tolk
- Tolk.Graph.of_stringof_string text is the sink of the graph text, the node of its last line.val · tolk
- Tolk.Helpers.Target.of_stringof_string s reads s, written [INTERFACE[:INDICES]+]DEVICE[:RENDERER[:ARCH]].val · tolk
- Tolk.Op.of_stringof_string s is the operation whose name is s, case included: of_string "ADD" is Ok Add, and of_string "add" and of_string "Ops.ADD" are errors.val · tolk
- Tolk.Ops.Axis_type.of_stringof_string s is the axis type named s in upper case, without its AxisType.val · tolk
- Ymir_fits.Fits.Header.of_stringof_string ~name s reads 80-byte records back to back up to END, or, when s holds a newline, lines of at most 80 bytes up to END, as .head files hold them.val · ymir
- Ymir_spk.of_stringof_string ~at ~name s is read of the file whose bytes are s.val · ymir
- Ymir_units.Unit.of_stringof_string s is Ok u iff to_string u = s, and Error msg otherwise.val · ymir
- Nx_ragged.of_stringsof_strings ss is the ragged array on the host whose row i holds the bytes of ss.(i) as uint8 values, from offset 0.val · nx
- Munin_float_text.Float_read.of_substringof_substring s pos len is of_string (String.sub s pos len), without the copy.val · munin
- Talon.Query.of_tableof_table t is the query of the rows of the table t.val · talon
- Nx.Rng.of_tensorof_tensor t is the key, or batch of keys, whose words are t, for words that were saved: the inverse of to_tensor.val · nx
- Talon.Column.of_tensorof_tensor ?validity x is the column of x's rows:val · talon
- Ymir.Earth.of_ttof_tt t ~delta_t is the Earth's orientation at the TT instants t, with ΔT = TT − UT1 delta_t, as of_utc reads its offset.val · ymir
- Tolk.Renderer.Estimates.of_uopsof_uops ~ignore_indexing uops estimates the linearized kernel uops:val · tolk
- Talon.Time.of_usof_us n is the stamp n microseconds after 1970-01-01 00:00:00 of its clock, or None if it is out of range.val · talon
- Talon.Time.Span.of_usof_us n is n microseconds, or None if the span is out of range.val · talon
- Ymir.Earth.of_utcof_utc t ~dut1 is the Earth's orientation at the UTC instants t, with UT1 − UTC dut1: UT1 is t's day and seconds since 0h moved by dut1 over days of 86400 s, so a leap second reads as SOFA's iauUtcut1 reads it.val · ymir
- Nx_io.Archive.of_valueof_value ~prefix s x is the archive with one entry for each array s walks in x, fixed arrays included, named under prefix as the module's preamble states.val · nx
- Hugin_kit.Ticks.of_valuesof_values ~locale ~notation s vs is the ticks at the values of vs that lie in the domain of s and are not missing for it, each once, labelled with the strings of locale, which defaults to Locale.default, and in notation if given (Labels).val · hugin
- Norn.Summary.of_weightedof_weighted u w summarises the weighted draws w of u, their log weights raw: it smooths them as Norn.Diag.psis does, and judges them by the raw weights' k̂.val · norn
- Ymir.Direction.of_xyzof_xyz f v is v / |v| along v's last axis.val · ymir
- Hugin_kit.Stack.offsetThe type for stack offsets.type · hugin
- Nx_array.Layout.offsetoffset l is the position of l's index (0, …, 0).val · nx
- Rig.Buffer.offsetoffset b is the number of bytes of b's memory before b's first byte.val · rig
- Talon.Expr.Temporal.offsetoffset step a is a, a date or a datetime, moved by step: an exact step adds ticks, and a calendar step moves the calendar date of a wall clock; a day of the month past the month's end becomes its last day.val · talon
- Talon.Tz.offset_soffset_s z t is the UTC offset of z at the POSIX second t, in seconds.val · talon
- Brot.Encoding.offsetsoffsets enc is the (start, end_) byte span of each token in the text that was encoded, before normalization: a token of "café" normalized to "cafe" reports the bytes of the accented text.val · brot
- Nx_ragged.offsetsoffsets r is r's offsets, length r + 1 of them.val · nx
- Ymir_units.Unit.ohmohm is kg m2 s-3 A-2.val · ymir
- Jera.Solution.okok s is is Converged s.val · jera
- Hugin_kit.Scheme.okabe_itookabe_ito is the eight colours of Okabe and Ito, which stay apart for viewers with any common colour vision deficiency: #000000 black, #e69f00 orange, #56b4e9 sky blue, #009e73 bluish green, #f0e442 yellow, #0072b2 blue, #d55e00 vermilion and #cc79a7 reddish purple.val · hugin
- Nx.Devices.onon is Placement.on v.val · nx
- Nx.Host.onon is Placement.on v.val · nx
- Nx.Placement.onon s holds the whole value on every device of s.val · nx
- Tolk.Ops.on_diskon_disk u is true iff u lives on a disk device.val · tolk
- Tolk.Heuristic_host.on_hoston_host k is true iff k renders for the CPU device.val · tolk
- Fehu.Render.on_renderon_render ~sink env wraps env so that every rendered frame after Env.reset and Env.step is passed to sink.val · fehu
- Nx_array.Dtype.oneone dt is dt's multiplicative identity: 1, and true for booleans.val · nx
- Tolk.Bigint.oneone is 1.val · tolk
- Ymir_units.Unit.oneone is the dimensionless unit 1.val · ymir
- One callThis page follows one call through rune: the gradient of a compiled function that holds a scan.guide · rune
- Vega.Schedule.one_cycleone_cycle ~max_value ~total_steps ?div_factor ?final_div_factor ?pct_start () is the 1cycle schedule.val · vega
- Nx.one_hotone_hot ~num_classes i is i's shape with an axis of num_classes appended, holding 1 at the class each element of i names and 0 at the others: an element outside [0, num_classes) gives a row of zeros.val · nx
- Talon.Kit.one_hotone_hot c q is q with its categorical column c replaced, in place, by one bool column per category, in dictionary order, named c ^ "_" ^ category: true where c is that category, false where it is another, and null where c is null:val · talon
- Kaun.Init.onesones fills with 1.0.val · kaun
- Nx.onesones dt s is the host value of shape s whose every element is one (true for booleans).val · nx
- Nx.ones_atones_at p dt s is ones dt s at p.val · nx
- Nx.ones_likeones_like x is ones (dtype x) (shape x).val · nx
- Talon.Expr.onlyonly a is a's one distinct non-null value, and a data error naming the frame when a has several.val · talon
- Rig_mlx5_abi.Entry.opThe type for operations.type · rig
- Tolk.OpOperations.module · tolk
- Tolk.Ops.opop u is u's operation.val · tolk
- Tolk.Ops.Upat.opop o is v ~op:(Op.Set.of_list [o]) ().val · tolk
- Nx_kernel.Prog.E.op1op1 k dt x is Op1 (k, dt, x).val · nx
- Nx_kernel.Prog.op1The type for kinds of one operand.type · nx
- Nx_kernel.Prog.op2The type for kinds of two operands.type · nx
- Nx_kernel.Prog.E.op3op3 k c x y is Op3 (k, c, x, y).val · nx
- Nx_kernel.Prog.op3The type for kinds of three operands.type · nx
- Tolk.Ops.op_in_backward_slice_with_selfop_in_backward_slice_with_self ~calls u ops is true iff u or a node of backward_slice ~calls u has an operation in ops.val · tolk
- Hugin.Role.opacityopacity is an opacity in [0;1].val · hugin
- Hugin_vg.Picture.opacityopacity a p paints p faded to the opacity a: p is painted on its own, over a transparent backdrop, and the result is composited with its alpha multiplied by a.val · hugin
- Tolk.Postrange.Scheduler.opaqueopaque k is k's opaque operations, but those of the bodies it calls: its calls (Ops.call), of code in the target's language among them (Ops.custom_function).val · tolk
- Tolk.Ops.opaque_call_bodiesopaque_call_bodies is the operations a call body can have: Op.t.Sink, Op.t.Program, Op.t.Linear, Op.t.Store and Op.t.Custom_function.val · tolk
- Munin.Store.open_open_ () is the store at root, made absolute against the current directory.val · munin
- Rig.open_open_ ~machine o is the open device named o.name on machine's machine (defaults to this machine).val · rig
- Rig_bnxt.open_open_ ~machine i holds NIC i of machine (defaults to Rig_pci.Machine.this) and boots it.val · rig
- Rig_mlx5.open_open_ name opens port 1 of the NIC name and maps its access regions, the NIC's pages through which the process rings doorbells.val · rig
- Rig_pci.Driver.open_open_ d m i f is f h fn, with the process holding function i of m by h and the function fn taken (Function.take).val · rig
- Rig.open_hostopen_host ~machine o is open_ of the host of the machine named machine, a driver's device of that machine that runs submissions and loads the code its driver takes.val · rig
- Rig.open_ioopen_io ~machine o is open_ for an io device (Rig_edge.Io).val · rig
- Rig_edge.openerThe type for openers of named devices.type · rig
- Sowilo.openingopening ~kernel img is the erosion of img followed by the dilation by the element reflected about its centre, whose neighbourhood holds the pixel dy rows above and dx columns left where kernel's holds the pixel dy rows below and dx columns right.val · sowilo
- Nx_kernel.operandThe type for a contraction's operand: an array, through any layout (strided, broadcast, windowed), never padded, and its scale.type · nx
- Rig_nv_abi.Template.operandThe type for the values of a template.type · rig
- Tolk.Tc.operandThe type for operands of a product.type · tolk
- Tolk.Ops.operand_boundsoperand_bounds u s is the bounds of u's source s as u reads it: an operation on a committed integer commits a weak integer operand to its type, which wraps it, so bounds that leave the type are one value wrapped, or the type's; an operation on a committed float commits a weak float operand to its type, which rounds it, and its bounds round with it.val · tolk
- Tolk.Stage.operandsoperands ren sink is the kernel sink, whose optimisations are applied, with each tensor core's operands staged (depth) that read global memory through an index whose terms that read the workgroup's threads and the thread's lanes form a dense box of strided dimensions: each element read once, and none of those terms reading another range.val · tolk
- Operations ReferenceEvery operation in sowilo, organized by category.guide · sowilo
- Rig_pci.Machine.opsThe type for the operations of a transport.type · rig
- Tolk.OpsThe intermediate representation: nodes, patterns and graph rewriting.module · tolk
- Tolk.OptKernel optimisations.module · tolk
- Optax ComparisonThis page maps Optax concepts and API to their Vega equivalents.guide · vega
- Talon.Query.optimizeoptimize q is the plan that running q runs: q with the same schema and the same rows, rewritten to read and compute less.val · talon
- OptimizersVega has no optimizer object.guide · vega
- Nx.Ptree.optionoption a reports presence as Walk.option does and walks the content with a at the same path.val · nx
- Nx.Ptree.Walk.optionoption walk c x reports whether x is present at c's path and, if it is, walks its content with walk at the same path.val · nx
- Talon.Expr.optionoption a is Some v where a is v and None where a is null.val · talon
- Talon.Column.optionsoptions k c is like values, with None for each null.val · talon
- Tolk.Ops.Upat.or_afteror_after p matches p or p ordered after anything.val · tolk
- Tolk.Ops.Upat.or_bitcastedor_bitcasted p matches p or a bitcast of it.val · tolk
- Tolk.Ops.Upat.or_castedor_casted p matches p or a cast of it.val · tolk
- Hugin_kit.Scheme.orangesoranges is Brewer's Oranges.val · hugin
- Rig_mlx5.Region.orderThe type for the order in which the NIC's writes land in a region.type · rig
- Talon.OrderSort keys.module · talon
- Nx.order_keyorder_key kd x is, for each element of x, an unsigned integer of kd whose order is the sort order: two keys compare, read unsigned, as their elements sort, and are equal iff the elements are equal in the sort order.val · nx
- Norn.Bij.orderedordered maps a vector u of K coordinates to the strictly increasing vector x with x0 = u0 and xk = x(k-1) + exp uk.val · norn
- OrderingThis page shows, with programs, how rig orders the work of several devices by the memory it uses.guide · rig
- Brot.Normalizer.original_spanoriginal_span a ~start ~stop is the span of the original text that the normalized bytes [start;stop) come from: from the first byte the byte at start comes from to the last byte the byte at stop - 1 comes from.val · brot
- Munin.History.orphansorphans h k is k's orphans, in log order.val · munin
- Hugin_kit.Scheme.orrdorrd is Brewer's OrRd, orange to red.val · hugin
- Talon.Expr.outThe type for outputs of shape 's: named expressions, as select, derive and aggregate take them.type · talon
- Rig.Out_of_memoryOut_of_memory (d, n) is raised when d cannot allocate n bytes, immediately or after the rounds of reclamation.exception · rig
- Nx.outerouter a b is of shape [|numel a; numel b|], its element (i, j) the product of a's ith element and b's jth, in C order.val · nx
- Hugin_font.Font.outlineoutline f g is the outline of g, unhinted, to be filled under the nonzero rule.val · hugin
- Hugin_font.Font.outlinesoutlines f is the format of the outlines of f: `Truetype for quadratic glyf outlines and `Cff for cubic CFF or CFF2 ones.val · hugin
- Quill.Cell.outputtype · quill
- Rig.Profile.output_chrome_traceoutput_chrome_trace oc events writes events to oc in Chrome's trace event format, JSON, which Perfetto and chrome://tracing load.val · rig
- Quill.Cell.output_of_tagoutput_of_tag s is the output Quill shows for the tag s:val · quill
- Talon.Expr.overover ~by ~order e is e evaluated in each partition of the enclosing frame (Frames).val · talon
- Tolk.Bigint.OverflowRaised when a conversion's result does not fit its type.exception · tolk
- Brot.Encoding.overflowingoverflowing enc is the list of overflow encodings produced by truncate when the input exceeds max_length.val · brot
- Tolk.Ops.overflowsoverflows u dt is true iff u's bounds reach outside dt's.val · tolk
- Hugin_vg_vector.Vector.overlapsoverlaps cut minx miny maxx maxy is true iff the box of these bounds meets cut, and false if a bound is NaN.val · hugin
- Rig.Buffer.overlapsoverlaps b b' is true iff b and b' share memory: they hold bytes of one word (Words), through views of one allocation, a borrow and the memory it maps, or two bigarrays over bytes of one word.val · rig
- Talon.Join.overlapsoverlaps (ls, le) (rs, re) matches the rows whose left period (ls, le) and right period (rs, re) overlap, as SQL-92's OVERLAPS predicate defines it (ISO/IEC 9075:1992, 8.11), in talon's total order.val · talon
- Rig_mlx5_abi.Completion.ownedowned b at ~count ~entries is true iff the completion at byte at of b is the one the NIC wrote for consumer count count of a ring of entries entries: its opcode is valid and its owner bit is bit log2 entries of count.val · rig
p
- Hugin_gg.P2Points of the plane.module · hugin
- Rig_pci.Page_table.pa_bitspa_bits t is the number of bits of a physical address t's entries hold: pa_bits of its format.val · rig
- Fehu.Space.packpack s v is v converted to the universal Value.t representation.val · fehu
- Tolk.Codegen.pack_argspack_args sink params is the lowered kernel sink reading its parameters params as bytes: one parameter, args, of eight bytes per parameter, holds each at its offset (Device.Tiny_elf.iter_sig).val · tolk
- Fehu.Space.packedtype · fehu
- Kaun.Cache_index.packedpacked ~seq lens is the whole index of lanes that each hold several sequences one after another: lane b holds sequences of lens.(b).(0), lens.(b).(1), ...val · kaun
- Nx.packedThe type for values whose dtype is hidden, such as a file's named values.type · nx
- Brot.Encoding.padpad enc ~target_length ~pad_id ~pad_type_id ~pad_token ~direction extends enc to exactly target_length tokens.val · brot
- Nx.padpad widths v x is x with fst widths.(i) elements v before its axis i and snd widths.(i) after it.val · nx
- Nx_kernel.padThe type for paddings of an operand of rank r, each array of lo, hi and interior of length r.type · nx
- Sowilo.padpad (top, bottom, left, right) img zero-pads the spatial dimensions.val · sowilo
- Tolk.Shape.padpad ~value u padding adds, on each axis, the given numbers of elements of value (default 0) before and after it; a negative number removes elements.val · tolk
- Brot.pad_lengthThe type for padding length strategies.type · brot
- Tolk.Shape.pad_topad_to ~value u shape pads the end of each axis to shape.val · tolk
- Brot.pad_tokenpad_token t is t's padding token, if any.val · brot
- Nx_kernel.Shape.paddedpadded p x is the shape a Padded load by p of an operand of shape x has.val · nx
- Brot.paddingThe type for padding configurations.type · brot
- Brot.paddingpadding length is a padding configuration for the given length strategy.val · brot
- Hugin_vg_vector.Vector.pagepage r is the context of the page of r, cut at a margin of the page's larger side around it.val · hugin
- Rig_pci.Machine.pagepage m is the page size of m in bytes.val · rig
- Quill_book.Render.page_htmlpage_html wraps a content fragment in the full page template with navigation.val · quill
- Rig_pci.Page_tableA GPU's physical memory and the page tables that map it.module · rig
- Rig_bnxt.Region.pagespages nic a n is a region over the n bytes at address a of the process, which start on a page of nic's machine (Rig_pci.Machine.page), pinned for nic as Rig_pci.Function.pin pins them until the region is deregistered or nic closes.val · rig
- Rig_edge.Io.pagespages d r is r's bytes as host memory, or None if d maps none.val · rig
- Rig_pci.Memory.pagespages r is the memory r's bytes are in and the physical ranges that hold them, as (address, bytes), in order from its first byte: addresses of the GPU's memory, of a peer's memory for Page_table.target.Peer, and bus addresses for system memory.val · rig
- Hugin_kit.Symbol.paintThe type for the ways a symbol is painted, which decide how it is sized: `Fill by the area it encloses, `Stroke by the length of its outline.type · hugin
- Hugin_vg_vector.Vector.paintWhat paints a leaf of a kind: its own colour, the colour an enclosing instance sets and the leaf inherits, or a colour an instance drawn in full gives it.type · hugin
- Nx.pairThe 0-d coordinates of one query-key pair: the query row, the key column, the row's visible length, its entry of lengths within [0, tk], and the number of query rows.type · nx
- Nx.Ptree.pairpair a b walks a pair's first component with a at Index 0, then its second with b at Index 1.val · nx
- Hugin_kit.Scheme.pairedpaired is Brewer's Paired, twelve colours in light and dark pairs.val · hugin
- Hugin.Theme.palettepalette th is the scheme of band scales of th.val · hugin
- Hugin_kit.Scheme.palettepalette cs is the qualitative scheme of the colours cs, in order.val · hugin
- Rig_pci.Page_table.pallocpalloc t n is the physical address of n' new bytes, n rounded up to 4 KiB, aligned to align (defaults to 4096) and zeroed if zero (defaults to true).val · rig
- pandas ComparisonThis page translates common pandas code into talon.guide · talon
- Hugin.Layout.panelThe type for panels.type · hugin
- Hugin.Layout.panelspanels l is the panels of l, in the order of the figure.val · hugin
- Hugin.Size.panelspanels w h gives each column of weight k (grid) a data area k *.val · hugin
- Hugin.Theme.paperpaper th is the paper of th.val · hugin
- Tolk.Setting.parallelparallel is the number of domains compiling kernels and running the beam search, from PARALLEL.val · tolk
- Hugin.Role.paramparam ~name ~equal is a new role name for a parameter of a mark, a value of any type: its domain unit holds no data, so only const binds it, and Mark.get gives the constant once per row.val · hugin
- Tolk.Device.Tiny_elf.paramThe type for the parameters of a program.type · tolk
- Tolk.Shape.paramparam ~shape slot dt is the parameter slot of type dt: a scalar without shape, flat storage of its greatest size viewed as shape otherwise, whose first element lies phase bytes past a multiple of align (defaults 0 and 16; see Ops.param_arg).val · tolk
- Tolk.Ops.param_argThe type for the arguments of Op.t.Param, Op.t.Buffer and Op.t.Alloc: storage, or a scalar variable.type · tolk
- Tolk.Ops.param_argparam_arg ~slot dtype is the argument with these fields.val · tolk
- Tolk.Shape.param_likeparam_like ~name u slot is a parameter in slot, named name (default none), that u can be passed to: a scalar variable without its value, or storage of u's shape.val · tolk
- Tolk.Cstyle.param_nameparam_name p is the name a kernel gives its parameter p.val · tolk
- Tolk.Shape.param_startparam_start body slot is the start (align, phase) that the storage parameter in slot of the call body body assumes (Ops.param_arg), or None if body takes no storage there.val · tolk
- Rig_nv_abi.Qmd.parametersparameters l d is the start of constant bank 0 for the launch of l with the values d, which the caller writes before the kernel's parameters.val · rig
- Munin.Run.paramsparams r is r's params, in the order its start gave them.val · munin
- Rig.Image.paramsparams fn is the number of parameter bytes, from byte 0, that fn reads, or None if its driver does not know it.val · rig
- Rig_nv.paramsThe type for the parameters of an RM call: bytes outside the OCaml heap, which the RM reads and writes in place.type · rig
- Rig_nv.paramsparams n is n zero bytes.val · rig
- Munin.Run.parentparent r is the id of the run r belongs to, if it named one.val · munin
- Quill_markdown.Edit.parseparse source parses source into blocks with byte ranges.val · quill
- Talon.Column.parseparse ty c is the column of type ty whose rows are the values that the rows of the string or binary column c write, null where c is null, or Error (row, reason) at the first non-null row that is not ty's text or holds a value ty does not, reason a phrase such as not a number.val · talon
- Talon.Expr.Str.parseparse ty a is the value of type ty that the text a writes, in the forms that talon.csv reads: true and false; decimal integers with an optional sign; decimal or scientific floats, inf, -inf and nan; ISO 8601 dates; and ISO 8601 datetimes, with an offset or Z for an instant and without one for a wall clock.val · talon
- Talon.Expr.Temporal.parseparse fmt ty a reads the text a in the format fmt as a value of ty, a date, datetime or clock type.val · talon
- Ymir_fits.Fits.Unit.parseparse ~scope s is the unit s spells.val · ymir
- Quill_project.parse_configparse_config s parses a quill.conf file.val · quill
- Ymir.Units.parsecparsec is 648000/π Unit.astronomical_unit, the distance at which one astronomical unit subtends one arcsecond.val · ymir
- Rig.Submission.partThe type for parts: a piece of a submission's work.type · rig
- Rig_program.partThe type for parts, as Rig.Submission.part.type · rig
- Rig_remote_proxy.Wire.partThe type for the parts of a handover, which the agent runs in order.type · rig
- Talon.Source.partThe type for parts: the units a source reads, such as files or row groups.type · talon
- Nx.Placement.partialpartial m sums cuts is mesh m cuts with the value a sum over the mesh axes sums.val · nx
- Ymir_units.Unit.pascalpascal is kg m-1 s-2.val · ymir
- Tolk.Ops.passThe type for how often a rewrite applies its rules to a node and its results.type · tolk
- Hugin_kit.Scheme.pastel1pastel1 is Brewer's Pastel1, nine colours.val · hugin
- Hugin_kit.Scheme.pastel2pastel2 is Brewer's Pastel2, eight colours.val · hugin
- Rig_nv_abi.Template.patchThe type for patches.type · rig
- Rig_nv_abi.Cubin.patchespatches c ~base is what c's relocations write over its image for an upload at the address base, in the order of its relocations: pairs (at, b), each writing the little-endian bytes b at the offset at.val · rig
- Hugin_gg.PathPaths.module · hugin
- Hugin_kit.Curve.pathpath c xs ys is curve c through the points (xs.(i), ys.(i)): one open subpath per run of at least two points (Missing points), in order, or Hugin_gg.Path.empty if there is none.val · hugin
- Hugin_kit.Symbol.pathpath paint a s is the outline of s at size a painted with paint, its centre at the origin.val · hugin
- Jera.Ode.pathpath y m ~tol ~budget f ~t0 ~t1 y0 is the solution from y0 at t0 to t1 as a function of time: a series of budget pieces, one per accepted step in increasing time, each the step's continuous extension.val · jera
- Munin.Artifact.pathpath a is the absolute path of a's blob, a file or a directory.val · munin
- Nx.Ptree.PathPaths from a value's root to one of its parts.module · nx
- Rig_amd.pathThe type for what a path gives a device.type · rig
- Rig_gpus.pathThe type for a path of rig.type · rig
- Rig_nv.pathThe type for what a path gives a device.type · rig
- Norn.Gaussian.pathfinderpathfinder u ?history ?elbo_draws ~steps lp k start fits each chain along its path to a mode (Pathfinder; Zhang, Carpenter, Gelman and Vehtari 2022).val · norn
- Rig_gpus.pathspaths ~firmware () is each path of rig.val · rig
- Nx.PatternAxis patterns.module · nx
- Talon.Expr.Str.patternThe type for patterns: what text is matched against.type · talon
- Tolk.Ops.Pattern_matcherPattern matchers.module · tolk
- Tolk.Transcendental.patternspatterns ~force ops rewrites each Op.t.Exp2, Op.t.Log2 and Op.t.Sin whose operation is not in ops, or every one if force, to xexp2, xlog2 and xsin.val · tolk
- Nx.Ptree.PayloadMaps and folds at any payload: a mask, a learning rate per array, a plan, dimensions.module · nx
- Tolk.Transcendental.payne_hanek_reductionpayne_hanek_reduction d is (r, q) with d = q * pi/2 + r and |r| <= pi/4, for 1 <= d, infinities excluded.val · tolk
- Rig_amd.peerpeer d d' is true iff map_peer d d' maps Device memory of d'.val · rig
- Rig_cuda.peerpeer d d' is true iff map_peer d d' maps Device memory of d'.val · rig
- Rig_edge.Driver.peerpeer d d' is true iff map_peer d d' maps Device memory of d'.val · rig
- Rig_metal.peerpeer d d' is true iff map_peer d d' maps Device memory of d'.val · rig
- Rig_nv.peerpeer d d' is true iff map_peer d d' maps Device memory of d'.val · rig
- Rig_remote_proxy.peerpeer d d' is true iff map_peer d d' maps Device memory of d'.val · rig
- Hugin_vg_vector.Vector.penThe type for pens as written: width, dash lengths and dash offset, in a frame.type · hugin
- Ymir_units.Quantity.perper u q is q's payload in Unit.(unit q / u).val · ymir
- Talon.Expr.per_listper_list ls r is, at each row, r over the frame of the elements of the lists that the outputs ls compute (Lists).val · talon
- Rig_mlx5_abi.Uar.per_regionper_region is 4, the doorbell registers of an access region.val · rig
- Peri-Event AnalysisAn electrophysiology session records when each neuron, or unit, fired and when each trial's events happened.guide · talon
- Nx.Rng.permutationpermutation n is a uniformly random ordering of 0, …, n - 1: the positions that sort n random 64-bit words.val · nx
- Nx.Rng.permutation_atpermutation_at p n is permutation n at p.val · nx
- Tolk.Shape.permutepermute u order is u with its axes in order, or u; negative axes count from the end.val · tolk
- Ymir_units.Unit.petapeta u is 1015 u.val · ymir
- Rig_pci.Page_table.pfreepfree t pa frees the block of physical memory at pa that palloc returned or that a mapping of alloc holds.val · rig
- Hugin_gg_kit.Pgon2Polygons.module · hugin
- Hugin_vg_vector.Vector.phasephase pen o is the dash offset of pen moved by o, within the length of its pattern made even.val · hugin
- Ymir_units.Unit.pipi is the number π.val · ymir
- Ymir_units.Unit.picopico u is 10-12 u.val · ymir
- Hugin_vg.PicturePictures.module · hugin
- Hugin_vg.Renderable.picturepicture r is the picture of r.val · hugin
- Talon.Expr.Str.piecespieces ss matches the strings ss in order without overlap: matches (pieces [ "special"; "requests" ]) is SQL's LIKE '%special%requests%'.val · talon
- Jera.PiecewisePiecewise Chebyshev series.module · jera
- Vega.Schedule.piecewise_constantpiecewise_constant ~boundaries ~values is a step function.val · vega
- Rig_pci.Function.pinpin f a n locks the n bytes at a of f's machine, which start on a page, and is Some runs, with runs the addresses at which f reaches them.val · rig
- Tolk.Simplify.pin_effectspin_effects sink is the kernel sink with each effect bound to the ranges it runs inside.val · tolk
- Nx.pinvpinv a is the […; n; m] Moore-Penrose pseudoinverse of a, from svd, or from eigh of the Hermitian a its lower triangle names under ~hermitian:true (default false).val · nx
- Rig_bin.Proc.pipepipe () is a pipe whose ends close on exec.val · rig
- Pipelines and IntegrationSowilo operations are pure functions on Nx tensors.guide · sowilo
- Talon.Expr.pivotpivot key v is, per frame, one record with a field per category of key's categorical type, in dictionary order, named as the category.val · talon
- Fehu.Render.Pixelmodule · fehu
- Ymir.Healpix.pixelpixel h d is the index of the pixel holding each direction of d, of d's batch shape, within the bound above, and -1 for a NaN or zero vector.val · ymir
- Ymir.Grid.pixelspixels ~at ~shape dtype t is the image of shape cells, [|rows; columns|], seen through t: the whole image, its own base, with measures at dtype and its window's start at at.val · ymir
- Ymir.Healpix.pixelspixels h is 12 nside², h's number of pixels.val · ymir
- Hugin_kit.Scheme.piygpiyg is Brewer's PiYG, pink to yellow-green.val · hugin
- Kaun.Cache_index.placeplace at index is index with its tensors whole on each device of at (Nx.Placement.uncut), as Nx.place: an index made from OCaml values (rows, whole, packed) lies on the host, and a model on other devices reads it where it computes.val · kaun
- Kaun.Rope.placeplace at t is t with its cosines and sines whole on each device of at (Nx.Placement.uncut), where a model on at reads them.val · kaun
- Nx.placeplace p x is x's elements at p.val · nx
- Nx.Ptree.placeplace (module U) p x is x with each array, fixed ones included, at p, as Nx.place.val · nx
- Nx.Rng.placeplace p k is k at p, as Nx.place.val · nx
- Nx_quant.placeplace p w is w with each part at p, as Nx.place.val · nx
- Rig_elf.placeThe type for where a symbol is.type · rig
- Tolk.Shape.placeholderplaceholder ~slot shape dt is a parameter of shape and dt (dt committed, Dtype.strong), or workgroup or register storage for those address spaces, in slot (default Ops.unique_num).val · tolk
- Tolk.Shape.placeholder_likeplaceholder_like u slot is a placeholder in slot of u's greatest shape and type.val · tolk
- Norn.Dist.placementplacement d is Nx.beside of d's first parameter: where its bounds lie, and where sample places its key, so a draw lies where the parameters do whatever the key's placement.val · norn
- Nx.placementplacement x is where x's bytes lie.val · nx
- Nx.PlacementWhere a value's elements lie over its set.module · nx
- Nx.Rng.placementplacement k is where k's words lie, Nx.placement of to_tensor k.val · nx
- Ymir_units.Unit.planckplanck is h, 6.62607015·10-34 kg m2 s-1.val · ymir
- Ymir.Cosmology.planck2013Planck 2013 XVI (Planck Collaboration 2014, A&A 571, A16), Table 5, Planck+WP+highL+BAO, best fit: H0 = 67.77, ωb = 0.022161, ωc = 0.11889.val · ymir
- Ymir.Cosmology.planck2015Planck 2015 XIII (Planck Collaboration 2016, A&A 594, A13), Table 4, TT,TE,EE+lowP+lensing+ext: H0 = 67.74, ωb = 0.02230, ωc = 0.1188.val · ymir
- Ymir.Cosmology.planck2018Planck 2018 VI (Planck Collaboration 2020, A&A 641, A6), Table 2, TT,TE,EE+lowE+lensing+BAO: H0 = 67.66, ωb = 0.02242, ωc = 0.11933.val · ymir
- Ymir.Transform.planeThe type for points [...; n] in a plane.type · ymir
- Hugin_kit.Scheme.plasmaplasma runs from dark blue through magenta to yellow.val · hugin
- Tolk.Helpers.pluralizepluralize st cnt is cnt followed by st, with an s unless cnt is 1.val · tolk
- Hugin_kit.Symbol.plusplus is two strokes of equal length crossing at their middles, one parallel to each axis.val · hugin
- Tolk.Linearizer.pm_add_control_flowpm_add_control_flow adds to each range of a context's sink, as its last source, the loop or the range it runs after in that context.val · tolk
- Tolk.Uop_weak.pm_cast_constpm_cast_const states the width of every constant still weak, at each consumer: a broadcastable consumer that derives a committed type from its sources converts its weak constants to their least upper type, and every remaining bare constant becomes a cast of the literal to its committed type (Ops.cconst), booleans included.val · tolk
- Tolk.Symbolic.pm_clean_up_group_sinkpm_clean_up_group_sink splices into each sink the sources of its Op.t.Noop, Op.t.Stack, Op.t.Sink and Op.t.After sources.val · tolk
- Tolk.Uop_weak.pm_commit_weakpm_commit_weak commits weak sources whose width something states:val · tolk
- Tolk.Ops.pm_drop_afterpm_drop_after replaces each Op.t.After with its first source.val · tolk
- Tolk.Symbolic.pm_drop_and_clausespm_drop_and_clauses drops from each gate's condition the clauses that run in none of the gated value's ranges.val · tolk
- Tolk.Decomp_dtype.pm_dtype_decompspm_dtype_decomps notes the narrow floats and 64-bit integers a kernel uses, and at the kernel's Op.t.Sink rewrites it to emulate those that the target lacks, in the promotion order of their types (Dtype.compare).val · tolk
- Tolk.Schedule.pm_flatten_linearpm_flatten_linear inlines each Op.t.Linear among the sources of an Op.t.Linear into it.val · tolk
- Tolk.Simplify.pm_flatten_rangepm_flatten_range replaces the sources of an Op.t.Reduce or Op.t.End after its value by the ranges they are or run inside, in order, each once.val · tolk
- Tolk.Gpudims.pm_group_gpudimspm_group_gpudims applies group_gpudims to sinks.val · tolk
- Tolk.Codegen.pm_linearize_cleanupspm_linearize_cleanups makes a gated Op.t.Store a store inside an Op.t.If on its gate, closed by an Op.t.Endif.val · tolk
- Tolk.Simplify.pm_load_collapsepm_load_collapse replaces a sum by load_collapse's, where it is one; and solves x + y < c for x when x is a Dtype.t.Weak_int that reads memory and y and c do not.val · tolk
- Tolk.Uop_weak.pm_lower_weakpm_lower_weak gives every weak node that nothing committed its default width (Ops.commit_dtype, with Dtype.t.Int32 as the default integer):val · tolk
- Tolk.Prepare.pm_mopspm_mops moves movements towards the storage they view:val · tolk
- Tolk.Gater.pm_move_gates_from_indexpm_move_gates_from_index moves gates onto loads and stores, with g a boolean node:val · tolk
- Tolk.Symbolic.pm_move_where_on_loadpm_move_where_on_load turns a selection between a load's index and 0 into a load gated by the selection's condition: each clause that the index does not already require, that is not constant, that runs in no range the index lacks and that reads no other Op.t.Index moves into the index's gate; the others stay in the selection.val · tolk
- Tolk.Gpudims.pm_range_to_specialpm_range_to_special replaces each Ops.Axis_type.t.Global and Ops.Axis_type.t.Local range by the hardware index (Op.t.Special) of its axis, its identity's last part: gidx0, lidx1, ..., keeping its bound, and splits ends (Linearizer.pm_split_ends), which drops the hardware indices from the ranges they close.val · tolk
- Tolk.Simplify.pm_reduce_collapsepm_reduce_collapse is the rewrites that compute a sum over a range in closed form: pm_reduce_unparented; integer comparisons x + y < c and x * y < c solved for x when y and c do not depend on a range and no value wraps (Ops.exact); the sum of a value over the part of a range where it is selected by bounds on the range, as the size of that part times the value, and +0.val · tolk
- Tolk.Simplify.pm_reduce_simplifypm_reduce_simplify is pm_reduce_unparented, and replaces a sum over ranges (Op.t.Reduce of Op.t.Add) that pm_reduce_collapse computes without any range by its closed form.val · tolk
- Tolk.Simplify.pm_reduce_unparentedpm_reduce_unparented removes from a sum, product or maximum (Op.t.Reduce) the ranges its value does not run inside: a sum over such ranges is multiplied by the product of their sizes, and a product raised to it, where that product is positive; where it is not, the reduction is its operation's identity.val · tolk
- Tolk.Shape.pm_remove_invalidpm_remove_invalid replaces each Ops.invalid that a gate or a stack holds by 0 of the holder's type, once the gates have been read.val · tolk
- Tolk.Simplify.pm_simplify_rangespm_simplify_ranges simplifies a kernel's ranges, with a context that starts empty:val · tolk
- Tolk.Symbolic.pm_simplify_validpm_simplify_valid simplifies each boolean conjunction with simplify_valid, and the weak integer value of each gate knowing its condition holds (uop_given_valid, without try_simplex) when it involves no Op.t.Index.val · tolk
- Tolk.Linearizer.pm_split_endspm_split_ends rewrites an Op.t.End of u into nested ends of one range each around u.val · tolk
- Tolk.Simplify.pm_split_rangespm_split_ranges splits a range of constant size n whose remainder by a constant c dividing n is taken, and that is neither a warp nor a device axis, into an outer range of size n / c and an inner one of size c, with a context that starts empty.val · tolk
- Tolk.Ops.pm_substitutepm_substitute replaces each node its context maps by its image.val · tolk
- Tolk.Shape.pm_uncast_constpm_uncast_const removes the cast from each committed constant source of a broadcastable node, leaving the bare weak literal, when the node's sources still derive the same least upper type and the node the same type.val · tolk
- Tolk.Simplify.pm_unpin_effectspm_unpin_effects undoes pin_effects where it is redundant: an Op.t.Index, or an effect of the index, whose storage is read through an Op.t.After of ranges alone reads it directly where it depends on those ranges without it.val · tolk
- Tolk.Tc.pm_wmma_fp8pm_wmma_fp8 dt makes a tensor core product of 8 lanes of 8-bit floats into Dtype.t.Float32 a product of its operands' bits as dt: the words in which AMD's builtins take them.val · tolk
- Hugin_vg_raster.pngpng ~density r is the PNG file of render ~density r: 8-bit RGBA, with a pHYs chunk of Float.round (density *.val · hugin
- Hugin.Coord.pointpoint p x y is the point of the page, in points, at which p puts the normalised position (x, y).val · hugin
- Rig.PointPoints on timelines.module · rig
- Point-in-Time FeaturesA model that predicts from an event's features may use only what was known when the event happened.guide · talon
- Hugin.Mark.pointspoints rows is (xs, ys), row i being at (xs.(i), ys.(i)) in points on the page: its positions mapped by the panel's projection.val · hugin
- Norn_model.pointwisepointwise m y p is the log likelihood of every point, in the order the observed sites and factors are performed: the points of a site are axis 0 of its factors (Norn.Dist.factors), and a scalar is one point.val · norn
- Norn.Dist.poissonpoisson ~rate is the count with probability rate^x exp (-rate) / x!, rate non-negative and finite.val · norn
- Nx.Rng.poissonpoisson rate is Poisson counts of rate, elementwise; a rate of 0 gives 0.val · nx
- Talon.Expr.Temporal.policyThe type for what localize and localize_ordered do with a reading that a zone repeats or skips, as the section's preamble describes.type · talon
- Rig_bnxt.Qp.pollpoll qp side takes the completion of the oldest entry of side whose completion has not been taken, if it has one.val · rig
- Rig_mlx5.Cq.pollpoll cq is the next completion of cq, which it consumes, or None if the NIC wrote none after those poll consumed.val · rig
- Hugin_gg.Path.polygonpolygon xs ys is polyline xs ys closed.val · hugin
- Ymir.Region.polygonpolygon dtype p vertices is the polygon whose vertices, on axis −2 of vertices in either orientation, are mapped by p into its plane, with edges straight there.val · ymir
- Hugin_gg.Path.polylinepolyline xs ys is the open subpath through the points (xs.(i), ys.(i)) in order, or empty for fewer than two points.val · hugin
- Vega.Schedule.polynomial_decaypolynomial_decay ~init_value ~end_value ~decay_steps ?power () decays from init_value to end_value over decay_steps using a polynomial schedule: end_value + (init_value - end_value) * (1 - step/decay_steps)^power.val · vega
- Kaun.Pool2-D max and average pooling.module · kaun
- Tolk.Shape.poolpool ~stride ~dilation u kernel is the windows of kernel over the last axes of u, each stride apart (default 1) and dilation between its elements (default 1): u's leading axes, then the number of windows along each pooled axis, then the kernel's axes.val · tolk
- Tolk.Ops.pop_constpop_const ~op u is (x, c) for u = op x c with c a constant, and (u, identity_element op) otherwise.val · tolk
- Tolk.Bigint.popcountpopcount n is the number of one bits of n.val · tolk
- Rig_remote.portport a is the port a listens at: the one listen was given, or the one the system chose for 0.val · rig
- Porting from the First HuginThe first hugin built a plot from mark functions with options per mark, then set titles, labels, limits and scales on the plot with functions of their own.guide · hugin
- Talon.Join.positionposition matches row i of the left with row i of the right.val · talon
- Ymir_spk.positionposition t ~target ~center tdb is state's position.val · ymir
- Ymir.Direction.position_angleposition_angle a b is b's bearing from a, east of north, in [+0, 2π) radians.val · ymir
- Hugin.Mark.positionspositions rows is (us, vs), row i being at the normalised position (us.(i), vs.(i)): its normalised x and y, 0.5 for a position the mark does not bind, the centre of its band on a band scale and the end of its interval for a stacked position (bind).val · hugin
- Hugin_vg_instances.Instances.positionspositions cut ~reach b is the box of the positions at which an instance of scale 1.val · hugin
- Kaun.Cache_index.positionspositions index, of shape [| batch; seq |], is the tokens' positions clamped to 0 and the last column of the positions' table (seq - 1 on a whole index), for rotating and for indexing a table of position embeddings: padding is 0.val · kaun
- Nx.positionspositions c is each index i of the 1-d c repeated c.{i} times, in increasing order.val · nx
- Rig_mlx5.Qp.postpost qp e writes e in qp's send ring.val · rig
- Brot.post_processorpost_processor t is t's post-processor, if any.val · brot
- Brot.Post_processorPost-processing tokenization output with special tokens.module · brot
- Hugin.Theme.posterposter is v ~size:20.val · hugin
- Tolk.PostrangeApplying kernel optimisations.module · tolk
- Hugin_font.Font.postscript_namepostscript_name f is the PostScript name of f (name ID 6), the name PDF output gives the font, read as family reads names, or "" if f has none.val · hugin
- Hugin_kit.Scale.powpow ~exponent () is a power scale with exponent exponent, which must be finite and positive.val · hugin
- Nx.powpow a b is a to the power b.val · nx
- Nx_kernel.Prog.E.powval · nx
- Tolk.Bigint.powpow b e is b to the power e.val · tolk
- Tolk.Ops.Elementwise.powpow x y is Op.t.Pow of x and y: promoted to a common type, but for an integer x and a weak integer y, which keeps its value at no width (Op.t.Pow's exponent is a number).val · tolk
- Tolk.Ops.powpow x y is Op.t.Pow of x and y: promoted to a common type, but for an integer x and a weak integer y, which keeps its value at no width (Op.t.Pow's exponent is a number).val · tolk
- Tolk.Ops.Upat.powpow x y is Op.t.Pow of x and y: promoted to a common type, but for an integer x and a weak integer y, which keeps its value at no width (Op.t.Pow's exponent is a number).val · tolk
- Ymir_units.Quantity.powpow n q is q's payload to the power n in Unit.(unit q ** n), computed by repeated products: integers wrap, and the exponent is exact whatever the payload's dtype, so pow 17 of -1.val · ymir
- Tolk.Transcendental.pow2ifpow2if q dt is 2q as a float, for q an integer in the range of normal exponents of that float: Dtype.t.Float64 for an Dtype.t.Int64 q, Dtype.t.Float32 for an Dtype.t.Int32 q, and dt for an Dtype.t.Int16 q.val · tolk
- Nx.pow_spow_s x c is pow x c.val · nx
- Brot.Decoder.pppp ppf decoder formats decoder for debugging.val · brot
- Brot.Encoding.pppp ppf enc formats enc as a table for inspection, one row per token: index, token, ID, byte offsets, word ID, type ID, attention and special-token masks.val · brot
- Brot.Normalizer.pppp ppf n formats n for inspection.val · brot
- Brot.Post_processor.pppp formats a post-processor for inspection.val · brot
- Brot.pppp formats a tokenizer for inspection.val · brot
- Brot.Pre_tokenizer.pppp ppf t formats t for inspection.val · brot
- Fehu.Info.pppp formats an info dictionary for debugging.val · fehu
- Fehu.Value.pppp formats a value for debugging.val · fehu
- Hugin.Coord.pppp ppf c formats c for debugging.val · hugin
- Hugin.Drawing.pppp ppf d formats d for debugging.val · hugin
- Hugin.Layout.pppp ppf l formats the page size, panels, ticks and guides of l and the warnings, for debugging and baselines.val · hugin
- Hugin.pppp ppf f displays f in a Quill notebook: it renders f as SVG as save does by default, and prints on ppf a one-line summary of f inside a Format.String_tag.val · hugin
- Hugin.Resolved.pppp ppf r formats the scopes of r, their fitted scales and the warnings, for debugging and baselines.val · hugin
- Hugin.Size.pppp ppf s formats s for debugging.val · hugin
- Hugin.Theme.pppp ppf th formats th for debugging.val · hugin
- Hugin.View.pppp ppf view formats the keys of view and their values for debugging.val · hugin
- Hugin_float_text.Float_print.pppp ppf x formats g x.val · hugin
- Hugin_font.Font.pppp ppf f formats the family, weight and slant of f for debugging, the family as Run.pp formats text.val · hugin
- Hugin_font.Run.pppp ppf r formats r for debugging.val · hugin
- Hugin_gg.Affine.pppp ppf m formats the coefficients of m for debugging.val · hugin
- Hugin_gg.Box2.pppp ppf b formats b's corners for debugging.val · hugin
- Hugin_gg.Color.pppp ppf c formats to_hex c.val · hugin
- Hugin_gg.P2.pppp ppf p formats p as (x, y) for debugging.val · hugin
- Hugin_gg.Path.pppp ppf p formats the segments p was built from as SVG path data, one segment per token: M x y, L x y, C c1x c1y c2x c2y x y and Z, numbers printed with %g, non-finite ones included.val · hugin
- Hugin_gg.Stroke.pppp ppf s formats s for debugging.val · hugin
- Hugin_gg_kit.Pgon2.pppp ppf p formats the rings of p in order as Ring2.pp does, separated by spaces.val · hugin
- Hugin_gg_kit.Ring2.pppp ppf r formats the points of r as SVG path data, M x y for the first point, L x y for each other and a final Z, with numbers printed with %g.val · hugin
- Hugin_kit.Curve.pppp ppf c formats the name of c, such as monotone_x.val · hugin
- Hugin_kit.Dash.pppp ppf d formats d for debugging: solid, or its lengths, such as 4 2.val · hugin
- Hugin_kit.Locale.pppp ppf l formats l for debugging.val · hugin
- Hugin_kit.Number.pppp ppf f formats f for debugging, such as plain 2 decimals trim.val · hugin
- Hugin_kit.Scale.pppp ppf s formats the properties s sets, for debugging and tests, a custom transform by its name.val · hugin
- Hugin_kit.Scheme.pppp ppf s formats s for debugging: a named scheme by its name, such as viridis, a made one by its constructor and colours, such as ramp(#000000 #ffffff), and a reversed one as reverse(rdbu).val · hugin
- Hugin_kit.Symbol.pppp ppf s formats the name of s, such as triangle.val · hugin
- Hugin_kit.Ticks.pppp ppf t formats the positions, labels, contexts and note of t for debugging and tests.val · hugin
- Hugin_kit.Time.pppp ppf t formats t at UTC in RFC 3339 form, such as 2026-10-01T12:30:00.25Z: the fraction of a second appears iff it is not zero, to the nanosecond with its trailing zeros removed, and a year outside [0;9999] is written with its sign and at least four digits, as ISO 8601 extends years.val · hugin
- Hugin_literal.Literal.pppp ppf s formats s as an OCaml string literal of its bytes that keeps its characters: valid UTF-8 is written as is, except that double quotes, backslashes and the control characters U+0000 to U+001F and U+007F to U+009F are escaped as %S escapes them, as are the bytes of malformed UTF-8.val · hugin
- Hugin_text.Text.Layout.pppp ppf l formats l for debugging: its box and its runs.val · hugin
- Hugin_text.Text.pppp ppf t formats t for debugging, in spans of one style.val · hugin
- Hugin_vg.Picture.pppp ppf p formats p for debugging and tests as nested s-expressions, one per case: paths as SVG path data, colours as Hugin_gg.Color.pp prints them, images by the shape of their tensor and arrays in full.val · hugin
- Hugin_vg.Renderable.pppp ppf r formats the size and the picture of r for debugging.val · hugin
- Jera.Solution.pppp ppf s formats the count of lanes in each status and the report get would raise for the first lane that did not converge.val · jera
- Jera.Tol.pppp ppf t formats t as "rel 1e-06 abs 1e-10" or "ulps 4", each number as the shortest decimal that reads as it: "abs 1000000".val · jera
- Munin.Provenance.pppp ppf p prints p as one field: value line per field.val · munin
- Munin.Value.pppp ppf v prints v for people and for tab-separated output: a boolean as true or false, an integer in decimal, a float in the fewest significant digits that read back to it, with a fraction or an exponent when it is integral, or as nan, inf or -inf, and a string with each backslash and control character escaped as a JSON string escapes it, so that a tab prints as a backslash and t.val · munin
- Norn.Bij.pppp ppf b formats b's name, such as exp or compose(exp, affine).val · norn
- Norn.Dist.pppp ppf d formats d's family and its parameters' dtypes and shapes, as normal(loc: float64 [], scale: float64 [8]).val · norn
- Norn.Evidence.pppp formats an estimate as ln Z = -412.31 ± 0.08, H = 5.20 nats, followed, for a spent budget, by , budget spent with 0.52 nats left or , budget spent at β = 0.73.val · norn
- Norn.Gaussian.pppp u ppf g formats g's dimension and rank.val · norn
- Norn.Summary.pppp ppf s formats s as a table, then its findings, one per line.val · norn
- Norn.Support.pppp ppf s formats s as a set: (-inf, inf), (0, inf), (-inf, 0), (0, 1), {0, 1, 2, ...}, {0, 1, ..., 10}, {false, true}, or a phrase such as simplex of 3.val · norn
- Norn_model.pppp ppf m formats m's site table: each site's name, role, family, shape, points and coordinates.val · norn
- Nx.Placement.pppp formats a placement on one device as the device's name, others as on [CUDA:0; CUDA:1], split ~axis:0 [CUDA:0; CUDA:1], partial [CUDA:0; CUDA:1] or a mesh's extents, sums and cuts.val · nx
- Nx.pppp ppf x prints x's dtype and shape, on its set off the host, then its elements as nested rows, as float32 [3] [1, 2, 3] or, 0-d, float32 [] 1.5.val · nx
- Nx.Ptree.Path.pppp formats a path as to_string.val · nx
- Nx_array.Dtype.pppp formats a dtype's name.val · nx
- Nx_array.Layout.pppp formats a layout's shape, strides and offset.val · nx
- Rig.Point.pppp formats a point as NAME:v, or ended:v once device is None.val · rig
- Rig.pppp formats a device's name.val · rig
- Talon.Binary.pppp ppf b formats b as 0x followed by two lowercase hexadecimal digits per byte, first byte first: 0x48690a.val · talon
- Talon.Error.pppp ppf e formats e for people: the places e was found at, coarsest first, then its message, separated by ": ":val · talon
- Talon.Expr.pppp ppf e formats e as it is written inside Expr.( … ), with fewest parentheses: (amount -.val · talon
- Talon.Kind.pppp ppf k formats k by the name of the value that builds it: bool, int, float, string, binary, date, instant, wall and span.val · talon
- Talon.pppp is pp_with limits.val · talon
- Talon.Query.pppp ppf q formats q's plan, which reads no data: the line query → and q's schema as Schema.pp formats it, then q's last step, with the steps it reads below it in a tree:val · talon
- Talon.Schema.pppp ppf s formats the columns of s in order, separated by commas, each as its name and its type: carrier string, delay float64.val · talon
- Talon.Time.Date.pppp ppf t formats t as YYYY-MM-DD.val · talon
- Talon.Time.pppp ppf t formats t as its civil date and time of day in ISO 8601's extended format, YYYY-MM-DDThh:mm:ss, followed by a point and 3, 6 or 9 digits, the fewest that are exact, when t is not a whole second: 2024-03-15T09:30:00, 1969-12-31T23:59:59.999999999.val · talon
- Talon.Time.Span.pppp ppf d formats d as a minus sign when it is negative, then each nonzero component in hours, minutes, seconds, milliseconds, microseconds and nanoseconds, largest first, suffixed h, m, s, ms, us and ns: 168h, 1h30m, 1s500ms, -250us.val · talon
- Talon.Type.pppp ppf t formats t as schemas and plans show it:val · talon
- Tolk.Dtype.pppp formats a data type as dtypes.val · tolk
- Tolk.Helpers.Target.pppp formats a target as of_string reads it, without trailing separators.val · tolk
- Tolk.Op.pppp formats an operation as "Ops." followed by its name: Ops.ADD.val · tolk
- Tolk.Op.Set.pppp formats a set as its operations in declaration order, between braces and separated by commas: {Ops.ADD, Ops.MUL}.val · tolk
- Tolk.Ops.Axis_type.pppp formats an axis type as AxisType.val · tolk
- Tolk.Ops.pppp formats a node and its sources, one per line and indented, as the calls that build them:val · tolk
- Tolk.Opt.pppp formats Opt(op=OptOps.SPLIT, axis=0, arg=(2, AxisType.UPCAST)).val · tolk
- Tolk.Shape.Sint.pppp formats an integer in decimal, and a node as Ops.pp does.val · tolk
- Tolk.Tc.pppp formats a tensor core as its types and fragments: TensorCore(dtype_in=dtypes.f16, dtype_out=dtypes.f32, frag_a=(('k1', 'm0', ...), ('k0',)), frag_b=..., frag_c=...).val · tolk
- Ymir.Cosmology.pppp formats a cosmology's fields.val · ymir
- Ymir.Frame.pppp prints name.val · ymir
- Ymir.Transform.pppp formats a transform's stages.val · ymir
- Ymir_fits.Fits.Header.pppp formats a header one record per line, trailing spaces dropped.val · ymir
- Ymir_fits.Fits.Image.pppp formats the element and shape, as float32 [4800; 4600], and the scaling and BLANK.val · ymir
- Ymir_fits.Fits.pppp formats an HDU's kind, EXTNAME and EXTVER, then Image.pp, or its data size for other kinds.val · ymir
- Ymir_fits.Fits.Table.pppp formats the kind, rows, columns and heap size, then one line per column.val · ymir
- Ymir_units.Constant.pppp ppf k formats k's name, its text and its unit's canonical text, as in Newtonian constant of gravitation = 6.67430(15)e-11 kg^-1 m^3 s^-2.val · ymir
- Ymir_units.Quantity.pppp ppf q formats q's payload as Nx.pp does, then its unit's canonical text.val · ymir
- Ymir_units.Unit.pppp ppf u formats u's canonical text.val · ymir
- Ymir_units.Vocabulary.pppp voc formats a unit as spell voc writes it, as in MJy sr^-1: 1e<decade> when decade is not 0, then each word, its prefix written with the SI's symbol (μ for micro) and its exponent as in canonical text (Canonical text); 1 for no item.val · ymir
- Tolk.Dtype.pp_addr_spacepp_addr_space formats an address space as AddrSpace.GLOBAL, AddrSpace.LOCAL, AddrSpace.REG or AddrSpace.ALU.val · tolk
- Tolk.Ops.pp_argpp_arg formats an argument as the literal that denotes it: None, 42, ConstFloat(1.5), True, Invalid, dtypes.f32, (AxisType.WEAK, 0), (Ops.ADD, 0), 'CPU', b'ab\x00'.val · tolk
- Tolk.Dtype.pp_constpp_const formats a constant as a literal: True, False, Invalid, an integer in decimal, and a float in the fewest significant digits that read back as the same float.val · tolk
- Tolk.Ops.pp_devicepp_device formats a device as its quoted name: 'CPU'.val · tolk
- Nx.pp_dtypepp_dtype is Dtype.pp: float32.val · nx
- Hugin_font.Font.pp_errorpp_error ppf e formats e for users.val · hugin
- Tolk.Ops.pp_estimatespp_estimates formats Estimates(ops=0, lds=0, mem=0).val · tolk
- Hugin_float_text.Float_print.pp_exactpp_exact ppf x formats exact x.val · hugin
- Norn.Summary.pp_findingpp_finding ppf f formats f as a sentence.val · norn
- Talon_arrow.pp_formatpp_format ppf f formats f for people: a line naming the framing and counting its columns, then one line per column, with its name and type, as Talon.Schema.pp formats a column, and its Arrow type after an arrow, as in a list[string] ← list<item: utf8>.val · talon
- Talon_csv.pp_formatpp_format ppf f formats f for people: a line with the number of columns and the dialect, followed, when sniff made f, by the number of rows its types were sniffed from; then one line per column, with its name and its type as Talon.Schema.pp formats a column, and, aligned after them, declared for a column whose type with_type changed in a sniffed format:val · talon
- Talon_json.pp_formatpp_format ppf f formats f for people: a line with its number of columns and its framing, followed, when sniff made f, by the number of rows its types were sniffed from; then one line per column, with its name and its type as Talon.Schema.pp formats a column, and, aligned after them, declared for a column whose type with_type changed in a sniffed format:val · talon
- Talon_parquet.pp_formatpp_format ppf f formats f for people: a line that counts its columns, then one line per column, with its name and the type it reads as, as Talon.Schema.pp formats a column, and its Parquet type, as Parquet's schema language writes it, aligned after an arrow.val · talon
- Hugin_kit.Time.pp_intervalpp_interval ppf i formats i for debugging, such as 3 months.val · hugin
- Talon.Type.pp_namepp_name ppf n formats the name n of a field or a column as pp formats a field name: as is, or quoted when it must be.val · talon
- Hugin_kit.Number.pp_notationpp_notation ppf n formats n in lowercase, such as percent.val · hugin
- Tolk.Bigint.pp_printpp_print formats integers as to_string does, and raises as it does.val · tolk
- Rune.pp_programpp_program formats a program: the dtype, shape and placement of each constant it reads, its parameters, each equation with its results' dtypes, shapes and placements and its operands, a construct with the programs it carries indented, and its outputs.val · rune
- Hugin_kit.Scale.pp_propertypp_property ppf p formats p in lowercase, as the argument that sets it is named, such as domain or tz_offset_s.val · hugin
- Talon.Type.pp_quotedpp_quoted ppf s formats s quoted, as pp quotes a category.val · talon
- Nx.pp_shapepp_shape ppf s prints s as nx's messages do: [2; 3].val · nx
- Talon.Time.pp_steppp_step ppf s formats Months n as nmo, Weeks n as nw, Days n as nd, and Exact d as Span.pp formats d: 1mo, 2w, -1d, 15m.val · talon
- Nx_array.Dtype.pp_valuepp_value dt formats a value of dt:val · nx
- Nx.Ptree.pp_visitpp_visit ppf v formats v as its path and what is there: "blocks.3.moe: case \"mxfp4\"", "0.keys: a leaf", "layers: length 4", "bias: Some", "window: int 8".val · nx
- Hugin.pp_warningpp_warning ppf w formats w for users: the id, then the message.val · hugin
- Talon.pp_withpp_with l ppf t formats t for people, reading only the rows it shows.val · talon
- Brot.Pre_tokenizer.pre_tokenizepre_tokenize t text splits text into pretokens with byte offsets.val · brot
- Brot.pre_tokenizerpre_tokenizer t is t's pre-tokenizer, if any.val · brot
- Brot.Pre_tokenizerPre-tokenization.module · brot
- Hugin_kit.Number.precisionThe type for precisions.type · hugin
- Kaun.Metric.precisionprecision ?average predictions labels is the fraction of each class's predicted instances that are correct — tp / (tp + fp) — combined according to average, which defaults to `Macro.val · kaun
- Brot.Normalizer.precompiledprecompiled charsmap is the normalizer of a SentencePiece model, as T5, ALBERT and XLNet ship it.val · brot
- Talon.Source.PredPredicates on one row: the filters that sources apply.module · talon
- Tolk.Bigint.predpred n is n - 1.val · tolk
- Norn_model.predictpredict m k p is data drawn given the latent values p.val · norn
- Rig_edge.Io.prefetchprefetch d r ~at ~len asks d to read the len bytes at at in r ahead, before a device other than the host reaches them through pages.val · rig
- Talon.Expr.Str.prefixprefix s matches s at the start of the text.val · talon
- Talon.Sel.prefixprefix p selects the columns whose name starts with p, in schema order.val · talon
- Ymir_units.Vocabulary.prefixingThe type for whether a symbol takes an SI prefix.type · ymir
- Rig.Image.prepareprepare i f may start making i's function f ready to launch, and returns without waiting.val · rig
- Rig_amd.prepareprepare i f may start making what entry i f makes and return without waiting, so that the device makes the entries of several functions concurrently: it is entry's asynchronous start.val · rig
- Rig_cuda.prepareprepare i f may start making what entry i f makes and return without waiting, so that the device makes the entries of several functions concurrently: it is entry's asynchronous start.val · rig
- Rig_edge.Driver.prepareprepare i f may start making what entry i f makes and return without waiting, so that the device makes the entries of several functions concurrently: it is entry's asynchronous start.val · rig
- Rig_metal.prepareprepare i f may start making what entry i f makes and return without waiting, so that the device makes the entries of several functions concurrently: it is entry's asynchronous start.val · rig
- Rig_nv.prepareprepare i f may start making what entry i f makes and return without waiting, so that the device makes the entries of several functions concurrently: it is entry's asynchronous start.val · rig
- Rig_remote_proxy.prepareprepare i f may start making what entry i f makes and return without waiting, so that the device makes the entries of several functions concurrently: it is entry's asynchronous start.val · rig
- Tolk.PreparePreparing a tensor graph for ranges.module · tolk
- Tolk.Prepare.prepare_rangeifyprepare_rangeify sink is the tensor graph of the function sink in the form ranges index.val · tolk
- Brot.Normalizer.prependprepend s is a normalizer that prepends s to non-empty text.val · brot
- Brot.Pre_tokenizer.prepend_schemeControls when metaspace prepends the replacement character.type · brot
- Pretrained TokenizersMost users start by loading an existing tokenizer rather than building one from scratch.guide · brot
- Quill_project.prev_notebookprev_notebook t nb is the notebook before nb, or None.val · quill
- Hugin_kit.Scheme.prgnprgn is Brewer's PRGn, purple to green.val · hugin
- Nx.printprint x prints x by pp on standard output, then a newline.val · nx
- Talon.Column.printprint c is the column of the canonical texts of c's rows, null where c is null, which parse reads back: parse (type_ c) (print c) is c.val · talon
- Ymir_fits.Fits.Unit.printprint u spells u with vocabulary, as MJy sr-1, scoped symbols last by name; "" for the dimensionless unit.val · ymir
- Quill_book.Render.print_page_htmlprint_page_html ~book_title ~chapters renders a print page containing all chapters.val · quill
- Quill_book.Theme.print_template_htmlprint_template_html is the page holding every chapter, for printing.val · quill
- Norn_model.Init.priorprior draws coordinates from the prior.val · norn
- Tolk.Dtype.prioritypriority dt is dt's rank in the promotion order, from -1 for Void to 15 for Float64.val · tolk
- Rune.ProbeValues observed inside functions.module · rune
- Rig_bin.ProcProcesses: starting them, learning how they ended, ending them.module · rig
- Brot.Post_processor.processprocess t enc ~add_special_tokens adds special tokens and sets type IDs on enc.val · brot
- Munin_sys.processprocess () is a new reading of this process.val · munin
- Rig_remote_proxy.Wire.processThe type for the processes of a job.type · rig
- Rig_amd_abi.Gpu.processorprocessor g is the processor g runs code objects of, as LLVM names it: "gfx" followed by g.target's major in decimal and its minor and stepping in hexadecimal, such as "gfx1201" or "gfx90a".val · rig
- Nx.prodprod ~axes x multiplies the elements from the first, as sum, and a product of none is 1.val · nx
- Tolk.Helpers.prodprod l is the product of the elements of l, 1 if l is empty.val · tolk
- Tolk.Shape.Sint.prodprod ss is the product of ss, Int 1 if empty.val · tolk
- Rig.ProfileProfiles of devices' work.module · rig
- Nx_kernel.ProgScalar programs: what a loop computes at each index.module · nx
- Rune.programThe type for traced programs.type · rune
- Rune.programprogram a b f x is the program that jit traces of f, a function of structure a to structure b, at a call over x, without compiling or computing it: f runs once, over stand-ins of x's arrays in C order.val · rune
- Tolk.Comgr_worker.programval · tolk
- Tolk.Mtlcompiler_worker.programval · tolk
- Tolk.Nvrtc_worker.programval · tolk
- Tolk.Spec.programprogram is shared as programs restrict it: every width is stated, so a constant appears only under the cast that types it and nothing else is weak; there is no movement but a shrink of storage by a constant length, no global buffer, no parameter that names a device, since a program runs on the devices of its call's arguments, and no Ops.invalid.val · tolk
- Tolk.Ops.program_infoThe type for the arguments of Op.t.Program.type · tolk
- Tolk.Shape.program_info_of_sinkprogram_info_of_sink ~target sink reads the program's launch dimensions from its Op.t.Special nodes, each named variable they read at slot -1, as a launch binds it by name; its parameters and buffers from its Op.t.Param nodes; and which buffers it reads and writes from its loads and stores, and those of the bodies of the calls it makes at the arguments they pass.val · tolk
- Tolk.Codegen.program_keyprogram_key ast ren is the key under which to_program keeps the program of the kernel's sink ast for ren, in memory and on disk: two kernels of one key take one program.val · tolk
- Tolk.Codegen.programsprograms () is the number of programs to_program and read have made in this process, from their kernels or read back from disk.val · tolk
- Rig_pci.Driver.progressThe type for how far detach took a function.type · rig
- Hugin.Mark.projectproject rows p is p, given in the panel's normalised positions, cut at the domain's edges (Path.crop) and mapped into points on the page by the panel's projection: an open subpath keeps its pieces within the domain, and a closed one the part of its region there, so fill a path closed.val · hugin
- Hugin.Coord.projectionThe type for projections: a coordinate system fitted to a panel's box.type · hugin
- Hugin.Mark.projectionprojection rows is the panel's projection, which puts the normalised position (x, y) at Coord.point (projection rows) x y on the page.val · hugin
- Tolk.Ops.promo_dtypepromo_dtype us is the common type of us: theirs if they share it, their least upper bound (Dtype.least_upper) otherwise.val · tolk
- Hugin_kit.Scale.propertyThe type for the properties of scales.type · hugin
- Rig_amd.Lock.protectprotect l f is f () run holding l, which it gives back as f returns or raises.val · rig
- Rig_nv.Lock.protectprotect l f is f () run holding l, which it gives back as f returns or raises.val · rig
- Ymir_units.Codata.proton_magnetic_momentproton_magnetic_moment r is μp in J T-1.val · ymir
- Ymir_units.Codata.proton_massproton_mass r is mp in kg.val · ymir
- Talon.Kind.provably_equalprovably_equal k0 k1 is Some Equal iff k0 and k1 are the same kind, neither being nor containing the extension kind: lists are equal when their elements are, and tensors when their dtypes are.val · talon
- Munin.ProvenanceHow and where a run started.module · munin
- Munin.Run.provenanceprovenance r is the provenance r's start recorded.val · munin
- Norn.Diag.psisThe type for smoothed importance weights.type · norn
- Norn.Diag.psispsis lr smooths the log importance ratios lr.val · norn
- Jera.Grid.ptreeptree dtype is the structure of grid values over breaks of dtype: the breaks of each axis as a list at breaks, and the coefficients at coefficients.val · jera
- Jera.Piecewise.ptreeptree s dtype is the structure of series with values of structure s over breaks of dtype: the breaks at breaks, the coefficients under coefficients, and the extension reported at extension as one of the cases "bounded", "hold" and "polynomial".val · jera
- Jera.Solution.ptreeptree s is the structure of answers of structure s, so a compiled function returns one: the answer at value, the error at error, the statuses at status, the counts at evaluations, the budget used at used and the report's data under report.val · jera
- Norn.Draws.ptreeptree u is the structure of draws of u: u's, with the same paths.val · norn
- Norn.Ensemble.ptreeptree u is the structure of states over u.val · norn
- Norn.Evidence.ptreeptree u is the structure of evidence estimates over u, so a compiled function may return one.val · norn
- Norn.Gaussian.ptreeptree u is the structure of Gaussians over u: mean, scale and directions, each walked as u, and variances.val · norn
- Norn.Hmc.ptreeptree u is the structure of states over u.val · norn
- Norn.Nested.ptreeptree u is the structure of states over u.val · norn
- Norn.Nuts.ptreeptree u is the structure of states over u.val · norn
- Norn.Smc.ptreeptree u is the structure of states over u.val · norn
- Norn.Stats.ptreeptree dtype is the structure of statistics at dtype, with tensors at lp, acceptance, step_size, n_steps, diverging, saturated and energy.val · norn
- Norn.Weighted.ptreeptree u is the structure of weighted draws of u: values, walked as u, and log_weights.val · norn
- Nx.PtreeStructures of arrays.module · nx
- Nx.Rng.ptreeptree is a key, or a batch of keys, as a description: one fixed array at the root, walked by walk, which rebuilds it as of_tensor does.val · nx
- Ymir.Direction.ptreeptree () is Nx.Ptree.v walk, the structure of directions.val · ymir
- Ymir.Earth.ptreeptree () is Nx.Ptree.v walk, the structure of orientations.val · ymir
- Ymir.Grid.ptreeptree () is the structure of grids: the kind, the base and window shapes and the dtype as static data, then the start and the transform.val · ymir
- Ymir.Observation.ptreeptree () is the structure of observations: the grid, then the data, variance, validity and area, each present or not.val · ymir
- Ymir.Region.ptreeptree () is the structure of regions: the shape's kind, its sizes and its placement.val · ymir
- Ymir.Time.ptreeptree () is Nx.Ptree.v walk, the structure of times.val · ymir
- Ymir.Transform.ptreeptree () is the structure of transforms: the number of stages, then each stage at its index, reporting its family, its sense and its static data (permutation, origin, code, frame, stated terms) and walking its parameters.val · ymir
- Hugin_kit.Scheme.pubupubu is Brewer's PuBu, purple to blue.val · hugin
- Hugin_kit.Scheme.pubugnpubugn is Brewer's PuBuGn, purple through blue to green.val · hugin
- Brot.Pre_tokenizer.punctuationpunctuation () separates punctuation characters from the text around them, whitespace included.val · brot
- Hugin_kit.Scheme.puorpuor is Brewer's PuOr which, in Brewer's order, runs from orange to purple.val · hugin
- Hugin_kit.Scheme.purdpurd is Brewer's PuRd, purple to red.val · hugin
- Hugin_kit.Scheme.purplespurples is Brewer's Purples.val · hugin
- Tolk.Helpers.Diskcache.putput ~table key value binds key to value in table.val · tolk
- PyTorch ComparisonThis guide explains how kaun relates to PyTorch, focusing on:guide · kaun
q
- Nx_quant.q4_kq4_k blocks is the Q4_K weight stored as blocks.val · nx
- Nx_quant.q6_kq6_k blocks is the Q6_K weight stored as blocks.val · nx
- Nx_quant.q8_0q8_0 blocks is the Q8_0 weight stored as blocks.val · nx
- Jera.Quad.qmcqmc key ~tol ~budget f box is the integral of f over each lane's box, of any dimension d up to 1111, by randomised quasi-Monte Carlo.val · jera
- Rig_nv_abi.QmdLaunch descriptors.module · rig
- Rig_bnxt.QpA queue pair is a send ring and a receive ring connected to one queue pair of a peer.module · rig
- Rig_mlx5.QpA queue pair is a send ring, and a receive ring if it was made with one, connected to one queue pair of a peer.module · rig
- Nx.qrqr a is (q, r) with a = q r, q's columns orthonormal and r upper triangular: q is […; m; k] and r […; k; n], or under ~factors:Complete […; m; m] and […; m; n].val · nx
- Nx_cpu.qrqr ~q ~r a stores q with orthonormal columns and r, upper triangular with +0 below its diagonal, a = q r.val · nx
- Nx_kernel.S.qrqr ~q ~r a stores q with orthonormal columns and r, upper triangular with +0 below its diagonal, a = q r.val · nx
- Nx_kernel.Shape.qrqr f a is the shapes of q and r: […; m; k] and […; k; n] under Reduced, […; m; m] and […; m; n] under Complete.val · nx
- Jera.QuadIntegrals.module · jera
- Nx.Ptree.quadquad a b c e walks a quadruple's components with a, b, c and e at Index 0 to Index 3, in that order.val · nx
- Hugin_gg.Path.quad_toquad_to c pt p is p with a quadratic Bézier from the current point to pt with control point c, stored as the cubic of the same curve.val · hugin
- Norn.Dist.quantilequantile d p is the value whose CDF is p, elementwise, read from the smaller tail: above 1/2 it inverts the survival at 1 - p, which is exact there.val · norn
- Nx.quantilequantile ~axis qs x is the quantiles qs of x along axis, or of all its elements without axis.val · nx
- Nx_ragged.quantilequantile qs r is Nx.quantile qs of each row of r, over the row's elements: entry (i, j) is quantile qs.(i) of row j, of shape [|Array.length qs; length r|].val · nx
- Talon.Expr.quantilequantile p a is the p-quantile of a's values, interpolating linearly between the two nearest ranks.val · talon
- Ymir_units.Constant.quantityquantity d k is k's value as a host scalar of d in unit k: its magnitude rounded once to d as Unit.ratio rounds, then its sign.val · ymir
- Ymir_units.QuantityValues in a unit.module · ymir
- Nx_quant.quantize_q8_0quantize_q8_0 x is the float weight x, [| …; n; k |], stored as Q8_0, x read at float32.val · nx
- Ymir_fits.Fits.Image.quantizedquantized ~tiles q h t stores t Rice-coded in integer steps of each tile's noise divided by q, in tiles of one row unless tiles, dithered by SUBTRACTIVE_DITHER_2, which keeps zeros exact; NaN pixels are undefined.val · ymir
- Ymir_units.Unit.quectoquecto u is 10-30 u.val · ymir
- Talon.QueryQueries: descriptions of tables to compute.module · talon
- Ymir_units.Unit.quettaquetta u is 1030 u.val · ymir
- Rig_edge.queueThe type for a device's queues.type · rig
- QuickstartThis gets you from zero to computing gradients and training a model in five minutes.manual
- quillQuill is a REPL and notebook environment for OCaml.library · quill
- QuillQuill -- notebook core library.module · quill
- Quill_bookStatic site rendering for notebook projects.module · quill
- Quill_markdownMarkdown notebook format.module · quill
- Quill_projectNotebook project model.module · quill
- Quill_serverWeb notebook server.module · quill
- Quill_topOCaml toplevel kernel for Quill.module · quill
- Quill_tuiTerminal notebook interface.module · quill
- Talon.Expr.Str.quotequote s is a regular expression that matches exactly s.val · talon
r
- Hugin_gg.Color.rr c is the encoded red component of c.val · hugin
- Nx_cpu.r2cr2c ~axes ~dst x stores the Forward transform of x, bins 0 to n/2 along the last of axes, n x's extent there.val · nx
- Nx_kernel.S.r2cr2c ~axes ~dst x stores the Forward transform of x, bins 0 to n/2 along the last of axes, n x's extent there.val · nx
- Nx_kernel.Shape.r2cr2c ~axes x is x but n/2 + 1 bins along the last of axes, n x's extent there, with axes as c2c takes them.val · nx
- Vega.radam_initradam_init is adam_init: RAdam shares Adam's state.val · vega
- Vega.radam_stepradam_step p ~lr st ~params ~grads is (params', st') after one rectified Adam step (Liu et al., 2020).val · vega
- Ymir_units.Unit.radianradian is the symbol rad.val · ymir
- Talon.Column.raggedragged dt c is c's rows as a ragged array without a copy, one row per row of c: a list column's offsets and its elements, stored as dt, or a string or binary column's offsets and bytes, dt being Nx.uint8.val · talon
- Ymir_fits.Fits.Table.raggedragged ~rows dtype name hdu is a Lists column's stored numbers, or a Text column's bytes as uint8, one list per cell.val · ymir
- Rig_remote_abi.railThe type for rails.type · rig
- Rig_remote_abi.railrail ~id ~local is a provider's new rail.val · rig
- Rig_remote_proxy.Link.railrail l ~id ~send ~receive is this machine's end of the rail id to l's peer, which carries send to it and receive from it, with the rail's areas zeroed (Rails) and its ready function.val · rig
- Hugin_kit.Scheme.rampramp cs is the sequential scheme that runs through the m colours cs in order, at evenly spaced normalised values from 0 to 1, interpolating in Oklab with Hugin_gg.Color.mix between consecutive ones.val · hugin
- Nx.Rng.randintrandint ~low ~high shape is integers drawn uniformly from [low, high) (low defaults to 0): a 64-bit draw times high - low, shifted down by 64 bits, whose biased low part, at most one draw in 2{^32}, takes a second draw (Lemire's multiply-shift).val · nx
- Nx.Rng.randint_atrandint_at p ~low ~high shape is randint ~low ~high shape at p.val · nx
- Fehu_envs.Random_walkOne-dimensional random walk.module · fehu
- Hugin.Mark.rangerange rows r is Some f if r is bound to a channel that reads a scale, and None if it is unbound, bound to a constant or reads no scale.val · hugin
- Hugin_kit.Time.rangerange ~tz_offset_s i t t' is the boundaries of i at tz_offset_s in [t;t'], from the past to the future.val · hugin
- Jera.Quad.RangeRanges of integration, one per element.module · jera
- Nx_array.Move.rangeThe type for the elements start + j·step, j < count, of an axis.type · nx
- Tolk.Ops.rangerange ~axis_type ~dtype end_ axis_id is the loop variable counting from 0 to end_, excluded, of type dtype (default Dtype.t.Weak_int) and role axis_type, which has no default: a Axis_type.t.Weak range promises that its trips are independent, and a Axis_type.t.Loop runs them in turn.val · tolk
- Tolk.Ops.range_startrange_start op is the index of the first source of an op node that is a range it ends: 1 for Op.t.Stage, Op.t.Reduce and Op.t.End, 0 for Op.t.Linear, and None otherwise: a call ends none of the ranges its arguments run inside.val · tolk
- Tolk.Ops.range_strrange_str ~color r is r's identity written with underscores, a negative part written m and its magnitude: 0_1, m1.val · tolk
- Tolk.RangeifyFrom tensor graphs to kernel graphs.module · tolk
- Nx_kernel.Prog.rangesranges ~axes ?outs p is the program of p's ranges over a box, of p's outputs, or with outs of the outputs at those indices, each of an integer or boolean dtype.val · nx
- Tolk.Ops.rangesranges u is the ranges u runs inside: u itself if it is a range, then those its sources run inside, minus those it closes.val · tolk
- Norn.Diag.rankrank u ~truth d is, for each element, the number of draws of d below the element of truth: from 0 to the number of draws.val · norn
- Nx_array.Layout.rankrank l is l's number of axes.val · nx
- Nx_ragged.rankrank r is the dense rank of each row of r: the number of distinct rows that order before it.val · nx
- Talon.Expr.rankrank a is the 1-based rank of a's value among the frame's non-null values in talon's total order, ties taking the lowest rank (SQL's RANK), as int64, and null where a is null.val · talon
- Norn.Diag.rank_uniformityrank_uniformity u ~draws ranks is, for each element, the simultaneous p-value of its ranks against the uniform distribution on 0, ..., draws (Säilynoja, Bürkner and Vehtari 2022), computed exactly, with no simulation: the probability, under uniformity, that the count of ranks below some point i / (draws + 1) strays from its expectation at least as far, in pointwise p-value, as the ranks' farthest count.val · norn
- Hugin.Mark.rasterraster draws a panel's picture of the mark as one image painted by the raster renderer at the density, over the device pixels the picture reaches (Picture.bounds), when the mark has more than 20,000 rows in the panel or more rows than the panel has device pixels, whatever the output.val · hugin
- Ymir_units.Unit.ratioratio d u w is the conversion from u to w, the exact number v of u / w, rounded once to d:val · ymir
- Ymir_fits.Fits.Image.rawraw ~window dtype hdu is the stored numbers of hdu's image as dtype.val · ymir
- Ymir_fits.Fits.Table.rawraw ~rows dtype name hdu is an Array column's stored numbers, of shape [rows] @ cell, when dtype holds every value of its element: bool for L, bit for X (most significant bit first), complex64 or complex128 for C and M.val · ymir
- Hugin_kit.Scheme.rdburdbu is Brewer's RdBu, red to blue.val · hugin
- Hugin_kit.Scheme.rdgyrdgy is Brewer's RdGy, red to grey.val · hugin
- Nx.rdiv_srdiv_s c x is div c x.val · nx
- Hugin_kit.Scheme.rdpurdpu is Brewer's RdPu, red to purple.val · hugin
- Hugin_kit.Scheme.rdylburdylbu is Brewer's RdYlBu, red through yellow to blue.val · hugin
- Hugin_kit.Scheme.rdylgnrdylgn is Brewer's RdYlGn, red through yellow to green.val · hugin
- Hugin_gg.Stroke.reachreach s is how far from its path a stroke with s paints: half the width of s, multiplied by the miter limit if s miters its joins, or by sqrt 2.val · hugin
- Hugin_vg_instances.Instances.reachreach m k p is how far the strokes of p under m reach beyond their paths, their lengths multiplied by k: the largest Hugin_gg.Stroke.reach of the pens of p, mapped through the transforms within p and through m.val · hugin
- Rig_pci.Memory.reachThe type for the facts that decide whose GPU memory a GPU maps (reaches).type · rig
- Rig_pci.Memory.reachreach m is the facts that decide whose GPU memory m's GPU maps.val · rig
- Tolk.Setting.reachWhat a setting's value reaches.type · tolk
- Rig.Buffer.reachedreached b access is true iff wait b access would return without blocking, as the work it waits for is done.val · rig
- Rig.Point.reachedreached p is true iff p's device has reached p's value, as wait would find it, also once a later device took p's slot.val · rig
- Rig.reachesreaches d d' is true iff d's work can address d''s own memory (Buffer.kind.Device) through a borrow on d (Buffer.borrow).val · rig
- Rig_pci.Memory.reachesreaches r o is true iff map_peer on a GPU of reach r maps the GPU memory of another GPU of reach o, given room.val · rig
- Tolk.Ops.reachesreaches ~gate ~calls u x is true iff x is u or a node u reaches, treating calls as calls says, through nodes other than x that gate accepts (default all).val · tolk
- Munin.History.readread run is run's history up to the log's last whole line, or, with limit, up to the first whole line that ends past limit bytes from the log's start.val · munin
- Nx.Repr.readread x f is f a, a an array on the host of x's elements for the extent of f: over x's own memory where x lies on one device whose memory the host addresses (the host, a memory device, Metal), and otherwise a copy on the host, made for any placement, including a value cut across devices.val · nx
- Nx_array.readread a f is f b, where b is a bigarray of a's shape over a's own bytes; an a with no element gives an empty b, whatever its memory.val · nx
- Rig.Buffer.readread b k f is f a, where a is b's bytes as elements of k, once every device's last write of b's memory is done.val · rig
- Rig.Submission.readread slot ~at is a ref through which the work reads the buffer slot.val · rig
- Rig_bin.Line.readread r is the complete lines fd r has for now, without their newlines.val · rig
- Rig_bnxt.Qp.readread qp ~src:(s, i) ~dst:(r, at) n posts the read of the n bytes at offset i of s into the n bytes at offset at of r, on qp's Send ring, whose peer serves it.val · rig
- Rig_bnxt.Region.readread r at n is a copy of the n bytes at offset at of r.val · rig
- Rig_edge.Io.readread d r ~at ~dst ~len reads the len bytes at at in r into host memory at dst.val · rig
- Rig_mlx5_abi.Completion.readread b at is the completion at byte at of b.val · rig
- Rig_pci.Window.readread w off n is a copy of the n bytes of w from byte off.val · rig
- Tolk.Codegen.readread ast ren is the program to_program returns for ast and ren when it can do so without rendering or compiling: the program this process has made under the same key, or else one kept on disk whose binary its compiler's table keeps too, which read then keeps as to_program would.val · tolk
- Ymir.Wcs.readread ~alt ~at f kw is the map from 0-based pixel indices, in tensor axis order, to directions in f, from kw's primary description or its alternate alt ('A' to 'Z'), its parameters at at.val · ymir
- Ymir_fits.Fits.readread path is the HDUs of the file path, primary first: every header copied, every data unit checked to lie inside the file, no data read.val · ymir
- Ymir_fits.Fits.Table.readread ~rows hdu is every column in one pass over the rows, in file order, with its name and cards, in the dtype its description names: its element's, or float64 when scaled.val · ymir
- Ymir_spk.readread ~at path is the segments of the SPK file path, in file order, the data of those of types 2 and 3 placed at at.val · ymir
- Tolk.Postrange.Scheduler.read_by_boundsread_by_bounds k is the ranges that the bound of one of k's ranges reads, as i in a loop j < x[i].val · tolk
- Tolk.Postrange.Scheduler.read_by_opaqueread_by_opaque k is the ranges that a source of one of k's opaque operations reads, as i in f(&x[i]) or popcount(x[i]).val · tolk
- Rig_remote.read_keyread_key file is the key in file: its bytes, 16 to 4096 of them, read through one open of it.val · rig
- Rig_bin.Line.readerThe type for the lines of a descriptor, read without blocking.type · rig
- Rig_bin.Line.readerreader fd reads the lines of fd, which it makes non-blocking.val · rig
- Talon.Source.readerThe type for readers of a part's batches.type · talon
- Munin.Session.readingThe type for readings.type · munin
- Reading a store without muninA store is a directory of plain files, which store_format.mli specifies in full.guide · munin
- Munin.History.readingsreadings h is the keys of the run's readings, sorted.val · munin
- Tolk.Heuristic_common.readsreads r b is true iff the index of b is or computes from r.val · tolk
- Tolk.Shape.reads_state_writtenreads_state_written c n is true iff the node n, storage or a state or view of it, names a state the call c writes through another argument or a later one.val · tolk
- Tolk.Shape.reads_storage_writtenreads_storage_written c i is true iff the call c, whose body shows its writes (written_params), only reads its argument i and writes the storage it holds (Ops.storage_base, Ops.whole) through another.val · tolk
- Nx.realreal dt z is z's real parts, stored in dt by the conversion rule (cast): exact where dt is as wide as the parts.val · nx
- Tolk.RealizeThe calls of a schedule, and the compilation of their kernels.module · tolk
- Rig_bin.Proc.reapreap pid is how pid ended, if it has, without waiting.val · rig
- Nx.rearrangerearrange ~sizes p x is x split, permuted and merged as p says: the result's element at an index is x's element where each name takes the same position.val · nx
- Nx.Ptree.rebuildrebuild s ~like xs is like with its arrays replaced by xs, in walk order.val · nx
- Kaun.Metric.recallrecall ?average predictions labels is the fraction of each class's true instances that are predicted — tp / (tp + fn) — combined according to average, which defaults to `Macro.val · kaun
- Rig_bnxt.Qp.receivereceive qp r ~at n posts the receive of up to n bytes at offset at of r, which the next send of qp's peer fills.val · rig
- Rig_mlx5.Qp.receivereceive qp l posts the receive of up to l.bytes bytes into l in qp's receive ring.val · rig
- Rig_mlx5_abi.Doorbell.receivereceive is the byte offset in a queue pair's record of its receive count, a 32-bit word holding the low 16 bits of the receive ring's producer count.val · rig
- Rig_mlx5_abi.ReceiveA receive entry is one receive of a queue pair's receive ring: the memory the next message from the peer lands in.module · rig
- Nx.reciprecip x is 1 / x; on integers, x for 1 and -1 and 0 otherwise.val · nx
- Nx_kernel.Prog.E.recipval · nx
- Tolk.Ops.Elementwise.reciprocalval · tolk
- Tolk.Ops.reciprocalval · tolk
- Tolk.Ops.Upat.reciprocalval · tolk
- Rig.Profile.recordrecord e adds e to each profile being taken when it is called, so take returns it among rig's own events, in time order.val · rig
- Talon.Expr.recordrecord os is the record of the outputs os, its fields named as they are, in order: record [ "x" := x; keep (Sel.names [ "flux" ]) ].val · talon
- Talon.RecordRecord cells.module · talon
- Talon.Type.recordrecord fields is Record with fields, in order.val · talon
- Ymir_fits.Fits.Header.recordsrecords h is h's records, 80 bytes each.val · ymir
- Hugin.rectrect () draws, for each row, the rectangle that its extents along x and y span (Mark.extent).val · hugin
- Hugin_gg.Path.rectrect b is the closed subpath around b from its corner (minx b, miny b) along the x axis first, clockwise on screen.val · hugin
- Hugin_gg.Color.redred is v 1.val · hugin
- Quill.Session.redoredo s restores the next document state.val · quill
- Hugin_kit.Scheme.redsreds is Brewer's Reds.val · hugin
- Nx_cpu.reducereduce p ~shape ~axes rs ~dsts loads reduces, for each (r, k) of rs, output k of p over the loop of shape shape along axes by r.val · nx
- Nx_kernel.S.reducereduce p ~shape ~axes rs ~dsts loads reduces, for each (r, k) of rs, output k of p over the loop of shape shape along axes by r.val · nx
- Nx_kernel.Shape.reducereduce ~axes rs y is the shape of each destination of reductions rs along axes of a loop of shape y: y without axes, and y itself for no axis.val · nx
- Tolk.Op.Set.reducereduce is the operations a Reduce can reduce with: Add, Mul and Max.val · tolk
- Tolk.Ops.reducereduce u op ranges is the kernel reduction of u with op over ranges.val · tolk
- Tolk.Ops.Upat.reduceval · tolk
- Nx.reduce_rangesreduce_ranges c ~lo ~hi x is one row per range, row i the rows of x from lo.{i} to hi.{i} folded by c.val · nx
- Tolk.Postrange.Scheduler.reduce_rangesreduce_ranges k is the ranges k's reductions close, each once.val · tolk
- Nx_kernel.reduce_refusalreduce_refusal p ~rank ~reductions:rv rs ~dsts:dv dsts ~loads:lv loads is as map_refusal for S.reduce's reductions rs of p, read by rv.val · nx
- Nx.reduce_segmentsreduce_segments c ~segments ids x is segments rows, row s the rows i of x with ids.{i} = s folded by c.val · nx
- Tolk.Postrange.Scheduler.reduceopreduceop k is the first of reduceops k.val · tolk
- Tolk.Setting.reduceop_split_sizereduceop_split_size is n such that the first kernel of a split reduction outputs at most 2n elements, from REDUCEOP_SPLIT_SIZE.val · tolk
- Tolk.Setting.reduceop_split_thresholdreduceop_split_threshold is how many times as many elements as its output a reduction's input must have for scheduling to split it, under split_reduceop, from REDUCEOP_SPLIT_THRESHOLD.val · tolk
- Tolk.Postrange.Scheduler.reduceopsreduceops k is k's reductions, in Ops.backward_slice's order.val · tolk
- Hugin.Mark.reducerThe type for reducers.type · hugin
- Kaun.Loss.reductionThe type for reducing per-element (or per-example) losses to a scalar.type · kaun
- Nx_kernel.reductionThe type for what a reduction computes from its terms.type · nx
- Nx_kernel.reductionsThe type for reading a collection 'c of reductions in place, as dtypes reads arrays: reductions c reductions, the ith kind c i of the program's output output c i.type · nx
- Rig.Submission.refThe type for a launch's or a fill's references to a submit's buffers and to functions.type · rig
- Rig_program.refThe type for refs, as Rig.Submission.ref.type · rig
- Rig_amd.refusalThe type for a path's refusal of a call.type · rig
- Rig_nv.refusalThe type for the RM's refusal of a call, with a message that names the call and the answer.type · rig
- Rig_remote_proxy.Wire.refuserefuse fd why sends the refusal why, cut to 4096 bytes, in place of a greeting, then closes fd.val · rig
- Nx_array.refusedrefused name r operands raises Invalid_argument for the refusal r that a kernel answered for operands, naming name, the reason r gives, and each operand's dtype and shape.val · nx
- Nx.Rng.refusingrefusing name f runs f in a scope that gives no keys.val · nx
- Talon.Expr.Str.regexregex r matches the regular expression r anywhere in the text; see Regular expressions for the language.val · talon
- Norn.Summary.regionThe type for where divergences gather: an element whose draws from transitions that diverged have a mean shift standard deviations from its mean, by more than one (Norn.Diag.divergent_shift).type · norn
- Rig_amd.regionThe type for memory of a device.type · rig
- Rig_bnxt.RegionA region is memory the NIC writes and reads.module · rig
- Rig_cuda.regionThe type for memory of a device.type · rig
- Rig_edge.Driver.regionThe type for memory of a device.type · rig
- Rig_edge.Io.regionThe type for memory of an io device.type · rig
- Rig_metal.regionThe type for memory of a device.type · rig
- Rig_mlx5.RegionA region is memory the NIC may read or write.module · rig
- Rig_nv.regionThe type for memory of a device.type · rig
- Rig_pci.Memory.regionThe type for ranges of memory a GPU addresses: what alloc, map_host or map_peer gave, until free.type · rig
- Rig_remote_proxy.regionThe type for memory of a device.type · rig
- Ymir.RegionPlanar shapes placed on a grid's world.module · ymir
- Rig_edge.Io.region_keyregion_key identifies the io library and its regions.val · rig
- Rig_mlx5.Region.registerregister nic a o m registers m with nic for access a and write order o.val · rig
- Rig_mlx5_abi.Uar.registerregister ~per_page ~size r is the system page and the byte offset in it of doorbell register r, with per_page access regions per system page and registers of size bytes.val · rig
- Tolk.Compiler_cpu.registersregisters arch is the vector register file of the processor clang arch compiles for, as the macros of the Clang run Setting.cc state its features: on x86_64, 16 registers, 32 with AVX-512, of 32 bytes with AVX and 16 without (a processor with AVX-512 is taken at 32 bytes); on arm64, 32 of 16; on riscv64, None.val · tolk
- Tolk.Renderer.registersThe type for a target's vector register file.type · tolk
- Hugin_font.Font.regularregular is Inter Regular, weight 400.val · hugin
- Jera.Tol.relrel r is v ~rel:r ~abs:0..val · jera
- Tolk.Tc.relabelrelabel tc is, for A and for B, each bit of its fragment paired with the bit it takes the place of when its elements are loaded as C's: its lane bits pair with C's lane bits in order, its element bits with the first of base_upcast_axes, from the last.val · tolk
- Rig.Claim.releaserelease h ends h's claims now, so a with_ that starts after it returns finds none of them.val · rig
- Rig.Submission.Run.releaserelease r frees r, which the calling thread has taken, keeping its bytes, so that a take may return it.val · rig
- Rig_nv_abi.Method.releaserelease addr v waits for the channel's earlier work to complete, every engine idle, then writes the 64-bit v into the semaphore at addr, so that any reader who sees v, the host and other devices included, sees the work's writes.val · rig
- Rig_nv_abi.Qmd.releaseThe type for what a launch writes when it completes: the 64-bit value at an address, 8-byte aligned, so that any reader who sees the value, the host and other devices included, sees the launch's writes.type · rig
- Rig_pci.Function.releaserelease f gives f back.val · rig
- Rig_pci.Memory.releaserelease m releases the function of m's GPU (Function.release), then drops the page tables create was given, which only a live GPU uses, and ends the GPU's mappings of other GPUs' memory, as their frees would, so that those GPUs' functions go back once released.val · rig
- Rig_pci.Space.releaserelease h gives back every hold of h, as free does.val · rig
- Rig_remote_proxy.Link.release_railrelease_rail l id ends l's rail id here: once it returns, neither thread reads or writes its end, whose memory lives while it is reachable, and its ready function must not be called again.val · rig
- Rig_nv_abi.Method.release_stamprelease_stamp addr v is release that also writes a stamp at addr + 8 (Semaphores).val · rig
- Rig_pci.Function.releasedreleased f is true iff a release of f has begun.val · rig
- Quill.Session.reloadreload doc s replaces the document, clearing history and statuses.val · quill
- Rig_elf.relocationThe type for relocations.type · rig
- Kaun.Fn.relurelu x is zero where x <= 0 and x elsewhere: -0.val · kaun
- Nx_kernel.Prog.E.remval · nx
- Tolk.Bigint.remrem a b is a - b * div a b: it has a's sign and is smaller than b in magnitude.val · tolk
- Rune.rematremat s f is f, s its signature as for jit, whose reverse mode keeps f's arguments and the values it captures, and computes f's intermediate values again in the backward pass: it trades computation for memory.val · rune
- Rig_mlx5.Region.remoteremote r at is offset at of r, as a peer's entries name it.val · rig
- Rig_mlx5_abi.Entry.remoteThe type for memory a peer registered with its NIC.type · rig
- Fehu.Info.removeremove k t is t without the binding for k.val · fehu
- Quill.Doc.removeremove id d removes the cell with identifier id from d.val · quill
- Ymir_fits.Fits.Header.removeremove k h is h without k's cards and their CONTINUE records.val · ymir
- Munin.Store.remove_aliasremove_alias s ~name alias removes name's alias alias, if it has one.val · munin
- Quill.Session.remove_cellremove_cell id s removes the cell with identifier id.val · quill
- Munin.Store.remove_tagsremove_tags s id tags removes tags from the run id, ignoring those it lacks.val · munin
- Tolk.Postrange.Scheduler.renren k is the renderer k optimises for.val · tolk
- Talon.Kit.renamerename pairs q is q with each column old of pairs renamed to the name it pairs with, in place:val · talon
- Fehu.Env.renderrender env produces a visualization of the current state.val · fehu
- Fehu.RenderVisualization primitives.module · fehu
- Fehu_envs.Cartpole.rendertype · fehu
- Fehu_envs.Grid_world.rendertype · fehu
- Fehu_envs.Mountain_car.rendertype · fehu
- Fehu_envs.Random_walk.rendertype · fehu
- Hugin_vg_pdf.renderrender r is the PDF document of r.val · hugin
- Hugin_vg_raster.renderrender ~density r is r drawn as a [|h; w; 4|] tensor of RGBA pixels with straight alpha, on the host, where w and h are the width and height of the page of r multiplied by density and rounded to the nearest integer.val · hugin
- Hugin_vg_svg.renderrender r is the SVG document of r, encoded in UTF-8.val · hugin
- Quill_book.RenderHTML rendering for project notebooks.module · quill
- Tolk.RenderNodes as text: expressions and program listings.module · tolk
- Tolk.Render.renderrender ~simplify u is u written as an expression, after Shape.simplify if simplify (default true).val · tolk
- Fehu.Env.render_modeRendering modes supported by environments.type · fehu
- Fehu.Env.render_moderender_mode env is the render mode chosen at construction, if any.val · fehu
- Fehu.Env.render_mode_to_stringrender_mode_to_string m is the string representation of m.val · fehu
- Tolk.Render.render_uirrender_uir uops writes the nodes uops, in order, one line each, as %i = op srcs : arg: its position among the lines, from 0; its operation in lower case, in its colour (Helpers.colored); its sources, separated by commas; and its argument after " : ", if it writes one.val · tolk
- Hugin.Drawing.renderablerenderable d is the picture of d on a page of the figure's size: the paper, then the panels' marks, the guides and the titles.val · hugin
- Hugin_vg.RenderableRenderables.module · hugin
- Tolk.Device.rendererrenderer t is the renderer of programs for t: the renderer of t's device, made for t.val · tolk
- Tolk.RendererRenderers: what code generation may emit for a target, and how it becomes a program.module · tolk
- Tolk_engine.rendererrenderer d is the renderer of d's target, which writes the source of its programs, made once per target (Tolk.Device.renderer).val · tolk
- Rendering and Extendingsave hides three stages.guide · hugin
- Rig_pci.Driver.renewrenew h resets the function h holds as its vendor does (reset of make), with its bus mastering off, and gives back the memory dead processes left for it.val · rig
- Ymir.Healpix.renumberrenumber h s p is the index in scheme s of each pixel whose index in h's scheme is in p, -1 for -1.val · ymir
- Norn_model.reparamreparam sel b m is m whose site at sel has the coordinates its value has under b d, d being the site's distribution on each run.val · norn
- Nx.repeatrepeat ~axis n x is x with each element repeated n times in place along axis: repeat ~axis:0 2 of [a; b] is [a; a; b; b].val · nx
- Tolk.Shape.repeatrepeat u repeats tiles u repeats times along each axis, the axes aligned to the right: a repeats longer than u's shape adds leading axes.val · tolk
- Brot.Decoder.replacereplace ~pattern ~by () replaces every literal occurrence of pattern with by in each token.val · brot
- Brot.Normalizer.replacereplace ~pattern ~replacement is a normalizer that replaces every occurrence of the string pattern with replacement, scanning left to right without overlap.val · brot
- Quill.Doc.replacereplace id c d replaces the cell identified by id with c.val · quill
- Talon.Expr.Str.replacereplace p ~by a is a with each match of p, found as split finds them, replaced by by: replace (literal "aa") ~by:"b" of "aaa" is "ba", and a row with no match is unchanged.val · talon
- Tolk.Ops.replacereplace u is u with the given fields replaced; u itself if none changes.val · tolk
- Quill_book.Render.replace_allreplace_all ~pattern ~with_ s replaces all occurrences of pattern in s with with_.val · quill
- Brot.Normalizer.replace_regexreplace_regex ~pattern ~replacement is a normalizer that replaces every match of the regular expression pattern with replacement.val · brot
- Nx.Ptree.reportThe type for what a walk reports.type · nx
- Nx.Reprmodule · nx
- Rig_remote_proxy.Link.requestrequest l r sends r after every message sent before it and is the agent's answer.val · rig
- Rig_remote_proxy.Wire.requestThe type for requests, each answered by a value of its parameter.type · rig
- Talon.Source.requestThe type for what talon asks of a source when it runs.type · talon
- Norn.Weighted.resampleresample u k ~n w is n equal-weight draws of w, by systematic resampling: one uniform v places the points (i + v) / n on the cumulative normalised weights, so a draw of normalised weight w_i is copied floor (n w_i) or ceil (n w_i) times.val · norn
- Rig_pci.Machine.reservereserve m ~base n reserves the n addresses from base on, in the address space of the process that holds m's functions, so that only Function.alloc_dma maps memory there.val · rig
- Fehu.Env.resetreset env () resets the environment to an initial state.val · fehu
- Fehu.Vec_env.resetreset t () resets all environments.val · fehu
- Rig_amd_pci.resetreset ~machine i stops whatever runs on GPU i of machine (defaults to Rig_pci.Machine.this) and, if its security processor's OS runs, resets the GPU whole (mode 1), which clears what any driver left.val · rig
- Rig_nv_pci.resetreset ~machine i resets GPU i of machine (defaults to Rig_pci.Machine.this), which ends what an earlier boot left running: it takes the GPU's function, turns its bus mastering off and resets the function (Rig_pci.Function.reset).val · rig
- Rig_pci.Driver.resetreset d m i takes function i of m, turns its bus mastering off, resets it as its vendor does (make), asks ready of make whether it is ready, and releases it, whatever the driver's reset or ready returns or raises.val · rig
- Rig_pci.Function.resetreset f resets f with the reset Linux has for it, which clears the state a previous driver left, and waits at most a second for f to answer again.val · rig
- Nx.reshapereshape s x is x's elements in C order at shape s.val · nx
- Tolk.Shape.reshapereshape u shape is u with shape shape, or u if unchanged; a size of -1 is inferred.val · tolk
- Sowilo.resizeresize ~height ~width img resizes to target dimensions.val · sowilo
- Hugin_kit.Time.resolutionThe type for resolutions, the units of int64 counts of instants.type · hugin
- Hugin.resolveresolve ~prev ~view f is f resolved under view, View.empty by default: its binds evaluated, its ids assigned, its scopes formed and its composition checked, then every scale's specification merged and fitted to summaries of its marks' data computed where their tensors live, categorical scales first.val · hugin
- Rig.Image.resolveresolve i f is i's function f, made ready to launch, or None if i has none.val · rig
- Tolk.Shape.resolveresolve ~default u is the value of the boolean node u if its simplification (simplify) has one possible value, and default (default true) otherwise.val · tolk
- Tolk.Shape.Sint.resolveresolve ~default c is c's value if known, and Shape.resolve's otherwise.val · tolk
- Tolk.Ops.resolve_returned_afterresolve_returned_after r effects is the value stored into the output r by the one store of effects into r's storage: r after that store if r is a parameter of the enclosing function.val · tolk
- Hugin.ResolvedResolved figures.module · hugin
- Munin.History.restartedrestarted h is true iff the follow that made h read its log again from its start.val · munin
- Ymir.Observation.restrictrestrict m o keeps a sample valid where it was valid and m holds.val · ymir
- Munin.History.resumeThe type for resumes.type · munin
- Munin.Session.resumeresume run ~step is a new session of run that continues from step.val · munin
- Munin.History.resumesresumes h is the run's resumes, in log order.val · munin
- Fehu.Gae.returnsreturns ~rewards ~terminated ~truncated ~gamma computes discounted cumulative returns.val · fehu
- Nx.Ptree.returnsreturns a is a result of a.val · nx
- Hugin_gg_kit.Ring2.reversereverse r is r with its points in reverse order: point i of reverse r is point length r - 1 - i of r.val · hugin
- Hugin_kit.Scheme.reversereverse s is s with its table in reverse order.val · hugin
- Jera.Sde.reversible_heunreversible_heun is the reversible Heun method (Kidger, Foster, Li and Lyons, 2021), Stratonovich: strong order 1/2, and 1 for additive noise.val · jera
- Tolk.Ops.Pattern_matcher.rewriterewrite m ctx u is the result of the first rule of m that matches u and does not decline, trying the namings of each pattern in Upat.match_'s order.val · tolk
- Tolk.Setting.rewrite_stack_limitrewrite_stack_limit bounds the work list of Ops.graph_rewrite, from REWRITE_STACK_LIMIT.val · tolk
- Nx.rfftrfft dt x is the forward transform of the real x along axes, the last of them keeping bins 0 to n/2 of its n points, the others being redundant: one bin for n = 0.val · nx
- Nx.rfftfreqrfftfreq dt ~d n is the host value of the frequency of each of rfft's n / 2 + 1 bins for n samples d apart: i / (d n) for i in [0, n / 2], computed as fftfreq's.val · nx
- Nx.rfftfreq_atrfftfreq_at p dt ~d n is rfftfreq dt ~d n at p, an _at form as Creation states.val · nx
- Sowilo.rgb_to_hsvrgb_to_hsv img converts RGB [0, 1] to HSV.val · sowilo
- Norn.Diag.rhatrhat u d is the rank-normalised split R-hat of each element: the larger of the R-hat of the split chains' normal scores (bulk) and of the normal scores of their distance to the median (tails).val · norn
- rigrig runs compiled work on the devices of a machine, such as its GPUs, from OCaml.library · rig
- RigDevices such as GPUs, and the order of the work they run.module · rig
- Rig_bin.Help.rigrig is rig's page.val · rig
- Rig_amdThe driver of AMD GPUs.module · rig
- Rig_amd_abiThe bytes an AMD GPU shares with its driver and with compiled code, as values.module · rig
- Rig_amd_amdgpuThe path that opens AMD GPUs through Linux's amdgpu kernel driver.module · rig
- Rig_amd_pciThe path that drives AMD GPUs over PCI, with no kernel driver.module · rig
- Rig_binmodule · rig
- Rig_bnxtBroadcom NetXtreme RoCE NICs, driven over PCI with no kernel driver.module · rig
- Rig_cudaNVIDIA GPUs, driven through the CUDA library.module · rig
- Rig_cuda_abiWhat compiled code needs from a CUDA device.module · rig
- Rig_diskFiles as an io device.module · rig
- Rig_edgeThe contract between rig and the drivers that run its devices.module · rig
- Rig_edge_primitivesThe OCaml primitives of rig_edge.h: the host's clock, rig's array of timeline words, the count of the locks a thread holds and its signal mask, which rig, its drivers and their C share.module · rig
- Rig_elfELF objects laid out in one image.module · rig
- Rig_gpusThe GPUs of this machine.module · rig
- Rig_metalThe Mac's GPU, driven through Apple's Metal framework.module · rig
- Rig_metal_abiWhat compiled code needs from a Metal device.module · rig
- Rig_mlx5ConnectX NICs of this machine, opened through Linux's RDMA subsystem.module · rig
- Rig_mlx5_abiConnectX NICs' formats, as values.module · rig
- Rig_nvThe driver of NVIDIA GPUs.module · rig
- Rig_nv_abiNVIDIA GPUs' formats, as values.module · rig
- Rig_nv_nvidiaNVIDIA GPUs of this machine, opened through NVIDIA's kernel driver.module · rig
- Rig_nv_pciNVIDIA GPUs driven over PCI, with no kernel driver.module · rig
- Rig_pciGPUs driven through their PCI function.module · rig
- Rig_poolmodule · rig
- Rig_programPrograms of steps on one machine's devices, as data that loads once and runs many times.module · rig
- Rig_remoteJobs, in which one process uses the devices of other machines through an agent on each.module · rig
- Rig_remote_abiWhat compiled code needs from another machine's devices.module · rig
- Rig_remote_proxyProxies: devices of another machine, driven through its agent.module · rig
- Rig_mlx5.Qp.ringring qp hands the NIC the entries posted since the last ring.val · rig
- Rig_mlx5_abi.ringThe type for a queue pair's rings.type · rig
- Hugin_gg_kit.Ring2Linear rings.module · hugin
- Hugin_gg_kit.Pgon2.ringsrings p is the rings of p, in order.val · hugin
- Tolk.Transcendental.rintkrintk d is the float d rounded to the nearest integer, halves away from zero, as the signed integer of d's width: Dtype.t.Int16, Dtype.t.Int32 or Dtype.t.Int64.val · tolk
- Jera.Ode.rk4rk4 is the classical Runge–Kutta method: order 4, four evaluations per step.val · jera
- Rig_nv.rmThe type for a GPU's resource manager, as a path reaches it: its allocations, frees, and each command the driver sends, which the path answers as NVIDIA's RM does.type · rig
- Kaun.Rms_normRoot mean square normalization over the feature axis.module · kaun
- Nx.rms_normrms_norm ~axes ~eps x is x / sqrt (mean x² + eps), the mean along axes, [-1] by default; eps defaults to 1e-6.val · nx
- Vega.rmsprop_initrmsprop_init p params is the initial state for optimizing params: all-zero averages and velocity, and step = 0.val · vega
- Vega.rmsprop_ptreermsprop_ptree p is Nx.Ptree.nest (module Rmsprop_state) p.val · vega
- Vega.rmsprop_stateThe state for rmsprop_step: the moving average of squared gradients and the momentum velocity, both with the shape of the parameters, and the number of completed steps as a scalar tensor.type · vega
- Vega.Rmsprop_stateThe structure of RMSprop states.module · vega
- Vega.rmsprop_steprmsprop_step p ~lr st ~params ~grads is (params', st') after one RMSprop step (Tieleman and Hinton, 2012).val · vega
- Nx.RngRandom numbers from one Threefry-2x32 generator.module · nx
- Tolk.Postrange.Scheduler.rngsrngs k is k's axes.val · tolk
- Brot.Post_processor.robertaroberta ~sep ~cls () is a RoBERTa-style post-processor.val · brot
- Tolk.Setting.rocm_pathrocm_path is the directory of the ROCm installation whose comgr library Compiler_amd.hip loads, from ROCM_PATH.val · tolk
- Hugin.RoleRoles.module · hugin
- Nx.rollroll ~axis k x is x with its elements along axis shifted k places toward the end, those past it wrapping to the start; a negative k shifts toward the start.val · nx
- Fehu.Collect.rolloutrollout env ~policy ~n_steps collects n_steps transitions.val · fehu
- Fehu.Render.rolloutrollout env ~policy ~steps ~sink () runs policy in env for up to steps steps.val · fehu
- Ymir_units.Unit.ronnaronna u is 1027 u.val · ymir
- Ymir_units.Unit.rontoronto u is 10-27 u.val · ymir
- Rig_bnxt.Qp.roomroom qp side is how many more entries side's ring of qp takes.val · rig
- Rig_mlx5.Qp.roomroom qp r is how many entries or receives qp takes before its ring r is full.val · rig
- Jera.RootZeros of functions of one variable.module · jera
- Munin.Store.rootroot s is the absolute path of s's directory.val · munin
- Nx.Ptree.Path.rootroot is the path of no step.val · nx
- Rig_pci.Page_table.rootroot t is the physical address of the root table.val · rig
- Rune.rootroot x ~residual solve is solve (), a value of structure x stated to be a zero of residual.val · rune
- Ymir_units.Quantity.rootroot n q is the real n-th root of q's payload in Unit.root n (unit q), by Nx.pow at the exponent 1/n, then, for n >= 3, one Newton step that corrects the rounding of 1/n.val · ymir
- Ymir_units.Unit.rootroot n u is u to the power 1/n.val · ymir
- Tolk.Shape.roprop u op axes reduces the axes axes of u with op: axes of size 1 are reshaped away, and the others are permuted to the front and reduced by one Op.t.Reduce.val · tolk
- Kaun.RopeRotary position embeddings: a frequency schedule and its rotation.module · kaun
- Hugin_gg.Affine.rotaterotate a turns points by a radians about the origin, from the positive x axis towards the positive y axis.val · hugin
- Ymir.Direction.rotaterotate g d is d in frame g: Frame.matrix (frame d) g times each vector.val · ymir
- Sowilo.rotate90rotate90 img rotates by k * 90 degrees counter-clockwise.val · sowilo
- Ymir.Transform.rotationrotation f g maps directions in f to the same directions in g, as Direction.rotate does.val · ymir
- Nx.roundround x rounds to the nearest integer, halves away from zero.val · nx
- Nx_kernel.Prog.E.roundval · nx
- Tolk.Helpers.round_upround_up num amt is the smallest multiple of amt greater than or equal to num, for positive amt.val · tolk
- Hugin_vg_vector.Vector.roundedrounded d v is v rounded to d decimals, as add_fixed writes it.val · hugin
- Rig.Buffer.rowrow ~src_offset ~dst_offset n is the box of one row of n bytes, at src_offset and dst_offset (default to 0), with pitches of n.val · rig
- Talon.Expr.rowThe shape of expressions with one value per row of their frame.type · talon
- Hugin.Mark.rowsThe type for the rows of a mark in one panel.type · hugin
- Hugin_vg.Picture.rowsThe type for the data a tagged picture draws.type · hugin
- Kaun.Cache_index.rowsrows ~context lens gives sequence b its own run of context slots, b * context to b * context + context - 1, and places its lens.(b) tokens at positions 0 to lens.(b) - 1.val · kaun
- Talon.Expr.rowsrows is the number of rows of the frame, as int64.val · talon
- Talon.rowsrows t is the number of rows of t.val · talon
- Ymir_fits.Fits.Table.rowsrows t is NAXIS2.val · ymir
- Nx.rpow_srpow_s c x is pow c x.val · nx
- Nx.rshiftrshift x n is x shifted n bits toward its low end: x / 2{^n} rounded toward negative infinity, so a signed x keeps its sign; 0, or -1 for a negative x, once n reaches the width.val · nx
- Nx.rsqrtrsqrt x is recip (sqrt x): inf at +0., -inf at -0., NaN below.val · nx
- Nx.rsub_srsub_s c x is sub c x.val · nx
- Hugin.rulerule () draws a straight segment for each row:val · hugin
- Hugin_vg.Picture.ruleThe type for fill rules, which say what area a path bounds.type · hugin
- Jera.Quad.RuleQuadrature rules on [−1, 1].module · jera
- Tolk.Ops.Pattern_matcher.ruleThe type for rules.type · tolk
- Tolk.Ops.Pattern_matcher.rulerule p f is f m for a node p matches, where m name is the node p names name.val · tolk
- Tolk.Shape.rulerule p f is Ops.Pattern_matcher.rule p f, which declines where f raises Not_a_number.val · tolk
- Tolk.Ops.Pattern_matcher.rule_ctxrule_ctx p f is rule, with the context passed to f.val · tolk
- Tolk.Shape.rule_ctxrule_ctx p f is Ops.Pattern_matcher.rule_ctx p f.val · tolk
- Munin.Store_format.rulesrules is the version of the statistics rules this library computes and reads in a statistics hint (Statistics of an ended run), 2.val · munin
- Tolk.Ops.rulesThe type for the rules of a rewrite and when they meet a node.type · tolk
- Fehu.Eval.runrun env ~policy () evaluates policy over n_episodes (default 10) episodes of at most max_steps (default 1000) steps each.val · fehu
- Hugin_font.RunGlyph runs.module · hugin
- Munin.Artifact.runrun a is the id of the run that produced a, which stays even after that run is deleted.val · munin
- Munin.History.runrun h is the run as h last read it: its tags, notes and status, and its end if h read one.val · munin
- Munin.RunRuns, as a listing reads them.module · munin
- Munin.Store.runrun s id is the run id of s, read as runs reads it, or None if s has no run id.val · munin
- Munin_tui.runrun store screen shows the dashboard of store, starting on screen, on the process's terminal until the user quits.val · munin
- Norn.Nested.runrun u ?batch ?tolerance ~budget ~prior ~likelihood k live steps from init until the live points' largest possible contribution changes ln Z by less than tolerance, or for budget steps, and is the evidence then.val · norn
- Norn.Smc.runrun u ?move ?resampled ~budget ~prior ~likelihood k start steps from init until β = 1, or for budget temperatures, and is the evidence then.val · norn
- Quill.Eval.runrun ~create_kernel doc creates a kernel, runs each code cell of doc in order with Kernel.execute, adds their outputs with Doc.add_output, shuts down the kernel, and returns the updated document.val · quill
- Quill_tui.runrun ~create_kernel path launches the notebook TUI for the file at path.val · quill
- Rig.Submission.RunRuns.module · rig
- Rig_bin.Help.runrun is rig run's page.val · rig
- Rig_bin.Runrig run: a job's launcher.module · rig
- Rig_bin.Run.runrun ~misuse ~rig ~firmware machines prog args runs prog with args as the controller of a job on machines, as written in --on: two or more, each a valid machine, named once.val · rig
- Rig_program.runrun ~waits p f runs p's steps once, in order, over the frame f (Order).val · rig
- Talon.Query.runrun q is q's rows as one batch: fold into a table, then one copy into a single batch.val · talon
- Tolk.Heuristic_common.runrun o r is the run of o along the range r: d if o reads r only as r / d, 1 otherwise.val · tolk
- Tolk_engine.runrun ~vars s slots runs s with the buffer of slots.(i) bound to its parameter i (an untagged Tolk.Op.t.Param), on the parameter's device, and each variable of the schedule bound to its value in vars (default []), or else to the value its call binds it to (Tolk.Realize.get_call_kernel_args).val · tolk
- Hugin_vg_vector.Vector.run_meetsrun_meets cut m r is true iff the ink of r mapped through m, or its origin if it has none, is finite and meets cut.val · hugin
- Tolk.Indexing.run_rangeifyrun_rangeify sink is the tensor graph sink with its elements indexed by ranges.val · tolk
- Rig_program.run_receivedrun_received ~waits w ~program ~region is run ~waits of the program that program gives for the value w starts with, over the frame w holds.val · rig
- runerune computes derivatives, maps, loops and compiled versions of OCaml functions whose arguments and results hold nx arrays.library · rune
- RuneDerivatives, maps, loops and compilation of functions over nx values.module · rune
- Munin.Store.runsruns s is the runs of s, newest first by start time, then by id.val · munin
- Rig_amd_abi.Code_object.runs_onruns_on co g is true iff g runs co: co is compiled for Gpu.processor g, or for a generic processor that LLVM lists it under.val · rig
- Rig.runs_on_hostruns_on_host d is true iff d's work is this process's code, which holds for host and a memory_device.val · rig
- Ymir_units.Codata.rydbergrydberg r is R∞, the Rydberg constant, in m-1.val · ymir
s
- Nx.Ptree.SStructures.module type · nx
- Nx_kernel.SThe type for kernel libraries.module type · nx
- Talon.Time.Span.ss n is like us in seconds.val · talon
- Nx.Placement.samesame p q is Some Equal iff p and q are of one set, whatever they cut.val · nx
- Nx_kernel.same_precisionsame_precision x y is true iff x and y have one precision, each float32 or complex64, or each float64 or complex128.val · nx
- Fehu.Buffer.samplesample buf ~batch_size draws batch_size transitions uniformly at random (with replacement).val · fehu
- Fehu.Space.samplesample s is a uniformly sampled value from s.val · fehu
- Jera.Ode.samplesample y m ~tol ~budget f ~at y0 is the state at each time of at, stacked on a new leading axis of each leaf, y0 first at at.(0), with budget attempts across all of them.val · jera
- Munin.Metric.sampleThe type for samples: the step the program gave, the time the sample was taken, and its value, read back exactly (Numbers and values).type · munin
- Munin.Metric.samplesample s i is s's ith sample.val · munin
- Munin.Session.samplesample s ~every readings calls each reading now, in the caller's thread, then every every seconds on a thread of its own until s finishes.val · munin
- Norn.Dist.samplesample k d is a draw of d from the key k.val · norn
- Norn.Ensemble.samplesample u lp k ~draws s is the state after draws transitions, the walkers' positions and their statistics, each walker a chain of the draws.val · norn
- Norn.Evidence.samplesample z is draws weighted by the posterior, their log weights normalised.val · norn
- Norn.Gaussian.samplesample u k ~n g is n draws of g on a new leading axis.val · norn
- Norn.Hmc.samplesample u lp k ~draws s is the state after draws transitions, their positions and their statistics.val · norn
- Norn.Nuts.samplesample u lp k ~draws s is the state after draws transitions, their positions and their statistics.val · norn
- Norn_model.samplesample d is a value of d at a new site.val · norn
- Talon.Kit.samplesample ~seed size q draws rows of q without replacement.val · talon
- Fehu.Buffer.sample_arrayssample_arrays buf ~batch_size is like sample but returns structure-of-arrays (observations, actions, rewards, next_observations, terminated, truncated) for direct use in training loops.val · fehu
- Hugin.savesave ~warn ~view ~theme ~size ~density file f writes the renderable of draw ~density (layout ~theme size (resolve ~view f)) to file in the format its extension names, ignoring case: PNG for .png, drawn at density with the density and the sRGB colour space recorded (Hugin_vg_raster.png); SVG for .svg; PDF for .pdf.val · hugin
- Quill_markdown.savesave figures path doc writes doc with its outputs to the file path names, through a link too, storing its images as figures says, and is the stamp of the notebook it wrote.val · quill
- Nx_io.save_jpegsave_jpeg path x writes the image x to path as a baseline JPEG file at quality 90, the standard tables of T.81's Annex K scaled to 20%, with a colour image's chroma subsampled 4:2:0.val · nx
- Brot.save_model_filessave_model_files t ~folder saves t's underlying model files (vocabulary and merges) to folder and returns the list of created file paths.val · brot
- Nx_io.save_npysave_npy path x writes x to the npy file path.val · nx
- Nx_io.save_npzsave_npz path a writes a's entries to the npz file path.val · nx
- Nx_io.save_pngsave_png path x writes the image x to path as a PNG file of 8 bits per sample, the bytes encode_png ?dpi ?srgb x.val · nx
- Brot.save_pretrainedsave_pretrained t ~path saves t to path in HuggingFace format.val · brot
- Nx_io.save_safetensorssave_safetensors path a writes a's entries to the SafeTensors file path, in the order of the format's reference implementation.val · nx
- Nx_io.save_txtsave_txt path x writes the 0-d value, vector or matrix x to the text file path, with no header or comment line.val · nx
- Saving and Pretrained WeightsA trained model is a set of tensors.guide · kaun
- Tolk.Setting.scachescache is where schedules are cached, from SCACHE: 0 nowhere, 1 in memory, 2 or more also in the Helpers.Diskcache, where a compiled schedule is kept whole too.val · tolk
- Nx.scalarscalar dt v is the 0-d host value v.val · nx
- Nx.scalar_atscalar_at p dt v is scalar dt v at Placement.uncut p, whole on each of p's devices however p cuts, since a 0-d value has no axis to cut.val · nx
- Nx.scalar_likescalar_like x v is scalar_at (beside x) (dtype x) v.val · nx
- Hugin.Mark.scalescale rows `X k is Some s, the fitted scale (Resolved.scale) that the channels on x of the panel read, if it has kind k, whether or not the mark binds x, and None if they read none or one of another kind, as in a swatch.val · hugin
- Hugin.Resolved.scalescale ~at r s is the scale named like s, of the kind of s, in the scope that holds the node at (Scopes), the root by default, fitted: its domain set and its nice and zero unset (Scale.fit), and its other properties those its channels merged, implied ones included.val · hugin
- Hugin_gg.Affine.scalescale sx sy scales x by sx and y by sy about the origin.val · hugin
- Hugin_kit.ScaleScales.module · hugin
- Hugin_text.Text.scalescale s t is t with every size factor and shift multiplied by s.val · hugin
- Nx.Ptree.scalescale s k x is x with each array a of the vector replaced by k' * a, k' k cast to a's format, part by part for a complex a.val · nx
- Nx_kernel.scaleThe type for a block scale of a contraction operand's array: an array of its rank, each extent dividing the operand's, 0 only along an empty axis.type · nx
- Vega.Loss_scale.scalescale ls x is x times the current scale, at x's dtype.val · vega
- Ymir.Observation.scalescale k o is o's data times k, in their units' product, with the variance times k².val · ymir
- Ymir.Time.scaleThe type for time scales.type · ymir
- Ymir.Time.scalescale t is t's scale.val · ymir
- Ymir.Transform.scalescale d maps x, in Unit.one, to x.(k) · d.(k) in component k, in d's unit: CDELT.val · ymir
- Hugin_vg_vector.Vector.scale_decimalsscale_decimals ctx s is the decimals of the pen an instance of scale s sets in the frame of ctx.val · hugin
- Hugin_vg_vector.Vector.scale_penscale_pen k pen is pen with its lengths multiplied by k.val · hugin
- Ymir_fits.Fits.Image.scaledscaled t is true iff BSCALE or BZERO scale the stored numbers beyond the offsets.val · ymir
- Kaun.Attention.scaled_dot_product_attentionscaled_dot_product_attention q k v is softmax (q @ kᵀ * scale) @ v: each of the n query rows takes a weighted average of the m value rows, weighted by the softmax of its scaled dot products with the key rows.val · kaun
- Scales, Colour and GuidesA scale maps data values to the interval [0, 1], and a role maps that interval to what it draws: a position across the panel, a colour, an area.guide · hugin
- Nx_cpu.scanscan ?carry p ~axis (r, k) ~dsts loads stores at each index the reduction r of output k of p over the indices along axis up to that index's, inclusive (Scans).val · nx
- Nx_kernel.S.scanscan ?carry p ~axis (r, k) ~dsts loads stores at each index the reduction r of output k of p over the indices along axis up to that index's, inclusive (Scans).val · nx
- Nx_kernel.Shape.scanscan ~axis y is the shape of a scan's destinations along axis of a loop of shape y: y.val · nx
- Rune.scanscan c x y f ~init xs folds the step f over the rows of xs, a value of structure x whose arrays share a leading length n: step i passes f row i of every array.val · rune
- Rune.scan'scan' f ~init xs is scan of one array each for the carry, the rows and the outputs.val · rune
- Nx_kernel.scan_refusalscan_refusal p ~rank r ~output:k ~dsts:dv dsts ?carry ~loads:lv loads is as map_refusal for S.scan's scan r of output k of p from a carry of dtype carry.val · nx
- Nx.scatterscatter ~combine ~axis p u x is x with each element of u combined into x at its own index with axis axis replaced by p's element there; p and u have one shape.val · nx
- Nx_cpu.scatterscatter c ~unique ~axis ~dst ~into idx updates stores into dst into with updates combined by c at the positions idx holds.val · nx
- Nx_kernel.S.scatterscatter c ~unique ~axis ~dst ~into idx updates stores into dst into with updates combined by c at the positions idx holds.val · nx
- Nx_kernel.Shape.scatterscatter ~axis into idx updates is the shape of a scatter into into: into.val · nx
- Sowilo.scharrscharr img computes Scharr gradients, which are more rotationally accurate than Sobel.val · sowilo
- Rig_nv_abi.Method.scheduleschedule addr schedules the launch descriptor at addr, 256-byte aligned and below 2{^40}, and the descriptors chained to it (Qmd.t.chain).val · rig
- Tolk.ScheduleSchedules: from tensor graphs to the calls that realize them.module · tolk
- Vega.ScheduleLearning-rate schedules.module · vega
- Tolk.Postrange.SchedulerSchedulers.module · tolk
- Talon.Query.schemaschema q is the names and types of the columns of q's rows.val · talon
- Talon.schemaschema t is the names and types of t's columns.val · talon
- Talon.SchemaSchemas: the names and types of a table's columns.module · talon
- Hugin.Theme.schemescheme th is the scheme of continuous scales of th.val · hugin
- Hugin_kit.Scale.schemescheme s is the colour scheme of s, if set.val · hugin
- Hugin_kit.SchemeColour schemes.module · hugin
- Ymir.Healpix.schemeThe type for pixel numberings.type · ymir
- Ymir.Healpix.schemescheme h is h's numbering.val · ymir
- Ymir_units.Unit.scopedscoped ~scope name is the symbol name of the data set scope.val · ymir
- Munin_tui.screenThe type for the screen a dashboard starts on.type · munin
- Tolk_engine.scrubThe type for what empties the caches of a schedule's devices before a time.type · tolk
- Tolk_engine.scrubscrub s is, for each device that runs a kernel of s and whose last-level cache's bytes its driver states (Rig.cache_bytes), a buffer of twice those bytes in the device's memory, written once, and a kernel compiled for the device that reads it whole.val · tolk
- Jera.SdeStochastic differential equations.module · jera
- Tolk.SearchSearching kernel optimisations by measuring them.module · tolk
- Nx.searchsortedsearchsorted ~side s v is, for each key of v, the number of keys of s before it (`Left), or before or equal to it (`Right).val · nx
- Ymir.Time.secondsecond t is each instant's seconds since its day's 0h, in [0, the day's length).val · ymir
- Ymir_units.Unit.secondsecond is the symbol s.val · ymir
- Hugin_kit.Time.secondsseconds k is the interval of k seconds.val · hugin
- Rig_elf.sectionThe type for sections.type · rig
- Nx.Ptree.Path.segThe type for a path's steps.type · nx
- Ymir_spk.segmentThe type for segments' descriptors.type · ymir
- Nx.Ptree.Path.segmentssegments p is p's steps, from the root.val · nx
- Nx_kernel.segmentsThe type for a contraction cut into groups by ids, an int64 array for S.contract (its shape for Shape.contract) of the batch extents then the segmented axis's.type · nx
- Ymir_spk.segmentssegments t is t's segments' descriptors, in file order.val · ymir
- Talon.SelColumn selectors.module · talon
- Kaun.Cache_index.selectselect columns index is index whose token i of lane b reads only the columns columns.(b).(i), of shape [| batch; seq; k |]: extend's seen has shape [| batch; seq; k; ...val · kaun
- Talon.Query.selectselect os q is q's rows with exactly the columns os, in order:val · talon
- Rig_bnxt.Qp.sendsend qp r ~at n posts the send of the n bytes at offset at of r to qp's peer, which lands them in the region of its next receive.val · rig
- Rig_mlx5_abi.Doorbell.sendsend is the byte offset in a queue pair's record of its send count, a 32-bit word holding the low 16 bits of the ring's producer count.val · rig
- Ymir.Direction.separationseparation a b is the angle between the rays, in [0, π] radians.val · ymir
- Kaun.Cache_index.seqseq index is the number of tokens per lane.val · kaun
- Brot.Decoder.sequencesequence decoders chains decoders left-to-right.val · brot
- Brot.Normalizer.sequencesequence ns is the composition of normalizers ns, applied left to right.val · brot
- Brot.Post_processor.sequencesequence processors chains processors left-to-right.val · brot
- Brot.Pre_tokenizer.sequencesequence ts chains multiple pre-tokenizers left-to-right.val · brot
- Fehu.Space.Sequencemodule · fehu
- Hugin.Mark.seriesseries rows is rows split into series: two rows are in one series iff they have the same index along every axis of the mark's shape but the last and the same category in every channel that reads a band scale and is bound to a role other than x, x2, y and y2.val · hugin
- Munin.History.seriesseries h k is k's samples that are not orphans, in log order.val · munin
- Munin.Metric.seriesThe type for series.type · munin
- Quill_server.serveserve ~create_kernel path starts the web notebook server for the notebook at path.val · quill
- Rig_remote.serveserve a kinds makes the process the agent of one job on its machine, and returns when the job ends.val · rig
- Quill_server.serve_dirserve_dir ~create_kernel ~toc root starts the web notebook server for a directory of notebooks at root.val · quill
- Munin.SessionWriting a run.module · munin
- Quill.SessionInteractive notebook sessions.module · quill
- Rig_bin.Version.sessionsession is what a session's first line carries: v and the version of the protocol between a job's processes, as "1.0.0~alpha3 protocol 5".val · rig
- Fehu.Info.setset k v t is t with k bound to v.val · fehu
- Hugin.View.setset k v view is view with k bound to v, replacing any value bound to a key of the name of k, whatever its sort.val · hugin
- Nx.setset idx v x is x with slice idx x replaced by v, broadcast to the selection's shape.val · nx
- Nx_array.setset a i x stores x at a's index i.val · nx
- Rig_nv.setset p (at, n) x writes the low n bytes of x to the field of n bytes at byte at of p, as get reads it.val · rig
- Tolk.Op.SetSets of operations.module · tolk
- Tolk.Ops.setset p x stores x through p, ends ends, and is p's storage after the store.val · tolk
- Ymir_fits.Fits.Header.setset ~comment v k x h is h with one card giving k the value x, at the place of k's first card or appended, its later cards removed.val · ymir
- Hugin_kit.Scheme.set1set1 is Brewer's Set1, nine colours.val · hugin
- Hugin_kit.Scheme.set2set2 is Brewer's Set2, eight colours.val · hugin
- Hugin_kit.Scheme.set3set3 is Brewer's Set3, twelve colours.val · hugin
- Rig_pci.Window.set32set32 w off x stores the low 32 bits of x at byte off of w.val · rig
- Rig_pci.Window.set64set64 w off x stores x at byte off of w.val · rig
- Rig_pci.Window.set8set8 w off b stores the low 8 bits of b at off of w.val · rig
- Munin.Store.set_aliasset_alias s a alias makes alias name the version a of its artifact, replacing what it named.val · munin
- Quill.Cell.set_attrsset_attrs a c is c with display attributes replaced by a.val · quill
- Rig.set_budgetset_budget d n sets d's budget to n, then frees cached memory (free_cache) until d holds at most n bytes or its cache is empty.val · rig
- Rig_pci.Function.set_bus_masterset_bus_master f on lets f master the bus, reaching system memory by DMA, iff on.val · rig
- Quill.Session.set_cell_attrsset_cell_attrs id attrs s sets the display attributes of cell id.val · quill
- Quill.Session.set_cell_kindset_cell_kind id kind s changes cell id to a code cell in the language l for `Code l, or to a text cell for `Text.val · quill
- Rig_pci.Function.set_config16set_config16 f off x writes the low 16 bits of x at byte off.val · rig
- Rig_pci.Function.set_config32set_config32 f off x writes the low 32 bits of x at byte off.val · rig
- Rig_pci.Function.set_config8set_config8 f off x writes the low 8 bits of x at byte off of f's configuration space.val · rig
- Rig_pci.Function.set_dma_bitsset_dma_bits f n states that f reaches system memory at addresses below 2{^n}.val · rig
- Quill.Doc.set_metadataset_metadata m d is d with metadata replaced by m.val · quill
- Munin.Store.set_notesset_notes s id (Some text) makes text the notes of the run id, replacing them whole, and set_notes s id None removes them.val · munin
- Rig_nv_abi.Method.set_objectset_object e cls binds the engine of class cls to e's subchannel.val · rig
- Quill.Cell.set_outputsset_outputs os c is c with outputs replaced by os.val · quill
- Quill.Cell.set_sourceset_source s c is c with source replaced by s.val · quill
- Rig_pci.Driver.set_stopset_stop h stop makes stop the driver's stop of the function h holds, which stop calls.val · rig
- Munin.Session.set_summaryset_summary s values appends a summary record, in which each value replaces the run's earlier value of its key.val · munin
- Hugin_kit.Scale.setssets p s is true iff s sets p.val · hugin
- Tolk.SettingSettings.module · tolk
- Quill.Kernel.severitytype · quill
- Vega.sgd_initsgd_init p params is the initial state for optimizing params: an all-zero velocity of params' shape and step = 0.val · vega
- Vega.sgd_ptreesgd_ptree p is Nx.Ptree.nest (module Sgd_state) p, the structure of an SGD state over parameters of structure p.val · vega
- Vega.sgd_stateThe state for sgd_step: the momentum velocity, with the shape of the parameters, and the number of completed steps as a scalar tensor.type · vega
- Vega.Sgd_stateThe structure of SGD states.module · vega
- Vega.sgd_stepsgd_step p ~lr st ~params ~grads is (params', st') after one step of gradient descent with heavy-ball momentum.val · vega
- Fehu.Space.shapeshape s is the dimensionality of s, if defined.val · fehu
- Hugin.Mark.shapeshape rows is the shape of the mark (Data and rows).val · hugin
- Norn.Bij.shapeshape b s is the shape of the coordinates of a value of shape s: s for the elementwise bijectors and ordered, s with its last axis one shorter for simplex and sum_to_zero, and s with its last two axes of n replaced by one of n (n - 1) / 2 for cholesky_corr.val · norn
- Norn.Dist.shapeshape d is the shape of d's values.val · norn
- Nx.shapeshape x is x's extents, a fresh array.val · nx
- Nx_array.Layout.shapeshape l is l's extents.val · nx
- Nx_array.Move.shapeshape m s is the shape of the result of m on an argument of shape s.val · nx
- Nx_kernel.ShapeShapes of destinations.module · nx
- Nx_quant.shapeshape w is w's logical shape, [| …; n; k |].val · nx
- Tolk.ShapeShapes, simplification and the nodes built on shapes.module · tolk
- Tolk.Shape.shapeshape u is u's shape.val · tolk
- Ymir.Grid.shapeshape g is the shape of g's cells: the data's last axes.val · ymir
- Ymir_fits.Fits.Image.shapeshape t is [|NAXISn; …; NAXIS1|].val · ymir
- Tolk.Heuristic_common.shape_atshape_at k i is the size of k's axis i.val · tolk
- Tolk.Postrange.Scheduler.shape_lenshape_len k is the number of k's axes.val · tolk
- Tolk.Shape.shape_optshape_opt u is u's shape, or None for nodes that have none, such as effects and program structure.val · tolk
- Tolk.Jit.shapingshaping ~beam () is Setting.shaping as jit_lower ~beam compiles under it: with BEAM at the width the lowering searches, and JITBEAM, which the lowering has read by then, unset (Setting.shaping_under).val · tolk
- Tolk.Setting.shapingshaping () is the name and the current value on the calling domain, as text, of each setting whose reach is Output, sorted by name: what the caches of programs, schedules and searches key on.val · tolk
- Tolk.Setting.shaping_undershaping_under bindings is shaping as it is in context bindings: each setting of bindings whose reach is Output shown at its bound value, the later binding of a setting winning.val · tolk
- Nx.Repr.shardsshards x is x's arrays, one per device of x's placement, in order: [| a |] for a value on one device.val · nx
- Hugin.shareshare pairs f is f with the scales named n regrouped by s for each pair (n, s) of pairs, overriding Scopes and looking through wrappers to find the node's children: grid [ [ a ]; [ b ] ] |> share [ ("x", `Shared) ] puts a and b on one x scale, labelled below b, and share [ ("color", `Independent) ] on a grid gives each cell its own colour scale and legend.val · hugin
- Rig.Claim.shareshare b puts b's memory outside the claims for good, through every view and borrow.val · rig
- Rig_pci.Space.shareshare h x is h's hold on the range of x, which then goes back once neither holds it, or None, holding nothing, if x was given back, as by a release of its holder on another thread.val · rig
- Rig.Buffer.share_byteshare_byte b b' is true iff b and b' share a byte of memory, seen as overlaps sees it, where overlaps holds too.val · rig
- Tolk.Heuristic_host.share_decodesshare_decodes k m upcasts the output axes along which the decoded operands of k's summed product m read one value, so a tile's rows share a decode of it.val · tolk
- Rig.Submission.Run.sharedshared r b n sets the dynamic shared memory of each group of b to n bytes.val · rig
- Tolk.Spec.sharedshared accepts the operations of every stage:val · tolk
- Rig_nv_abi.Method.shared_memory_windowshared_memory_window addr makes kernels' shared memory appear at the address addr.val · rig
- Hugin.sharingThe type for the ways a node shares a scale.type · hugin
- Rig_edge_primitives.shieldshield () starts a pair as mask does, inside which unblock gives nothing back, for code that must take no signal even in its waits, such as a finaliser or an exit handler (rig_shield).val · rig
- Talon.Expr.shiftshift n a is a's value n rows earlier in the frame, or -n rows later if n is negative, and null past the frame's edges.val · talon
- Ymir.Transform.shiftshift r maps x to x - r: the plane whose origin is r.val · ymir
- Tolk.Bigint.shift_leftshift_left n k is n * 2{^k}.val · tolk
- Tolk.Bigint.shift_rightshift_right n k is n / 2{^k} rounded toward negative infinity.val · tolk
- Tolk.Postrange.Scheduler.shift_toshift_to k r amount target splits the axis r into (r', s): r' of r's size divided by amount, and s, of size amount and target's role, a new range unless new_rng is given.val · tolk
- Tolk.Ops.Elementwise.shlval · tolk
- Tolk.Ops.shlval · tolk
- Tolk.Ops.Upat.shlval · tolk
- Tolk.Transcendental.shlshl x n is x * 2n: x shifted left by n bits if x is an integer.val · tolk
- Hugin_float_text.Float_text.shortestshortest f x is the decimal with the fewest digits that reads as x in f, reading to nearest, ties to even; of those, the nearest to x; and at a tie between two, the one whose last digit is even.val · hugin
- Munin_float_text.Float_text.shortestshortest f x is the decimal with the fewest digits that reads as x in f, reading to nearest, ties to even; of those, the nearest to x; and at a tie between two, the one whose last digit is even.val · munin
- Norn_float_text.Float_text.shortestshortest f x is the decimal with the fewest digits that reads as x in f, reading to nearest, ties to even; of those, the nearest to x; and at a tie between two, the one whose last digit is even.val · norn
- Hugin_vg_instances.Instances.showsshows cut ~reach at s b is true iff b scaled by s, moved to at and grown by reach meets cut: an instance at at of scale s of a picture of box b, whose pens reach reach, may paint within cut.val · hugin
- Tolk.Ops.Elementwise.shrval · tolk
- Tolk.Ops.shrval · tolk
- Tolk.Ops.Upat.shrval · tolk
- Tolk.Transcendental.shrshr x n is x // 2n, the division rounding towards negative infinity: x shifted right by n bits if x is an integer.val · tolk
- Tolk.Shape.shrinkshrink u bounds keeps, on each axis, the elements from its start to its end, excluded; None keeps the axis whole.val · tolk
- Tolk.Shape.shrink_toshrink_to u shape keeps the first elements of each axis.val · tolk
- Talon.Kit.shuffleshuffle ~seed q is q's rows in the order of the hash of each row's place in q, counted across its batches:val · talon
- Quill.Kernel.shutdownshutdown k starts k's graceful shutdown.val · quill
- Ymir_units.Vocabulary.sisi is the SI's vocabulary: the symbols of the base units, kg A m s K mol cd rad, then sr and Hz, then the derived units, then g, t, L and l, then the units accepted for use with the SI, as the SI Brochure writes them (Ω, °, ′, ″ in UTF-8).val · ymir
- Rig_pci.Driver.siblingsThe type for what detach does with the other functions of a device, such as a GPU's audio function or a network adapter's other ports.type · rig
- Hugin.sideThe type for the sides of a panel or of a scope's figure.type · hugin
- Rig.Submission.sideThe type for the ends of a copy.type · rig
- Rig_bnxt.Qp.sideThe type for a queue pair's rings.type · rig
- Rig_remote_proxy.Wire.sideThe type for the memory on one side of a copy.type · rig
- Ymir_units.Unit.siemenssiemens is A2 s3 kg-1 m-2.val · ymir
- Ymir_units.Unit.sievertsievert is m2 s-2, which is gray.val · ymir
- Nx.sigmoidsigmoid x is 1 / (1 + exp (-x)), in [0, 1]: +0.val · nx
- Kaun.Loss.sigmoid_bcesigmoid_bce logits targets is the reduction over all elements of the binary cross-entropy between sigmoid logits and targets: -(y log p + (1 - y) log (1 - p)) with p = sigmoid logits and y = targets.val · kaun
- Nx.signsign x is -1, 0 or 1 by x's sign, and NaN for a NaN.val · nx
- Nx_kernel.Prog.E.signval · nx
- Tolk.Bigint.signsign n is -1, 0 or 1 as n is negative, zero or positive.val · tolk
- Rig_bin.Proc.signalsignal pid s sends signal s to pid.val · rig
- Rig.signaledsignaled d is the last value d has reported.val · rig
- Rig_amd.signaledsignaled d is the value in d's word, read with acquire order.val · rig
- Rig_cuda.signaledsignaled d is the value in d's word, read with acquire order.val · rig
- Rig_edge.Driver.signaledsignaled d is the value in d's word, read with acquire order.val · rig
- Rig_metal.signaledsignaled d is the value in d's word, read with acquire order.val · rig
- Rig_nv.signaledsignaled d is the value in d's word, read with acquire order.val · rig
- Rig_remote_proxy.signaledsignaled d is the value in d's word, read with acquire order.val · rig
- Rig_bin.Proc.signalssignals interrupts handles SIGCHLD, SIGPIPE and each signal of interrupts: each wakes wait, and the first interrupt is noted.val · rig
- Tolk.Bigint.signed_extractsigned_extract n off len is extract n off len read as a len-bit two's complement integer, in [-2{^len-1};2{^len-1}-1].val · tolk
- Hugin_float_text.Float_text.significantsignificant n x is |x| rounded to n significant digits, to nearest, ties to even.val · hugin
- Munin_float_text.Float_text.significantsignificant n x is |x| rounded to n significant digits, to nearest, ties to even.val · munin
- Norn_float_text.Float_text.significantsignificant n x is |x| rounded to n significant digits, to nearest, ties to even.val · norn
- Kaun.Fn.silusilu x is x * Nx.sigmoid x, also known as Swish.val · kaun
- Norn.Bij.simplexsimplex maps a vector of K - 1 coordinates to a vector of K positive components summing to one, by the isometric log-ratio: the coordinates are the value's centred logarithms in an orthonormal basis of the vectors summing to zero.val · norn
- Tolk.Renderer.Estimates.simplifysimplify e is e with each count simplified (Shape.ssimplify).val · tolk
- Tolk.Shape.simplifysimplify u is u rewritten with symbolic to a fixed point.val · tolk
- Tolk.SimplifyRange and reduction simplification.module · tolk
- Tolk.Symbolic.simplify_validsimplify_valid valid is the conjunction valid with each clause simplified knowing that the clauses before it hold (uop_given_valid), after removing duplicates and ordering the clauses so that one whose expression others depend on, then one of tighter bounds, comes first.val · tolk
- Tolk.Decomp_op.simplifying_patternssimplifying_patterns ops lowers, for a target with the operations ops:val · tolk
- Norn.Draws.simulatesimulate a b f k d is map a b of f k_ij over d, where the draw j of chain i has the key Nx.Rng.fold_in k (i * n + j), n being the number of draws per chain.val · norn
- Norn_model.simulatesimulate m k is a draw of every site from the model: latent values and data.val · norn
- Nx.sinsin x is the sine of x radians.val · nx
- Nx_kernel.Prog.E.sinval · nx
- Nx_kernel.Prog.singlesingle p is Some n iff p is of_node ~ins n where n reads each of p's operands in order: a Const or a Coord over zero operands, or a kind of m operands reading nodes 0 to m - 1 over m operands.val · nx
- Nx.sinhsinh x is the hyperbolic sine.val · nx
- Nx_kernel.Prog.E.sinhval · nx
- Hugin_vg_vector.Vector.sinkThe type for what receives a path.type · hugin
- Tolk.Ops.sinksink ~kernel us is the Op.t.Sink of us, the root of the kernel kernel if given.val · tolk
- Tolk.Ops.Upat.sinkval · tolk
- Tolk.Ops.sintThe type for symbolic integers: an integer, or an integer node whose value is known when the program runs.type · tolk
- Tolk.Shape.SintSymbolic integers.module · tolk
- Tolk.Ops.sint_to_uopsint_to_uop ~dtype s is s as a node of type dtype, by default Dtype.t.Weak_int.val · tolk
- Tolk.Dtype.sintssints is the signed integers, from Int8 to Int64.val · tolk
- Ymir.Transform.sipsip ~seed (a, b) maps pixel offsets (u, v) from CRPIX, in Unit.one, to (u + f, v + g) with SIP's polynomials f = Σ a.(p).(q) uᵖ v^q and g = Σ b.(p).(q) uᵖ v^q; a and b are [...; n; n], n the order plus one.val · ymir
- Fehu.Buffer.sizesize buf is the number of stored transitions.val · fehu
- Hugin.Layout.sizesize l is the width and height of the figure, in points.val · hugin
- Hugin.Role.sizesize is an area, in square points.val · hugin
- Hugin.SizeFigure sizes.module · hugin
- Hugin.Theme.sizesize th is the base size of th, in points.val · hugin
- Hugin_font.Run.sizesize r is the em size of r.val · hugin
- Rig_amd_abi.Code_object.sizesize co is the length of co's image, in bytes: the size of the object's image as Rig_elf lays it out, rounded up to whole 32-bit words.val · rig
- Rig_mlx5_abi.Completion.sizesize is 64, the bytes of a completion.val · rig
- Rig_mlx5_abi.Doorbell.sizesize is 8, the bytes of a record.val · rig
- Rig_mlx5_abi.Entry.sizesize is 64, the bytes of a basic block.val · rig
- Rig_mlx5_abi.Receive.sizesize is 16, the bytes of a scatter segment.val · rig
- Rig_nv_abi.Cubin.sizesize c is the length of c's image: its laid-out sections, then zeros up to the next multiple of 4 KiB and 4 KiB more.val · rig
- Tolk.Heuristic_common.size_atsize_at k i is the size of k's axis i.val · tolk
- Munin.Artifact.size_bytessize_bytes a is the size of a's content, in bytes; for a directory, the sum of its files' sizes.val · munin
- Tolk.Ops.size_srcsize_src n is the first source of a Op.t.Param, Op.t.Buffer or Op.t.Alloc of n elements: the weak integer constant n, or the empty Op.t.Stack for a scalar (None).val · tolk
- Tolk.Helpers.size_to_strsize_to_str s is the size of s bytes, with two decimals in the largest of GB, MB and KB it reaches (powers of 1024), or as s B below a kilobyte.val · tolk
- Nx.Ptree.SkeletonSkeletons of values.module · nx
- Hugin_font.Font.slantThe type for slants: upright, italic, or upright letterforms slanted.type · hugin
- Hugin_font.Font.slantslant f is `Italic if the OS/2 table of f marks it italic, or its head table does when it has no OS/2 table; `Oblique if OS/2 marks it oblique; and `Normal otherwise.val · hugin
- Rig_amd.sleepsleep d ~seen ~still_ms blocks on the device's events until the word differs from seen, and must return, or raise Fault, at the latest after still_ms milliseconds.val · rig
- Rig_cuda.sleepsleep d ~seen ~still_ms blocks on the device's events until the word differs from seen, and must return, or raise Fault, at the latest after still_ms milliseconds.val · rig
- Rig_edge.Driver.sleepsleep d ~seen ~still_ms blocks on the device's events until the word differs from seen, and must return, or raise Fault, at the latest after still_ms milliseconds.val · rig
- Rig_metal.sleepsleep d ~seen ~still_ms blocks on the device's events until the word differs from seen, and must return, or raise Fault, at the latest after still_ms milliseconds.val · rig
- Rig_nv.sleepsleep d ~seen ~still_ms blocks on the device's events until the word differs from seen, and must return, or raise Fault, at the latest after still_ms milliseconds.val · rig
- Rig_remote_proxy.sleepsleep d ~seen ~still_ms blocks on the device's events until the word differs from seen, and must return, or raise Fault, at the latest after still_ms milliseconds.val · rig
- Nx.sliceslice idx x selects from x axis by axis: each entry selects along the next axis, or a mask along its axes, independently of the others, N excepted, and axes past the list are whole.val · nx
- Talon.Expr.Str.sliceslice ~offset ~length a is the at most length scalar values of a from offset, counted from the end when offset is negative, as Query.slice counts rows.val · talon
- Talon.Query.sliceslice ~offset ~length q is q's rows at the positions offset to offset + length - 1 that q has, in order, a negative offset counting from the end: slice ~offset:0 ~length:10 q is q's first ten rows, and slice ~offset:(-10) ~length:10 q its last ten.val · talon
- Nx.sliding_windowsliding_window ~axis ~window ~step x is x's windows of window consecutive elements along axis (default -1), one every step (default 1).val · nx
- Nx.slogdetslogdet dt a is (sign, logabs) with det a = sign · exp logabs, from lu's pivots: logabs in dt is the sum of their magnitudes' logarithms, and sign is 1.val · nx
- Rig_program.slotThe type for the buffers a step names.type · rig
- Tolk_engine.slotsslots s is, for each slot that a parameter of s takes, a buffer of the parameter's device holding the parameter's bytes, starting at its phase, and None for a slot no parameter takes: slots run and time accept.val · tolk
- Tolk.Shape.smaxsmax ss is the greatest of ss, a Op.t.Max node if some are symbolic, simplified.val · tolk
- Norn.SmcSequential Monte Carlo by tempering.module · norn
- Tolk.Shape.sminsmin ss is the least of ss, as smax.val · tolk
- Talon_arrow.sniffsniff b is the format of the file or stream whose bytes are b, every field of its schema, in order, read as its Arrow type says.val · talon
- Talon_csv.sniffsniff ?rows ?nulls r is the format of the CSV text that r reads, inferred from its first records as sniffing says, with:val · talon
- Talon_json.sniffsniff ?rows r is the format of the JSON text that r reads, inferred from its first rows as sniffing says.val · talon
- Talon_parquet.sniffsniff b is the format of the Parquet file whose bytes are b, read from its footer alone: each column of the file's schema, in order, reading as its Parquet type says.val · talon
- Sowilo.sobelsobel img computes Sobel gradients.val · sowilo
- Nx.softmaxsoftmax ~axes ~scale ~where x is exp (scale x) / Σ exp (scale x) along axes, [-1] by default; scale defaults to 1..val · nx
- Kaun.Loss.softmax_cross_entropysoftmax_cross_entropy logits targets is the reduction over examples of the cross-entropy between softmax logits and targets along the last axis: -Σ targets * log_softmax logits.val · kaun
- Kaun.Loss.softmax_cross_entropy_sparsesoftmax_cross_entropy_sparse logits labels is softmax_cross_entropy with integer class labels: labels has logits' shape without the last axis, and each label is the index of the true class, in [0;classes - 1].val · kaun
- Kaun.Fn.softplussoftplus x is log(1 + exp(x)), a smooth approximation of relu: Nx.softplus, which does not overflow for large x and whose derivative is Nx.sigmoid.val · kaun
- Nx.softplussoftplus x is log (1 + exp x): exp x to the format's precision far below 0., x far above it, never overflowing.val · nx
- Hugin_kit.Dash.solidsolid is an unbroken line, v [].val · hugin
- Jera.SolutionAnswers of solves.module · jera
- Jera.Linear.solvesolve x s a r is the u with a u = r, found by s.val · jera
- Jera.Minimize.solvesolve x m ~lo ~hi ~tol ~budget f x0 is a local minimum of f from x0 by m, in the box [lo, hi] when either side is given: values of x0's structure whose float tensors bound each coordinate from below and from above.val · jera
- Jera.Ode.solvesolve y m ~tol ~budget f ~t0 ~t1 y0 is the state at t1 of the solution from y0 at t0, scalars; t1 may equal or precede t0.val · jera
- Jera.System.solvesolve x m ~linear ~tol ~budget f guess is a zero of f near guess by m.val · jera
- Nx.solvesolve a b is x with matmul a x = b, for a square a, from lu's factors and two triangular solves; b is as solve_triangular takes it.val · nx
- Nx.solve_triangularsolve_triangular a b is x with matmul a x = b, or under transpose with matmul aᵀ x = b or matmul aᴴ x = b, reading a's triangle (default Lower) alone, without its diagonal under ~unit_diag:true (default false), whose ones it takes instead.val · nx
- Nx_cpu.solve_triangularsolve_triangular t ~adjoint ~unit_diagonal ~dst a b stores x with a x = b, or aᴴ x = b under adjoint.val · nx
- Nx_kernel.S.solve_triangularsolve_triangular t ~adjoint ~unit_diagonal ~dst a b stores x with a x = b, or aᴴ x = b under adjoint.val · nx
- Nx_kernel.Shape.solve_triangularsolve_triangular a b is b, […; n; r], for a square of n and b of a's batch.val · nx
- Nx.sortsort ~descending ~axis x is x's elements along axis (default -1) in the sort order, or its reverse where descending (default false), and the positions they come from.val · nx
- Nx_cpu.sortsort ~axis ~descending ~values ~positions x orders each slice of x along axis, stably, into values, bits unchanged, and their int64 positions along axis into positions.val · nx
- Nx_kernel.S.sortsort ~axis ~descending ~values ~positions x orders each slice of x along axis, stably, into values, bits unchanged, and their int64 positions along axis into positions.val · nx
- Nx_kernel.Shape.sortsort ~axis ~k x is the shape of a sort's values and positions: x with k elements along axis with Some k, x with None.val · nx
- Talon.Query.sortsort ks q is q's rows ordered by the keys ks in turn, each by talon's total order in its direction, nulls last unless Order.nulls_first (see Order).val · talon
- Norn.Dist.sortedsorted n d is n independent draws of the scalar d sorted along a new last axis, with the exact density n! times the product of d's on nondecreasing vectors.val · norn
- Quill.Cell.sourcesource c is the source text of cell c.val · quill
- Quill_markdown.Edit.sourcesource t is the original source string.val · quill
- Talon.SourceSources: tables read in batches.module · talon
- Talon_arrow.sourcesource f b is the source, named arrow, of the rows of the file or stream whose bytes are b, read as f says.val · talon
- Talon_csv.sourcesource f open_ is the source, named csv, of the records that the readers open_ () read, as f says, in decode's batches.val · talon
- Talon_json.sourcesource f open_ is the source, named json, of the rows that the readers open_ () read, as f says, in decode's batches.val · talon
- Talon_parquet.sourcesource f b is the source, named parquet, of the rows of the Parquet file whose bytes are b, read as f says.val · talon
- Tolk.Jit.source_digestsource_digest is the digest of this library's sources, which a key of what they make depends on.val · tolk
- Tolk.Source_digestmodule · tolk
- sowiloDifferentiable computer vision on Nx tensors.library · sowilo
- SowiloDifferentiable computer vision on Rune.module · sowilo
- Sowilo vs. OpenCV -- a practical comparisonThis guide explains how Sowilo's image processing model relates to Python's OpenCV, focusing on:guide · sowilo
- Fehu.SpaceObservation and action spaces.module · fehu
- Rig_pci.Page_table.spacespace t is the virtual address space t maps.val · rig
- Rig_pci.SpaceVirtual addresses the GPUs of one vendor share.module · rig
- Hugin.spanspan ~rows ~cols f is f as a grid cell that covers rows rows and cols columns, both 1 by default.val · hugin
- Nx_array.Layout.spanspan l is (lo, hi): every position l reaches lies in [lo, hi), with 0 <= lo, and (0, 0) if l has no element.val · nx
- Quill_markdown.Edit.spanA byte range within the source string.type · quill
- Rig.Profile.spanspan name f is f (), recorded as a span named name on the calling domain's lane of host, from the call until f returns or raises, in each profile taken when the call starts.val · rig
- Rig_pci.Page_table.spanspan t is the number of virtual addresses the tables reach from base: 2{^bits}.val · rig
- Talon.Col.spanspan name is v Kind.span name, which binds every duration and clock type.val · talon
- Talon.Expr.spanspan d is the span d.val · talon
- Talon.Kind.spanspan reads duration and clock columns of every unit.val · talon
- Talon.Time.spanThe type for spans: a signed count of nanoseconds, from -263 to 263 \- 1, about 292 years either way.type · talon
- Talon.Time.SpanSpans.module · talon
- Rig.Buffer.spansspans b is true iff b's bytes are all of its memory, as for a buffer create, of_string, of_bigarray or of_io made, a borrow of one, or a view of all of one.val · rig
- Rig_bin.Proc.spawnspawn ~env prog args ~stdin ~stdout ~stderr starts prog, found through PATH, with args and env (defaults to this process's), and is its pid.val · rig
- Fehu.Space.specStructural description of a space.type · fehu
- Fehu.Space.specspec s is the structural description of s.val · fehu
- Tolk.Setting.specspec is whether the graphs passed between stages are checked against their specifications, from SPEC: 0 checks nothing, and any other value checks them.val · tolk
- Tolk.SpecSpecifications: the nodes each stage of compilation may hold.module · tolk
- Tolk.Ops.specialspecial end_ name is the hardware index name, bounded by end_.val · tolk
- Brot.Encoding.special_tokens_maskspecial_tokens_mask enc is 1 for special tokens (CLS, SEP, padding) and 0 for content tokens.val · brot
- Hugin_kit.Scheme.spectralspectral is Brewer's Spectral, red through yellow to blue.val · hugin
- Ymir_units.Unit.speed_of_lightspeed_of_light is c, 299792458 m s-1.val · ymir
- Ymir_units.Vocabulary.spellspell voc u spells u in as few words as it can:val · ymir
- Ymir_units.Vocabulary.spellingThe type for spellings.type · ymir
- Brot.Pre_tokenizer.splitsplit ~pattern () splits on a literal string pattern.val · brot
- Jera.SplitSplitting methods for separable Hamiltonians.module · jera
- Kaun.Attention.splitsplit ~head_dim l x projects x, of shape [| batch; seq; embed |], through l and splits the result into heads: shape [| batch; heads; seq; head_dim |], where heads is l's output width divided by head_dim.val · kaun
- Nx.Devices.splitsplit ~axis is Placement.split ~axis v.val · nx
- Nx.Host.splitsplit ~axis is Placement.split ~axis v.val · nx
- Nx.Placement.splitsplit ~axis s cuts axis into one window per device of s, in order, as evenly as its extent allows: of n elements over k devices, the first n mod k windows hold n / k + 1 and the others n / k.val · nx
- Nx.Rng.splitsplit ~n k is n keys derived from k (n defaults to 2), independent of each other.val · nx
- Nx.splitsplit ~axis n x is x cut along axis into n runs of consecutive elements, in order.val · nx
- Talon.Expr.Str.splitsplit p a is the pieces of a between the matches of p, found left to right without overlap, as a list[string]: k matches give k + 1 pieces, so a leading, trailing or repeated p gives an empty piece, and no match gives one piece, a itself.val · talon
- Tolk.Heuristic_common.splitsplit k axis amount target applies a split the heuristic has checked and returns the axes it made, with top choosing the outer end.val · tolk
- Tolk.Ops.splitThe type for the loop a host program's launch splits into blocks that the host's cores run at once.type · tolk
- Tolk.Shape.splitsplit ~axis u sizes is the consecutive slices of u along axis (default 0) of sizes elements.val · tolk
- Nx.split_atsplit_at ~axis offsets x is x cut along axis before each of offsets, in order.val · nx
- Nx.Rng.split_batchsplit_batch ~n k is split ~n k as one batch of shape [|n; 2|]: row i holds (split ~n k).(i)'s words.val · nx
- Tolk.Setting.split_reduceopsplit_reduceop lets scheduling split a large reduction into two kernels to expose more parallelism, from SPLIT_REDUCEOP.val · tolk
- Brot.Pre_tokenizer.split_regexsplit_regex ~pattern () is split on the matches of the regular expression pattern rather than on the occurrences of a string.val · brot
- Tolk.Postrange.split_targetssplit_targets target is the roles of the axes a split can move to target: global, local and weak axes to Opt.target.Upcast, and global and weak axes to Opt.target.Local.val · tolk
- Tolk.Ops.split_uopsplit_uop u op is the operands of the tree of op nodes rooted at u, left to right: [u] if u's operation is not op.val · tolk
- Nx.sqrtsqrt x is the square root, NaN below -0., correctly rounded on floats.val · nx
- Nx_kernel.Prog.E.sqrtval · nx
- Tolk.Ops.Elementwise.sqrtval · tolk
- Tolk.Ops.sqrtval · tolk
- Tolk.Ops.Upat.sqrtval · tolk
- Hugin_kit.Symbol.squaresquare is a square with sides parallel to the axes.val · hugin
- Nx.squaresquare x is mul x x.val · nx
- Nx.squeezesqueeze ~axes x is x without the axes axes; without axes, without every axis of extent 1.val · nx
- Tolk.Shape.squeezesqueeze ~axis u removes axis if its size is 1, or every axis of size 1.val · tolk
- Jera.Sde.sra1sra1 is Rößler's (2010) SRA1, for additive noise, where the diffusion does not depend on the state: strong order 3/2, reading the Lévy area; 1/2 otherwise.val · jera
- Tolk.Ops.srcsrc u is u's sources, in order.val · tolk
- Tolk.Ops.src_without_bodysrc_without_body u is a call's arguments, and any other node's sources.val · tolk
- Tolk.Render.srendersrender s is an integer s in decimal, and a node s as render writes it.val · tolk
- Tolk.Shape.ssimplifyssimplify u is simplify u as an Int if it is an integer constant, a Sym otherwise.val · tolk
- Jera.Ode.ssprk3ssprk3 is the three-stage strong-stability-preserving method of Shu and Osher (1988): order 3, three evaluations per step.val · jera
- Hugin_kit.StackStacks.module · hugin
- Nx.stackstack ~axis xs is xs joined along a new axis at position axis of the result (default 0), in order.val · nx
- Tolk.Shape.stackstack ~axis us is the Op.t.Stack of us, each converted to their common type (Ops.ccast) except `Invalid constants, with the new axis moved to axis (default 0).val · tolk
- Tolk.Indexing.stack_selectstack_select i lanes is the lane of lanes at the index i, a tree of selections on the bits of i, the most significant first: of two lanes, where (i mod 2 < 1) l0 l1.val · tolk
- Tolk.Setting.stagestage stages a tensor core's operands through shared memory where the target has it, and a decoded operand through a panel on the host (Stage), and gives a staged core its own tile (Heuristic.hand_coded_optimizations), from STAGE.val · tolk
- Tolk.StageStaging a tensor core's operands through shared memory, and a decoded operand through a panel on the host (Decoded panels on the host).module · tolk
- Hugin_vg.Picture.stampstamp ~fills ~strokes ~scales xs ys p paints a copy of p, an instance, at each point (xs.(i), ys.(i)), in the order of i.val · hugin
- Hugin_vg_vector.Vector.stampThe type for stamps as the walk writes them.type · hugin
- Quill_markdown.stampA stamp identifies a file's content as stored: its device, inode, size and modification time.type · quill
- Quill_markdown.stampstamp path is the stamp of the file path names, through links, or None if there is none.val · quill
- Talon.Time.stampThe type for stamps of the clock 'c.type · talon
- Quill_book.Theme.standalone_htmlstandalone_html is the page of a single notebook outside a site, with style_css inlined.val · quill
- Quill_book.Render.standalone_page_htmlstandalone_page_html ~title ~content ~live_reload_script renders a self-contained HTML page for a single notebook.val · quill
- Norn.Dist.standardizestandardize d is the bijector whose coordinates are d's draws standardised: (x - loc) / scale for normal, student_t, cauchy, laplace and logistic, (log x - loc) / scale for lognormal, L⁻¹ (x - loc) for mvn, and the same for iid of these.val · norn
- Hugin_kit.Symbol.starstar is a regular five-pointed star pointing up, its inner corners on the lines that join its points.val · hugin
- Fehu.Space.Discrete.startstart s is the starting value of s.val · fehu
- Munin.Session.startstart ~experiment () creates a run of experiment in store and is its session.val · munin
- Rig_amd_abi.Counter.startstart g l ~filter_address is the packets, as little-endian 32-bit words, that wait until the queue's dispatches before them complete, reset every counter of g, select l's, and start counting the waves that the filter word at filter_address selects (filter).val · rig
- Rig_amd_abi.Thread_trace.startstart g ~wgps ~size ~address is the packets, as little-endian 32-bit words, that start tracing on every shader engine of g whose arrays run work (wgps), engine e writing at most size bytes into its buffer at address + (e * size).val · rig
- Munin.Run.started_atstarted_at r is the time of r's start.val · munin
- Norn.Ensemble.stateThe type for states of walkers over positions 'u.type · norn
- Norn.Hmc.stateThe type for states of chains over positions 'u.type · norn
- Norn.Nested.stateThe type for states of nested sampling over positions 'u.type · norn
- Norn.Nuts.stateThe type for states of chains over positions 'u.type · norn
- Norn.Smc.stateThe type for states of tempering over positions 'u.type · norn
- Rig_remote_proxy.Link.stateThe type for the states of a job.type · rig
- Ymir_spk.statestate t ~target ~center tdb is (p, v), target's position in km and velocity in km/s from center at the instants tdb, each [...; 3] on ICRS's axes, which J2000's are in JPL's ephemerides.val · ymir
- Vega.Loss_scale.staticstatic at s is the fixed scale s at at, whole on each of its devices: step never changes it.val · vega
- Station ResamplingA weather station reports once a day, and an analysis often wants months.guide · talon
- Fehu.Eval.statsThe type for evaluation statistics.type · fehu
- Hugin_kit.StatsStatistics.module · hugin
- Kaun.Batch_norm.StatsRunning statistics: the exponential moving averages of per-feature batch mean and (population) variance maintained by training-mode apply and consumed by eval-mode apply.module · kaun
- Munin.History.statsstats h k is the statistics of k's samples that are not orphans, or None if there is none.val · munin
- Munin.Metric.statsThe type for the statistics of a series of at least one sample.type · munin
- Norn.Ensemble.statsThe type for statistics of transitions, one element per walker.type · norn
- Norn.Ensemble.statsstats dtype is the structure of statistics at dtype.val · norn
- Norn.Hmc.statsThe type for statistics of transitions.type · norn
- Norn.Hmc.statsstats dtype is Stats.ptree dtype.val · norn
- Norn.Nuts.statsstats dtype is Stats.ptree dtype.val · norn
- Norn.StatsStatistics of Hamiltonian transitions.module · norn
- Munin.History.stats_ofstats_of run is the statistics of each key of run that has a sample, sorted by key, equal to what stats gives for the run's history.val · munin
- Jera.Solution.statusThe type for a lane's outcome.type · jera
- Munin.Run.statusThe type for statuses (The writer and the status of a run).type · munin
- Munin.Run.statusstatus r is r's status as read.val · munin
- Quill.Kernel.statustype · quill
- Quill.Kernel.statusstatus k is k's status.val · quill
- Rig_bin.Proc.statusstatus st is the exit status a shell shows for st: its own, or 128 + N for signal N.val · rig
- Rig_mlx5_abi.Completion.statusThe type for the outcomes of entries.type · rig
- Nx.stdstd ~axes ~ddof x is the standard deviation σ: the square root of the variance var states, rounded once to the dtype.val · nx
- Talon.Expr.stdstd a is the sample standard deviation of a's values, dividing by n - 1, and null for fewer than two values.val · talon
- Ymir_units.Unit.stefan_boltzmannstefan_boltzmann is σ, 2π5k4/(15h3c2).val · ymir
- Jera.Piecewise.steffensteffen x y is Steffen's (1990) interpolant, once continuously differentiable: degree 3.val · jera
- Fehu.Env.stepThe type for step results.type · fehu
- Fehu.Env.stepstep env action advances the environment by one timestep.val · fehu
- Fehu.Vec_env.stepThe type for batched step results.type · fehu
- Fehu.Vec_env.stepstep t actions steps all environments with the given actions.val · fehu
- Jera.Split.stepstep m ~kick ~drift h s is s after one step h of m.val · jera
- Munin.Artifact.stepstep a is the step the producing run gave for a, if it gave one.val · munin
- Munin.Metric.stepstep s i is the step of s's ith sample.val · munin
- Norn.Ensemble.stepstep u lp k s is one transition of every walker from s.val · norn
- Norn.Hmc.stepstep u lp k s is one transition of every chain from s.val · norn
- Norn.Nested.stepstep u ~prior ~likelihood k s deletes s.batch live points and replaces them, or is s if s has taken s.budget steps.val · norn
- Norn.Nuts.stepstep u lp k s is one transition of every chain from s.val · norn
- Norn.Smc.stepstep u ~prior ~likelihood k s moves the particles of s to the next temperature.val · norn
- Rig_program.stepThe type for the steps of a run.type · rig
- Talon.Time.stepThe type for steps: the units in which temporal expressions floor and offset stamps and dates.type · talon
- Vega.Loss_scale.stepstep p s ls ~grads update x is one optimizer step from x under the loss scale ls, and the scale for the next step.val · vega
- Hugin_kit.Curve.step_afterstep_after goes from each point parallel to the x axis as far as the next point's x, then parallel to the y axis to the next point: a value holds until the next one.val · hugin
- Hugin_kit.Curve.step_beforestep_before goes from each point parallel to the y axis as far as the next point's y, then parallel to the x axis to the next point: a value holds since the previous one.val · hugin
- Munin.History.step_metricstep_metric h k is the key that k's latest metric record names as its x axis, if it names one.val · munin
- Hugin_kit.Curve.step_midstep_mid goes from each point parallel to the x axis halfway to the next point's x, then parallel to the y axis to the next point's y, then parallel to the x axis to the next point: steps fall midway between points.val · hugin
- Rig_program.step_pointsstep_points p is a fresh array for the step_points of p's frames, each element the same point of Rig.host, which is reached.val · rig
- Fehu.Env.step_resultstep_result ~observation () constructs a step result.val · fehu
- Hugin_kit.Scale.steppedstepped s is true iff s sets stepped to true.val · hugin
- Munin.Run.stepssteps r is the largest step r declared it will log, if it declared one.val · munin
- Ymir_units.Unit.steradiansteradian is rad2.val · ymir
- Nx.stftstft dt ~window x is the short-time transform of x along its last axis: frames of window samples, one every step (default window / 4, at least 1), each multiplied by win and transformed by rfft.val · nx
- Norn.Evidence.stopThe type for the reasons a run stopped.type · norn
- Norn.Evidence.stopstop z is why the run that gave z stopped.val · norn
- Rig_amd.stopstop d ~fault stops the lost d and must not wait for its work.val · rig
- Rig_amd_abi.Counter.stopstop g ~wgps l ~address is the packets, as little-endian 32-bit words, that wait until the queue's dispatches before them complete, stop counting, and store l's values from address as t lays them out, 0 for each work-group processor that wgps says runs no work.val · rig
- Rig_amd_abi.Thread_trace.stopstop g ~wgps ~address is the packets, as little-endian 32-bit words, that stop tracing on every shader engine of g whose arrays run work (wgps) and wait until each has written its trace out.val · rig
- Rig_cuda.stopstop d ~fault stops the lost d and must not wait for its work.val · rig
- Rig_edge.Driver.stopstop d ~fault stops the lost d and must not wait for its work.val · rig
- Rig_edge.Io.stopstop d ends d.val · rig
- Rig_metal.stopstop d ~fault stops the lost d and must not wait for its work.val · rig
- Rig_nv.stopstop d ~fault stops the lost d and must not wait for its work.val · rig
- Rig_pci.Driver.stopstop h calls the driver's stop (set_stop), then releases the function (Function.release).val · rig
- Rig_remote_proxy.stopstop d ~fault stops the lost d and must not wait for its work.val · rig
- Tolk_engine.storageThe type for where a schedule's storage that no buffer is bound to lives.type · tolk
- Tolk.Ops.storage_basestorage_base u is the storage u targets: base u without bitcasts and Op.t.Afters.val · tolk
- Tolk.Dtype.storage_fmtstorage_fmt dt is the format character of the storage of dt values: fmt for most, 'H', a 16-bit unsigned integer, for Bfloat16, and 'B', an 8-bit unsigned integer, for the 8-bit floats.val · tolk
- Tolk.Shape.storage_phasestorage_phase u is the alignment and phase (align, phase) (Ops.param_arg) of the storage u views: its storage's, moved by the bytes a shrink skips when it shrinks the storage seen whole, a buffer or a stage the schedule allocates, in row-major order.val · tolk
- Munin.Session.storestore s is the store that holds s's run.val · munin
- Munin.StoreStores: a root directory and what it holds.module · munin
- Talon.Expr.storestore ty a is a typed as ty, as if a met an operand of type ty: a literal, null, const, $ or of_option result takes ty, and an a of a type that ty contains is widened to it.val · talon
- Tolk.Ops.storestore p x is the Op.t.Store of x through p, guarded by gate.val · tolk
- Tolk.Ops.Upat.storeval · tolk
- Tolk.Shape.store_callstore_call dst src is the call that copies src into dst.val · tolk
- Munin.Store_formatThe files of a munin store.module · munin
- Tolk.Heuristic_common.stored_addressesstored_addresses k is the addresses k stores to.val · tolk
- Talon.Expr.StrText.module · talon
- Brot.streamThe type for streaming decodings.type · brot
- Hugin_gg.Affine.stretchstretch m is the largest factor by which m scales a length, the largest singular value of linear m: a vector of length l maps to one of length at most stretch m *.val · hugin
- Nx_array.Layout.stridestride l i is the stride of l's axis i.val · nx
- Tolk.Heuristic_common.stridestride r i is the stride of the range r in the index i, when a term of i is r or r * c: 1 or c.val · tolk
- Nx_array.Layout.stridesstrides l is l's strides.val · nx
- Talon.Col.stringstring name is v Kind.string name, which binds string and categorical types.val · talon
- Talon.Expr.stringstring s is the text s.val · talon
- Talon.Kind.stringstring reads string and categorical columns.val · talon
- Talon.Type.stringstring is String.val · talon
- Tolk.Setting.stringstring ~reach key default is a setting whose initial value is the variable key as written, or default if key is unset.val · tolk
- Ymir_fits.Fits.Value.stringstring is a quoted string, with doubled quotes read as one, the CONTINUE records after a string ending in & joined (§4.2.1.2), and trailing spaces dropped: '' is "" and ' ' is " ".val · ymir
- Hugin.stringsstrings ~scale ~title a is the labelled categories named by the elements of a, one datum per element, with scale and title as for num: its shape is [|Array.length a|], and it contributes its distinct strings to its scale's domain in order of first appearance.val · hugin
- Brot.Decoder.stripstrip ~content ~start ~stop () removes up to start leading and stop trailing occurrences of content from each token.val · brot
- Brot.Normalizer.stripstrip ?left ?right () is a normalizer that strips Unicode whitespace from text boundaries.val · brot
- Brot.Normalizer.strip_accentsstrip_accents removes every mark (Mn, Mc, Me).val · brot
- Quill_book.Render.strip_html_tagsstrip_html_tags s removes HTML tags from s and collapses whitespace.val · quill
- Tolk.Helpers.strip_parensstrip_parens s is s without its first and last characters if they are parentheses enclosing the whole of s, and s otherwise.val · tolk
- Hugin.Role.strokestroke is the colour that draws lines and outlines.val · hugin
- Hugin_gg.StrokeStroke styles.module · hugin
- Hugin_vg.Picture.strokestroke s c q paints the outline of q under s with c.val · hugin
- Hugin_vg_vector.Vector.strokeThe type for strokes as written.type · hugin
- Hugin_kit.Symbol.strokedstroked is the symbols designed to be stroked: circle, plus, times, triangle, asterisk, square and diamond.val · hugin
- Tolk.Dtype.strongstrong dt commits dt to a width: Weak_int is default_int (), Weak_float is default_float (), and any other dt is itself.val · tolk
- Nx.Ptree.Walk.structurestructure s c x walks x, a value of the description s, at c's path: each of its arrays as tensor walks one, each of its reports as s makes it, at s's paths after c's.val · nx
- Ymir.Cosmology.StructureStructure is walk as a structure whose arrays are all positions: Nx.Ptree.instantiate (module Cosmology.Structure) describes a cosmology of tensors.module · ymir
- Ymir_units.Quantity.StructureStructure is walk as a structure whose arrays are all positions.module · ymir
- Tolk.Ops.structure_ranksstructure_ranks u numbers u and the nodes it reaches, call bodies included, in the order of compare_structure, and is the function that gives each its number: of two of them n0 and n1, n0's number is less than n1's exactly when compare_structure n0 n1 < 0, and the two are equal exactly when it is 0.val · tolk
- StructuresA model, an optimizer state or a decoding cache is an ordinary OCaml record of arrays.guide · nx
- Sowilo.structuring_elementstructuring_element shape (h, w) is a host structuring element of the given shape and size.val · sowilo
- Norn.Dist.student_tstudent_t ~df ~loc ~scale is Student's t with df degrees of freedom, in (0, inf), located and scaled.val · norn
- Quill_book.Theme.style_cssstyle_css is the stylesheet of every page, which Build writes as style.css at the root of a site.val · quill
- Hugin_text.Text.subsub t is t as a subscript: every size factor k and shift d become 0.7 *.val · hugin
- Nx.subsub a b is a - b, as add.val · nx
- Nx_kernel.Prog.E.subval · nx
- Nx_ragged.subsub r ~offset ~length is rows offset to offset + length - 1 of r, over r's values, in O(1).val · nx
- Nx_wide.subsub a b is a - b within 3u² / (1 - 4u) relative, and exactly zero when a = b.val · nx
- Rig_pci.Window.subsub w off n is the n bytes of w from byte off on.val · rig
- Tolk.Bigint.subsub n0 n1 is n0 - n1.val · tolk
- Tolk.Ops.Elementwise.subsub x y is x + neg y.val · tolk
- Tolk.Ops.subsub x y is x + neg y.val · tolk
- Tolk.Ops.Upat.subsub x y is x + neg y.val · tolk
- Ymir.Observation.subsub a b is a's data less b's, as add combines them.val · ymir
- Ymir_units.Quantity.subsub a b is map2 Nx.sub a b.val · ymir
- Nx.sub_ssub_s x c is sub x c.val · nx
- Rig.SubmissionPrepared submissions.module · rig
- Rig.Submission.Run.submissionsubmission r is the submission whose blocks r holds.val · rig
- Rig.submitsubmit ~waits r ~buffers runs the work of r's submission s (Submission.Run.submission) once and returns its point (d, v), where d is s's device and v is one more than submitted d.val · rig
- Rig_program.submitThe type for submissions: parts on device device.type · rig
- Rig.Buffer.submit_copysubmit_copy ~waits ~src ~dst boxes copies the bytes of each box of boxes from src into dst, one box after the other, and is a point that is reached once every byte is there.val · rig
- Rig.Buffer.submit_fillsubmit_fill ~waits b c sets every byte of b to c, after the points of waits (defaults to [||]), and is a point that is reached once they are set, returned as submit_copy returns it.val · rig
- Rig.submittedsubmitted d is the last value assigned on d's timeline (Timelines), or, before any, the value d's timeline started after: 0 for a device that took a slot no device took before, and the last value the slot's earlier device was given otherwise.val · rig
- Hugin_font.Font.subsetsubset f gs is an OpenType file holding the face of f cut down to the glyphs it keeps: glyph 0, the glyphs gs, and the components of every composite glyph it keeps.val · hugin
- Nx_kernel.Prog.substsubst ?outs ~ins f p is the program over operands of dtypes ins that computes p's outputs, or with outs the outputs of p at those indices, in that order, with each leaf replaced.val · nx
- Tolk.Ops.substitutesubstitute ~calls ~pass u subs is u with each node of subs replaced by its pair, top-first, rewriting with extra_pm as well, which reads the substitution, and treating calls and results as graph_rewrite does.val · tolk
- Talon.Kit.subtract_intervalssubtract_intervals ~by (s, e) rest q is the values that q's set of a key holds and rest's does not, written q |> Kit.subtract_intervals (s, e) rest, in normal form (Normal form).val · talon
- Tolk.Bigint.succsucc n is n + 1.val · tolk
- Talon.Expr.Str.suffixsuffix s matches s at the end of the text.val · talon
- Talon.Sel.suffixsuffix s selects the columns whose name ends with s, in schema order.val · talon
- Nx.sumsum ~axes x adds the elements from +0: a sum of -0.val · nx
- Nx_wide.sumsum ~axes w adds w along axes (default all, a negative axis counting from the end) in a balanced tree of adds, ⌈log₂ n⌉ deep for n summands, whose association depends only on the shape: within ((1 + 3u² / (1 - 4u))^⌈log₂ n⌉ - 1) · Σ|wᵢ|.val · nx
- Talon.Expr.sumsum a is the sum of a's values: int64 over integers, a's type over floats and durations.val · talon
- Norn.Bij.sum_to_zerosum_to_zero maps a vector of K - 1 coordinates to a vector of K components summing to zero, by an orthonormal basis of that subspace.val · norn
- Munin.History.summarysummary h is the run's summary: each key's latest value, sorted by key.val · munin
- Norn.SummarySummaries of draws, with their findings.module · norn
- Hugin_text.Text.supsup t is t as a superscript: every size factor k and shift d become 0.7 *.val · hugin
- Ymir.Frame.supergalactictype · ymir
- Ymir.Frame.supergalacticsupergalactic is supergalactic coordinates, defined on galactic: the north supergalactic pole at Galactic (47.37°, +6.32°), the origin at l = 137.37°.val · ymir
- Norn.Dist.supportsupport d is the set of d's values.val · norn
- Norn.SupportSets of values a distribution gives positive density.module · norn
- Tolk.Support_memoryMemory planning.module · tolk
- Tolk.Renderer.supported_dtypessupported_dtypes r is the data types of Dtype.all, in order, that r has natively, without Dtype.t.Float64 if Dtype.t.Int64 is emulated (setting Setting.emulated_dtypes): a double cannot be bitcast without a 64-bit integer.val · tolk
- Nx.svdsvd dt a is (u, s, vh), a the product of u, the diagonal matrix of s and vh, s the […; k] singular values in dt, descending and not negative, a zero one +0; u and vh have orthonormal columns and rows: […; m; k] and […; k; n], or under ~factors:Complete […; m; m] and […; n; n].val · nx
- Nx_cpu.svdsvd ~s ~vectors:(~u, ~vh) a stores a = u diag s vh, s descending and not negative, a zero one +0.val · nx
- Nx_kernel.S.svdsvd ~s ~vectors:(~u, ~vh) a stores a = u diag s vh, s descending and not negative, a zero one +0.val · nx
- Nx_kernel.Shape.svdsvd f a is the shapes of s, u and vh: […; k], then […; m; k] and […; k; n] under Reduced, […; m; m] and […; n; n] under Complete.val · nx
- Nx.svdvalssvdvals dt a is svd's s of a, computed without its vectors by an iteration that may round otherwise: its elements lie within the kernel library's bounds of s's, not always bit for bit.val · nx
- Nx.swapaxesswapaxes a b x is x with its axes a and b exchanged.val · nx
- Rune.switchswitch b i branches is f (), f the branch at position i of branches, a 0-d int32: that branch alone computes.val · rune
- Tolk.Symbolic.symsym is Shape.symbolic and pm_simplify_valid, followed by:val · tolk
- Tolk.Shape.sym_compilesym_compile s var is the function computing s in an environment, each variable v of s read by var v: sym_compile s var env is sym_infer s vars for vars binding each variable v to var v env.val · tolk
- Tolk.Shape.sym_compile_zsym_compile_z s var is sym_compile's function with values of any size: sym_compile_z s var env is the exact value of s, an integer whether or not it fits an int, such as a kernel's estimated operations over a loop bounded by a value loaded from memory (Ops.estimates).val · tolk
- Tolk.Shape.sym_infersym_infer s vars is the value of s with each variable named in vars bound to its value.val · tolk
- Hugin.Role.symbolsymbol is a marker shape.val · hugin
- Hugin_kit.SymbolSymbols.module · hugin
- Rig_elf.symbolThe type for the entries of a symbol table.type · rig
- Rig_elf.symbolsymbol o name is Some offset for the first symbol of o.symbols, by index, named name whose place is Image { offset; _ }, and None if there is none.val · rig
- Ymir_units.Unit.symbolsymbol name is the symbol name.val · ymir
- Tolk.Shape.symbolicsymbolic is symbolic_simple and commutative, followed by rewrites that match deeper:val · tolk
- Tolk.SymbolicSymbolic simplification for code generation.module · tolk
- Tolk.Shape.symbolic_simplesymbolic_simple folds nodes whose value follows from their operation and their sources alone:val · tolk
- Hugin_kit.Scale.symbolssymbols s is the symbols of s, if set, as a fresh array.val · hugin
- Hugin_kit.Scale.symlogsymlog ~constant () is a symmetric logarithmic scale with constant constant, which defaults to 1.val · hugin
- Munin.Session.syncsync s makes every record s wrote survive a power loss.val · munin
- Nx.syncsync x waits for the work x depends on, then drains the checks pending on x's set, raising the earliest failure, as stated above.val · nx
- Ymir.Healpix.synthesissynthesis h a is the map [...; pixels h], in h's scheme, of the coefficients a, [...; lmax + 1; lmax + 1], summed at each pixel's centre.val · ymir
- Jera.SystemZeros of systems.module · jera
- Talon.Tz.systemsystem () is of_dir d for the system's zoneinfo directory d: the value of the environment variable TZDIR when it is set and not empty, and "/usr/share/zoneinfo" otherwise.val · talon
- Tolk.Helpers.systemsystem ~input cmd is the output of the command cmd, a program and its arguments separated by spaces, run with input on its standard input or, without input, with the process's: what it writes on its standard output and standard error, together, without the white space that starts and ends it.val · tolk
- Tolk.Host_config.systemval · tolk
- System monitoringA run records its machine beside its metrics: how busy the CPUs were, how much memory the process held, what its GPUs drew.guide · munin
t
- Brot.Decoder.tThe type for decoders.type · brot
- Brot.Encoding.tThe type for tokenization encodings.type · brot
- Brot.Normalizer.tThe type for normalizers.type · brot
- Brot.Post_processor.tThe type for post-processors.type · brot
- Brot.Pre_tokenizer.tThe type for pre-tokenizers.type · brot
- Brot.tThe type for tokenizers.type · brot
- Fehu.Buffer.tA replay buffer of transitions.type · fehu
- Fehu.Collect.tThe type for trajectories.type · fehu
- Fehu.Env.tEnvironment handle.type · fehu
- Fehu.Info.tThe type for info dictionaries.type · fehu
- Fehu.Space.tThe type for spaces over values of type 'a.type · fehu
- Fehu.Value.tThe type for universal values.type · fehu
- Fehu.Vec_env.tThe type for vectorized environments.type · fehu
- Hugin.Coord.tThe type for coordinate systems.type · hugin
- Hugin.Drawing.tThe type for drawings.type · hugin
- Hugin.Layout.tThe type for laid-out figures.type · hugin
- Hugin.Resolved.tThe type for resolved figures.type · hugin
- Hugin.Role.tThe type for roles taking channels of domain values 'd and range 'r.type · hugin
- Hugin.Size.tThe type for sizes.type · hugin
- Hugin.tThe type for figures.type · hugin
- Hugin.Theme.tThe type for themes.type · hugin
- Hugin.View.tThe type for views: keys bound to values.type · hugin
- Hugin_font.Font.tThe type for fonts.type · hugin
- Hugin_font.Run.tThe type for glyph runs.type · hugin
- Hugin_gg.Affine.tThe type for affine maps.type · hugin
- Hugin_gg.Box2.tThe type for boxes.type · hugin
- Hugin_gg.Color.tThe type for colours.type · hugin
- Hugin_gg.P2.tThe type for points.type · hugin
- Hugin_gg.Path.tThe type for paths.type · hugin
- Hugin_gg.Stroke.tThe type for stroke styles.type · hugin
- Hugin_gg_kit.Field2.tThe type for fields.type · hugin
- Hugin_gg_kit.Pgon2.tThe type for polygons.type · hugin
- Hugin_gg_kit.Ring2.tThe type for rings.type · hugin
- Hugin_kit.Curve.tThe type for curves.type · hugin
- Hugin_kit.Dash.tThe type for dash patterns.type · hugin
- Hugin_kit.Locale.tThe type for locales.type · hugin
- Hugin_kit.Number.tThe type for number formats.type · hugin
- Hugin_kit.Scale.tThe type for scales over domain values of type 'd.type · hugin
- Hugin_kit.Scheme.tThe type for colour schemes.type · hugin
- Hugin_kit.Symbol.tThe type for symbols.type · hugin
- Hugin_kit.Ticks.tThe type for the ticks of an axis.type · hugin
- Hugin_kit.Time.tThe type for instants.type · hugin
- Hugin_text.Text.Layout.tThe type for layouts.type · hugin
- Hugin_text.Text.tThe type for texts.type · hugin
- Hugin_vg.Picture.tThe type for pictures.type · hugin
- Hugin_vg.Renderable.tThe type for renderables.type · hugin
- Jera.Grid.tThe type for grid series over breaks of dtype 'b.type · jera
- Jera.Linear.tThe type for solvers of systems on values of type 'x.type · jera
- Jera.Minimize.tThe type for methods for objectives of type 'f over values of type 'x.type · jera
- Jera.Ode.tThe type for one-step methods with tags 'k over states 'y and times of dtype 't.type · jera
- Jera.Piecewise.tThe type for piecewise series with values of structure 'v over breaks of dtype 'b: each float leaf of the coefficients has shape [pieces; degree + 1] @ value, and evaluation at points of shape q gives it shape q @ value.type · jera
- Jera.Quad.Box.tThe type for boxes of dtype 'b.type · jera
- Jera.Quad.Range.tThe type for ranges of dtype 'b.type · jera
- Jera.Quad.Rule.tThe type for rules with tags 'k.type · jera
- Jera.Sde.Brownian.tThe type for Brownian paths of dtype 'b.type · jera
- Jera.Sde.tThe type for methods.type · jera
- Jera.Solution.tThe type for answers of type 'a.type · jera
- Jera.Split.tThe type for splittings: the coefficients a of the kicks and b of the drifts.type · jera
- Jera.System.tThe type for methods for systems on values of type 'x.type · jera
- Jera.Tol.tThe type for tolerances.type · jera
- Kaun.Attention.Cache.tThe type for key-value caches over payload 'a.type · kaun
- Kaun.Attention.tThe type for attention parameters over payload 'a: the query, key, value and output projections.type · kaun
- Kaun.Batch_norm.Stats.tThe type for running statistics over payload 'a, at tensor payloads each of shape [| features |].type · kaun
- Kaun.Batch_norm.tThe type for batch-norm parameters over payload 'a: per-feature scale gamma and shift beta, at tensor payloads each of shape [| features |].type · kaun
- Kaun.Cache_index.tThe type for cache indices of batch lanes of seq tokens each, on the set 'd.type · kaun
- Kaun.Conv.tThe type for convolution parameters over payload 'a.type · kaun
- Kaun.Embedding.tThe type for embedding parameters over payload 'a.type · kaun
- Kaun.Init.tThe type for initializers producing float tensors with layout 'b, on the set 'd.type · kaun
- Kaun.Layer_norm.tThe type for layer-norm parameters over payload 'a.type · kaun
- Kaun.Linear.tThe type for linear-layer parameters over payload 'a.type · kaun
- Kaun.Rms_norm.tThe type for RMS-norm parameters over payload 'a.type · kaun
- Kaun.Rope.tThe type for rotary schedules on the set 'd: the inverse frequencies of one attention head, head_dim / 2 of them, and their angles' cosines and sines at the positions 0 to context - 1, in float32.type · kaun
- Munin.Artifact.tThe type for artifact versions.type · munin
- Munin.History.tThe type for histories.type · munin
- Munin.Provenance.tThe type for provenance.type · munin
- Munin.Run.tThe type for runs as read.type · munin
- Munin.Session.tThe type for sessions.type · munin
- Munin.Store.tThe type for stores.type · munin
- Munin.Value.tThe type for values.type · munin
- Norn.Bij.tThe type for bijectors over tensors of element type 'f.type · norn
- Norn.Dist.tThe type for distributions over values 'x whose log densities have element type 'f.type · norn
- Norn.Draws.tThe type for draws of 'u.type · norn
- Norn.Evidence.tThe type for evidence estimates over positions 'u.type · norn
- Norn.Gaussian.tThe type for Gaussians over 'u whose log densities have element type 'f.type · norn
- Norn.Stats.tThe type for statistics of transitions over element type 'f.type · norn
- Norn.Summary.tThe type for summaries.type · norn
- Norn.Support.tThe type for supports.type · norn
- Norn.Weighted.tThe type for weighted draws of 'u.type · norn
- Norn_model.Init.tThe type for strategies over latent values 'p.type · norn
- Norn_model.tThe type for models with latent values 'p, observed values 'y and log densities of element type 'f.type · norn
- Nx.Devices.tThe type for device sets of brand 'd: distinct devices, and the kernels that compute on them.type · nx
- Nx.Mesh.tThe type for named axes over a set of brand 'd.type · nx
- Nx.Pattern.tThe type for axis patterns, parsed and checked.type · nx
- Nx.Placement.tThe type for placements over a set of brand 'd.type · nx
- Nx.Ptree.Path.tThe type for paths.type · nx
- Nx.Ptree.S.tThe type for values with parameter 'a and fixed arrays on 'd.type · nx
- Nx.Ptree.Skeleton.tThe type for a value's skeleton: its visits.type · nx
- Nx.Ptree.tThe type for descriptions of structures 's whose arrays are on 'd.type · nx
- Nx.tThe type for arrays of 's-format elements read as 'v, on the device set 'd.type · nx
- Nx_array.Dtype.tThe type for dtypes.type · nx
- Nx_array.Layout.tThe type for layouts.type · nx
- Nx_array.Move.tThe type for movements.type · nx
- Nx_array.tThe type for arrays of elements of storage format 's read as 'v.type · nx
- Nx_io.Archive.tThe type for archives of values on the set 'd.type · nx
- Nx_io.Gguf.tThe type for the contents of a GGUF file.type · nx
- Nx_kernel.Prog.E.tThe type for expressions of dtype ('v, 's) Nx_array.Dtype.t.type · nx
- Nx_kernel.Prog.tThe type for programs.type · nx
- Nx_kernel.Shape.tThe type for shapes.type · nx
- Nx_quant.tThe type for quantised weights on the set 'd.type · nx
- Nx_ragged.tThe type for ragged arrays of values of type 'v stored as 's on the set 'd.type · nx
- Nx_wide.tThe type for arrays of double-word numbers of the float format 's on the set 'd.type · nx
- Quill.Cell.ttype · quill
- Quill.Doc.tThe type for notebook documents.type · quill
- Quill.Kernel.tThe type for kernels.type · quill
- Quill.Session.tThe type for notebook sessions.type · quill
- Quill_markdown.Edit.tA parsed markdown source split into blocks with byte ranges.type · quill
- Quill_project.tA project.type · quill
- Rig.Buffer.tThe type for buffers.type · rig
- Rig.Claim.tThe type for the claims of a with_.type · rig
- Rig.Image.tThe type for loaded images: a binary in a device's format, whose functions that device's work runs.type · rig
- Rig.Point.tThe type for points.type · rig
- Rig.Submission.Run.tThe type for runs: the storage of one submission's launch and fill blocks, outside the OCaml heap, for one submit at a time.type · rig
- Rig.Submission.tThe type for prepared submissions.type · rig
- Rig.tThe type for devices.type · rig
- Rig_amd.Lock.tThe type for a path's locks.type · rig
- Rig_amd.tThe type for open devices of the driver.type · rig
- Rig_amd_abi.Capability.tThe type for what compiled code needs from an AMD device.type · rig
- Rig_amd_abi.Code_object.tThe type for code objects: an ELF object for AMD GPUs, and what its relocations write over its image.type · rig
- Rig_amd_abi.Counter.tThe type for a counter as a run counts it.type · rig
- Rig_amd_abi.Gpu.tThe type for GPUs.type · rig
- Rig_bin.Line.ttype · rig
- Rig_bnxt.Qp.tThe type for reliable-connected queue pairs.type · rig
- Rig_bnxt.Region.tThe type for regions.type · rig
- Rig_bnxt.tThe type for NICs the process holds.type · rig
- Rig_cuda.tThe type for open devices: a GPU's primary context, two streams and a timeline word.type · rig
- Rig_cuda_abi.tThe type for what compiled code needs from a CUDA device.type · rig
- Rig_edge.Driver.tThe type for open devices of the driver.type · rig
- Rig_edge.Io.tThe type for open io devices.type · rig
- Rig_elf.tThe type for objects laid out in their image (Objects).type · rig
- Rig_metal.tThe type for open devices.type · rig
- Rig_metal_abi.tThe type for what compiled code needs from a Metal device.type · rig
- Rig_mlx5.Cq.tThe type for completion queues.type · rig
- Rig_mlx5.Qp.tThe type for reliable-connected queue pairs.type · rig
- Rig_mlx5.Region.tThe type for registered memory.type · rig
- Rig_mlx5.tThe type for open NIC ports.type · rig
- Rig_mlx5_abi.Completion.tThe type for completions.type · rig
- Rig_mlx5_abi.Entry.tThe type for work entries.type · rig
- Rig_nv.Lock.tThe type for a path's locks.type · rig
- Rig_nv.tThe type for open devices of the driver.type · rig
- Rig_nv_abi.Cubin.tThe type for cubins.type · rig
- Rig_nv_abi.Gpu.tThe type for GPUs.type · rig
- Rig_nv_abi.Launch.tThe type for kernels set up for launch on a GPU.type · rig
- Rig_nv_abi.Local_memory.tThe type for local memory.type · rig
- Rig_nv_abi.Qmd.tThe type for a launch's values, as its descriptor takes them.type · rig
- Rig_nv_abi.Template.tThe type for templates.type · rig
- Rig_pci.Driver.tThe type for a driver.type · rig
- Rig_pci.Function.tThe type for functions the process has taken.type · rig
- Rig_pci.Machine.tThe type for machines.type · rig
- Rig_pci.Memory.tThe type for the memory of one GPU.type · rig
- Rig_pci.Page_table.tThe type for the page tables and physical memory of one GPU.type · rig
- Rig_pci.Space.tThe type for virtual address spaces.type · rig
- Rig_pci.Window.tThe type for windows.type · rig
- Rig_program.tThe type for descriptions.type · rig
- Rig_remote.tThe type for jobs, seen from their controller.type · rig
- Rig_remote_abi.tThe type for the records of proxies.type · rig
- Rig_remote_proxy.Link.tThe type for links.type · rig
- Rig_remote_proxy.tThe type for proxies.type · rig
- Rune.Probe.tThe type for probes of 'a values.type · rune
- Rune.Total.tThe type for totals of ('v, 's, 'd) Nx.t values.type · rune
- Talon.Binary.tThe type for byte strings.type · talon
- Talon.Column.tThe type for columns.type · talon
- Talon.Error.tThe type for errors.type · talon
- Talon.Expr.tThe type for expressions of shape 's whose values read as 'a.type · talon
- Talon.Ext.tThe type for declarations of extensions whose values read as 'e and are stored as 's.type · talon
- Talon.Kind.tThe type for kinds whose cells read as OCaml values of type 'a.type · talon
- Talon.Order.tThe type for sort keys.type · talon
- Talon.Query.tThe type for queries.type · talon
- Talon.Record.tThe type for record cells.type · talon
- Talon.Schema.tThe type for schemas.type · talon
- Talon.Sel.tThe type for selectors.type · talon
- Talon.Source.Pred.tThe type for predicates.type · talon
- Talon.Source.tThe type for sources.type · talon
- Talon.tThe type for tables: named, typed columns of equal length.type · talon
- Talon.Time.Date.tThe type for dates.type · talon
- Talon.Time.Span.tThe type for spans.type · talon
- Talon.Type.tThe type for column types whose cells read as 'a.type · talon
- Tolk.Bigint.tThe type for integers.type · tolk
- Tolk.Compiler_worker.tThe type for workers.type · tolk
- Tolk.Dtype.tThe type for data types.type · tolk
- Tolk.Dtype.Value.tThe type for values.type · tolk
- Tolk.Helpers.Target.tThe type for targets.type · tolk
- Tolk.Op.Set.tThe type for sets of operations.type · tolk
- Tolk.Op.ttype · tolk
- Tolk.Ops.Axis_type.tThe type for axis types, in the order ranges nest: a kernel's ranges sort by axis type, then identity.type · tolk
- Tolk.Ops.Elementwise.ttype · tolk
- Tolk.Ops.Nodes.tThe type for sets of nodes.type · tolk
- Tolk.Ops.Pattern_matcher.tThe type for pattern matchers.type · tolk
- Tolk.Ops.tThe type for nodes.type · tolk
- Tolk.Ops.Upat.tThe type for patterns.type · tolk
- Tolk.Opt.tThe type for kernel optimisations.type · tolk
- Tolk.Postrange.Scheduler.tThe type for schedulers.type · tolk
- Tolk.Renderer.Compiler.tThe type for compilers.type · tolk
- Tolk.Renderer.Estimates.tThe type for estimates.type · tolk
- Tolk.Renderer.tThe type for renderers.type · tolk
- Tolk.Setting.tThe type for settings of type 'a.type · tolk
- Tolk.Shape.Sint.ttype · tolk
- Tolk.Spec.tThe type for specifications.type · tolk
- Tolk.Support_memory.Tlsf_allocator.tThe type for allocators.type · tolk
- Tolk.Tc.tThe type for tensor cores.type · tolk
- Tolk_engine.tThe type for linked schedules.type · tolk
- Vega.Adafactor_state.tThe type for values with parameter 'a and fixed arrays on 'd.type · vega
- Vega.Adagrad_state.tThe type for values with parameter 'a and fixed arrays on 'd.type · vega
- Vega.Adam_state.tThe type for values with parameter 'a and fixed arrays on 'd.type · vega
- Vega.Adan_state.tThe type for values with parameter 'a and fixed arrays on 'd.type · vega
- Vega.Lion_state.tThe type for values with parameter 'a and fixed arrays on 'd.type · vega
- Vega.Loss_scale.tThe type for loss scales on the set 'd: the current scale factor and the number of consecutive finite steps since it last changed, both scalar tensors.type · vega
- Vega.Rmsprop_state.tThe type for values with parameter 'a and fixed arrays on 'd.type · vega
- Vega.Schedule.tThe type for learning-rate schedules.type · vega
- Vega.Sgd_state.tThe type for values with parameter 'a and fixed arrays on 'd.type · vega
- Ymir.Cosmology.Structure.tThe type for values with parameter 'a and fixed arrays on 'd.type · ymir
- Ymir.Cosmology.tThe type for cosmologies with payloads 'p.type · ymir
- Ymir.Direction.tThe type for directions in frame 'f: float64 vectors [...; 3], each read as the ray it spans.type · ymir
- Ymir.Earth.tThe type for the Earth's orientation at a batch of instants.type · ymir
- Ymir.Frame.tThe type for celestial frames.type · ymir
- Ymir.Grid.tThe type for grids of cells whose world is 'w, with measures at dtype 'e.type · ymir
- Ymir.Healpix.tThe type for HEALPix pixelisations of frame 'f's sphere, an nside and a scheme.type · ymir
- Ymir.Observation.tThe type for observations on a grid whose world is 'w, at dtype 'e.type · ymir
- Ymir.Region.tThe type for regions placed from world 'w, sized at dtype 'e.type · ymir
- Ymir.Time.tThe type for times: batches of instants on the scale 's.type · ymir
- Ymir.Transform.tThe type for maps from points of type 'a to points of type 'b.type · ymir
- Ymir_fits.Fits.Header.tThe type for headers: records in file order, without END.type · ymir
- Ymir_fits.Fits.Image.tThe type for an image's description: shape, stored format, scaling and BLANK.type · ymir
- Ymir_fits.Fits.Table.tThe type for a table's description: rows, columns, heap size.type · ymir
- Ymir_fits.Fits.Value.tThe type for how a value field reads and prints as an 'a.type · ymir
- Ymir_spk.tThe type for SPK files' segments, their data on the device set 'd.type · ymir
- Ymir_units.Codata.tThe type for CODATA releases.type · ymir
- Ymir_units.Constant.tThe type for measured constants.type · ymir
- Ymir_units.Quantity.Structure.tThe type for values with parameter 'a and fixed arrays on 'd.type · ymir
- Ymir_units.Quantity.tThe type for values 'p in a unit.type · ymir
- Ymir_units.Unit.tThe type for units.type · ymir
- Ymir_units.Vocabulary.tThe type for vocabularies.type · ymir
- Hugin_kit.Scheme.tabletable s is the table of s (Continuous reading), from the colour of 0 to the colour of 1, as a fresh array.val · hugin
- Ymir_fits.Fits.TableBinary and ASCII tables.module · ymir
- Jera.Ode.tableautableau ~a ~b ~c is the explicit Runge–Kutta method of Butcher tableau (a, b, c) with s stages: stage i evaluates the field at t + c.(i) h on y + h Σ_(j<i) a.(i).(j) k_j, and the step is y + h Σ_i b.(i) k_i.val · jera
- Hugin_kit.Scheme.tableau10tableau10 is the ten colours of Tableau 10 by Maureen Stone: #4e79a7, #f28e2c, #e15759, #76b7b2, #59a14f, #edc949, #af7aa1, #ff9da7, #9c755f and #bab0ab.val · hugin
- Rig_pci.Page_table.tablesThe type for where page tables come from.type · rig
- Rig_pci.Page_table.tablestables t ~va n is the physical addresses of the tables from the root down to the one whose entries would map the n bytes from va, root first.val · rig
- Hugin_vg.Picture.tagThe type for tags.type · hugin
- Hugin_vg.Picture.tagtag t p paints p tagged with t.val · hugin
- Munin.Run.tagstags r is r's tags as read, sorted.val · munin
- Ymir.Time.taitype · ymir
- Ymir.Time.taitai is International Atomic Time.val · ymir
- Talon.Kit.tailtail n q is the last n rows of q:val · talon
- Nx.taketake ~axis p x replaces x's axis axis by p's axes, reading x at each position p holds; without axis, flatten x is read.val · nx
- Nx_quant.taketake ~axis p w is the rows or matrices of w at the positions p along axis, gathered from its packed parts with no value decoded, p read as Nx.take reads it: at each position in range, dequant dt (take ~axis p w) is Nx.take ~axis p (dequant dt w).val · nx
- Nx_ragged.taketake ~indices r is the ragged array whose row j is row indices.{j} of r.val · nx
- Rig.Profile.taketake f is f () and the events recorded while it ran, in time order and, at equal times, longest first, then in the order they were recorded.val · rig
- Rig.Submission.taketake s is a run of s that the calling thread has taken for its stores and its next submit: the first of s's runs that is free, or else a new one, whose bytes read 0.val · rig
- Rig_pci.Function.taketake m bus takes the function at bus on m for this process alone.val · rig
- Talon.taketake indices t is the rows of t at indices, in order, as one batch: take (Nx.Rng.permutation ~key (rows t)) t shuffles every column.val · talon
- Nx.take_along_axistake_along_axis ~axis p x has p's shape broadcast with x's off axis; its element at j is x's element at j with axis axis replaced by p's element at j.val · nx
- Tolk.Ops.take_uniquetake_unique n makes every number unique_num returns afterwards greater than n, up to max_int, past which no number is left.val · tolk
- Tolk.Memory.takentaken si is the storage that the entry si of a schedule takes, each once, in the order its calls take it: the buffer, call-local storage or parameter (Op.t.Buffer, Op.t.Alloc, Op.t.Param) that each argument of its calls names or views (Ops.storage_base).val · tolk
- Hugin.Theme.talktalk is v ~size:16.val · hugin
- talonTalon is a library of tables: named, typed columns of equal length.library · talon
- TalonTables: named, typed columns.module · talon
- Talon_arrowArrow IPC files and streams.module · talon
- Talon_csvCSV files.module · talon
- Talon_jsonJSON and NDJSON files.module · talon
- Talon_parquetParquet files, read and written.module · talon
- Nx.tantan x is the tangent of x radians.val · nx
- Nx_kernel.Prog.E.tanval · nx
- Nx.tanhtanh x is the hyperbolic tangent.val · nx
- Nx_kernel.Prog.E.tanhval · nx
- Jera.Quad.tanh_sinhtanh_sinh ~tol f range is the integral of f over each element of range by a double-exponential rule: tanh-sinh on a finite range, exp-sinh on Range.from and sinh-sinh on Range.line.val · jera
- Hugin_vg_vector.Vector.targetThe type for targets: what a renderer writes for each case of a picture, into a 'b, a stamp's definition being a 'd.type · hugin
- Rig_amd_abi.Code_object.targettarget co is the processor co is compiled for, as LLVM names it: a GPU, such as "gfx1201", or a generic processor, such as "gfx12-generic".val · rig
- Rig_pci.Page_table.targetThe type for the memory an entry points into.type · rig
- Tolk.Helpers.TargetCompilation targets.module · tolk
- Tolk.Opt.targetThe type for the roles a split gives the axis it makes, in the order of their axis types (Ops.Axis_type.compare): an axis across the threads of a workgroup, or an axis unrolled into the lanes of a thread, which a reduction's axis unrolls as its loop.type · tolk
- Tolk_engine.targettarget d is what d's programs are compiled for, a function of d alone.val · tolk
- Ymir_units.Codata.tau_masstau_mass r is mτ in kg.val · ymir
- Tolk.TcTensor cores.module · tolk
- Tolk.Setting.tc_min_globalstc_min_globals is the number of global axes below which tensor core optimization does not upcast its N axis, from TC_MIN_GLOBALS.val · tolk
- Tolk.Setting.tc_opttc_opt is how hand-coded optimizations (Heuristic) fit a kernel to a tensor core, from TC_OPT: 2 pads the kernel's axes to the core's dimensions, 0 and 1 take only kernels those dimensions divide.val · tolk
- Tolk.Setting.tc_selecttc_select is the tensor core kernel optimization uses, from TC_SELECT: -1 tries the target's tensor cores in order and uses the first that fits, n uses only the n-th.val · tolk
- Ymir.Time.tdbtype · ymir
- Ymir.Time.tdbtdb is Barycentric Dynamical Time at the geocentre.val · ymir
- Brot.Post_processor.templatetemplate ~single () is a template-based post-processor.val · brot
- Rig_nv_abi.TemplateBytes around the caller's values.module · rig
- Quill_book.Theme.template_htmltemplate_html is the page of a chapter, with a sidebar, search and page navigation.val · quill
- Talon.Expr.TemporalTemporal values.module · talon
- Nx.Ptree.tensortensor is one array, at the root.val · nx
- Nx.Ptree.Walk.tensortensor c x walks x, a fixed array, at c's path.val · nx
- Nx_io.Archive.tensortensor ~shape dt name a is a's entry name, as stored.val · nx
- Talon.Kind.tensortensor dt reads tensor columns of dtype dt and any cell shape.val · talon
- Talon.Type.tensortensor dt shape is Tensor of dt with a copy of shape.val · talon
- Tolk.Spec.tensortensor is shared with the operations of tensor graphs: float-only unary math; global buffers with a size and a device, and storage declared without a buffer (Op.t.Alloc); variables without a device; unlowered Op.t.Specials; movement, the shape of a reshape or an expand being integers; reductions; copies to a device other than a disk; Op.t.Detach, Op.t.Contiguous_backward and Op.t.Stage; programs as compilation fills them in: a sink, then its linear form, its source and its binary; and calls of opaque bodies wherever their arguments start, since scheduling decides what each becomes.val · tolk
- Nx_io.Gguf.tensor_infoThe type for an array's description in the file.type · nx
- Nx.tensordottensordot ~axes:(la, lb) a b is the sum of the products of a and b over the axes la of a, paired in order with the axes lb of b (default a's last two with b's first two).val · nx
- Nx_io.Gguf.tensorstensors g is the archive of g's arrays.val · nx
- Ymir_units.Unit.teratera u is 1012 u.val · ymir
- Ymir_units.Unit.termThe type for terms.type · ymir
- Norn_model.termsterms m y c is each site's term of log_density m y at the position c, of shape [chain], by name in the order the model performs them, then the factors' sum, zeros when there is none.val · norn
- Ymir_units.Unit.termsterms u is u's terms, each (t, num, den) with t's exponent num/den reduced, den >= 1 and num <> 0.val · ymir
- Tolk.Op.Set.ternaryternary is the operations of three operands: Where and Mulacc.val · tolk
- Ymir_units.Unit.teslatesla is kg s-2 A-1.val · ymir
- Fehu.Space.Textmodule · fehu
- Hugin.Mark.texttext ~halign ~valign rows c pt s draws s set in the faces of the theme at its base size, aligned about the point pt of the page by halign and valign (Alignment), `Center and `Middle by default, its characters without a colour of their own painted with c.val · hugin
- Hugin.Role.texttext is a text.val · hugin
- Hugin.texttext ~x ~y ~text () sets each row's text in the theme's faces at its base size, centred on the row's point moved dx points right and dy points up, both 0.val · hugin
- Hugin_font.Run.texttext r is the text r renders.val · hugin
- Hugin_text.TextRich text.module · hugin
- Quill.Cell.texttext ?id ?attrs source is a text cell with the given source.val · quill
- Ymir_fits.Fits.Value.texttext is the value field as written, trimmed: a string's text keeps its quotes.val · ymir
- The dashboardmunin without a command, on a terminal, shows the store's runs.guide · munin
- The munin commandmunin reads a store and changes what may change in it: a run's tags and notes, an artifact's aliases, and what is deleted.guide · munin
- The Tokenization PipelineBrot processes text through up to 6 stages, each optional and independently configurable:guide · brot
- Hugin.Mark.themetheme rows is the theme of the layout being drawn.val · hugin
- Hugin.ThemeThemes.module · hugin
- Quill_book.ThemeThe default book theme: its stylesheet and HTML page templates.module · quill
- Norn.Draws.thinthin u ~every d is every every-th draw of each chain of d, from the first.val · norn
- Rig_pci.Machine.thisthis is the machine this process runs on.val · rig
- Ymir_units.Codata.thomson_cross_sectionthomson_cross_section r is σe in m2.val · ymir
- Rig_amd_abi.Thread_traceThread traces, which a GPU's shader engines record while they run waves.module · rig
- Tolk.Postrange.Scheduler.threadlessthreadless k is k's ranges that take no threads: those unread_by_opaque, and the weak ranges whose steps share storage with the code outside them (Rewriting).val · tolk
- Rig.Image.threadsthreads fn is the most threads a group of fn may have, or None if its driver does not know it.val · rig
- Rig.Submission.Run.threadsthreads r b x y z sets each group of b to x by y by z threads.val · rig
- Tolk.Tc.threadsthreads tc is the number of threads of the warp: 2 to the number of lanes.val · tolk
- Tolk.Postrange.threads_fixedthreads_fixed ast is true iff the builder of the kernel ast, or of a body it calls, fixed its threads: it holds workgroup or register storage (Dtype.addr_space.Local, Dtype.addr_space.Reg), a barrier, a matrix multiply-accumulate or a call of code in the target's language.val · tolk
- Nx_kernel.Prog.E.threefryval · nx
- Tolk.Decomp_op.threefry2x32threefry2x32 x key is the Threefry-2x32 hash of the Dtype.t.Uint64 counter x under key, with 20 rounds, as 32-bit additions, rotations and exclusive ors: the two 32-bit words of the result, high word second, in a Dtype.t.Uint64.val · tolk
- Sowilo.thresholdthreshold t img is true where img's element is greater than t, and false elsewhere, NaN elements included.val · sowilo
- Rig_pci.Window.throughthrough tr a n is the n bytes at a of the machine tr reaches.val · rig
- Hugin_kit.Ticks.tickThe type for labelled ticks.type · hugin
- Hugin.Mark.ticksticks rows r is Some us, the normalised positions of the major ticks that layout froze for the scale r reads, in increasing order, whether or not a guide shows them, or None if r is unbound, bound to a constant or reads no scale.val · hugin
- Hugin_kit.Scale.ticksticks s is the guide values of s, if set, as a fresh array.val · hugin
- Hugin_kit.TicksAxis ticks.module · hugin
- Nx.tiletile reps x is x repeated reps.(i) times, end to end, along its axis i.val · nx
- Tolk.Stage.tiletile ren sink core ~core_threads ~workgroups ~sizes:(n, m, k) ~in_run is the splits of a tensor core's staged tile, as (axis, amount, target) with axis 0, 1 or 2 for the core's N, M and K, in the order to apply them, and whether the core has two stages and the kernel keeps core_threads / 2 threads on some tile, which an unknown count of workgroups does not show, when admits and depth stage the core; None otherwise.val · tolk
- Hugin.timetime ~scale ~valid ~title r counts is the instants of the integer counts, one datum per element, the element n being Time.v r n, n units of r after the epoch, with scale, valid and title as for num.val · hugin
- Hugin_kit.Scale.timetime () is a temporal scale whose ticks and nice domains follow the calendar at the offset tz_offset_s (see Time.tz_offset_s); unset counts as 0.val · hugin
- Hugin_kit.TimeInstants and the calendar.module · hugin
- Hugin_kit.Time.timeThe type for times of day (hh, mm, ss): the hour hh in [0;23], and the minute mm and the second ss in [0;59].type · hugin
- Jera.Ode.timeThe type for times.type · jera
- Munin.Metric.timetime s i is the time of s's ith sample.val · munin
- Talon.TimeCivil time values: stamps, spans and dates.module · talon
- Tolk_engine.timetime ~vars s slots is the time in seconds of one run of s on slots with vars (run), reporting nothing on standard output whatever Tolk.Setting.debug holds.val · tolk
- Ymir.TimeInstants on astronomical time scales.module · ymir
- Fehu.Env.time_limittime_limit ~max_episode_steps env enforces a maximum episode length.val · fehu
- Talon.Kit.time_rangetime_range ty ~start ~stop step name is the table of one column name of ty, the values offset (k * step) start below stop, in order, for k from 0, so that adjacent ranges meet at stop.val · talon
- Tolk.Helpers.time_to_strtime_to_str ~w t is the duration of t seconds right-aligned in w columns with two decimals, followed by a two-column unit: s above 10 seconds, ms above 10 milliseconds, and us otherwise.val · tolk
- Hugin_kit.Symbol.timestimes is plus turned by an eighth of a turn.val · hugin
- Ymir_units.Quantity.timestimes u q is q's payload in Unit.(unit q * u).val · ymir
- Tolk.Device.Tiny_elfA compiled program's parameters, as a runtime packs them.module · tolk
- Ymir.Frame.tirstype · ymir
- Ymir.Frame.tirstirs is the Terrestrial Intermediate Reference System: CIRS turned about the CIP by the Earth rotation angle (IERS Conventions 2010, 5.4.2), the Earth's frame less polar motion.val · ymir
- Hugin.titletitle ~align s f is f with the title s above it, set bold in the theme's faces at 1.2 em (Derived lengths).val · hugin
- Hugin.title'title' ~align t f is title with the rich text t: title s f is title' (Text.v s) f.val · hugin
- Quill_project.title_of_filenametitle_of_filename "01-intro.md" is "Intro".val · quill
- Tolk.Support_memory.Tlsf_allocatorTwo-level segregated fit allocators.module · tolk
- Nx.to_arrayto_array x is x's elements in C order.val · nx
- Nx_array.to_arrayto_array a is a's elements in C order of indices.val · nx
- Nx.to_bigarrayto_bigarray k x is a fresh bigarray of x's shape holding its elements.val · nx
- Nx_array.Dtype.to_bigarray_kindto_bigarray_kind dt is the kind of dt's elements, if Bigarray has their format: None for Bfloat16, the float8 and float4 dtypes, Int4, Uint4, Uint32, Uint64, Bool and Bit.val · nx
- Munin.Value.to_boolto_bool v is b for `Bool b, and None otherwise.val · munin
- Tolk.Dtype.Value.to_boolto_bool v is true iff v is not zero: a NaN is true.val · tolk
- Tolk.Shape.to_boolto_bool u is the value of the boolean node u.val · tolk
- Talon.Time.Date.to_civilto_civil t is (y, m, d), the year, month and day of t.val · talon
- Hugin_kit.Time.to_date_timeto_date_time ~tz_offset_s t is the civil date and time of t at tz_offset_s, its nanoseconds t.nsec dropped.val · hugin
- Talon.Time.Date.to_daysto_days t is the number of days from 1970-01-01 to t.val · talon
- Talon.to_denseto_dense dt ~axes value t is t's column value scattered into a tensor, and its validity.val · talon
- Nx_array.to_deviceto_device d a is a's elements on d, also when a is on d, in a fresh buffer of at most Dtype.bytes of one element per index of a, an axis of stride 0 counted once.val · nx
- Munin.Value.to_floatto_float v is f for `Float f, the float nearest n for `Int n, and None otherwise.val · munin
- Sowilo.to_floatto_float img is img cast to float32 and scaled to [0, 1] by dividing by 255.val · sowilo
- Tolk.Bigint.to_floatto_float n is the float nearest to n, ties to even: an infinity if n is past the greatest float.val · tolk
- Tolk.Dtype.Value.to_floatto_float v is v as a float: 0.val · tolk
- Tolk.Shape.to_floatto_float u is the value of the float node u, as to_bool.val · tolk
- Tolk.Helpers.to_function_nameto_function_name s is ansistrip s with each character other than an ASCII letter, digit or underscore replaced by its Unicode code point in uppercase hexadecimal, at least two digits.val · tolk
- Sowilo.to_grayscaleto_grayscale img converts RGB to single-channel grayscale using ITU-R BT.601 weights: 0.299 * R + 0.587 * G + 0.114 * B.val · sowilo
- Hugin_gg.Color.to_hexto_hex c is c as #rrggbbaa in lowercase, each pair the component times 255.val · hugin
- Quill_markdown.Edit.to_htmlto_html source renders CommonMark source to an HTML fragment.val · quill
- Fehu.Space.Discrete.to_intto_int e is the integer value of the discrete element e.val · fehu
- Munin.Value.to_intto_int v is n for `Int n, and None otherwise.val · munin
- Tolk.Bigint.to_intto_int n is n.val · tolk
- Tolk.Dtype.Value.to_intto_int v is to_z v as an int.val · tolk
- Tolk.Op.to_intto_int o is the position of o in the declaration order of t, counting from 0.val · tolk
- Hugin_kit.Time.to_int64to_int64 r t is Some n for the greatest n such that v r n is not after t, and None if that n does not fit in an int64.val · hugin
- Tolk.Bigint.to_int64to_int64 n is n.val · tolk
- Brot.Decoder.to_jsonto_json decoder serializes decoder to HuggingFace JSON format.val · brot
- Brot.Normalizer.to_jsonto_json n is n serialized to HuggingFace-compatible JSON.val · brot
- Brot.Post_processor.to_jsonto_json t is t serialized to HuggingFace tokenizer.json format.val · brot
- Brot.Pre_tokenizer.to_jsonto_json t serializes t to HuggingFace JSON format.val · brot
- Brot.to_jsonto_json t is t serialized to HuggingFace JSON format.val · brot
- Fehu.Info.to_listto_list t is the bindings of t in key order.val · fehu
- Nx.Devices.to_listto_list s is s's devices in order.val · nx
- Tolk.Op.Set.to_listto_list s is the operations of s in declaration order.val · tolk
- Tolk.Ops.Nodes.to_listto_list s is the nodes of s in the order they joined it.val · tolk
- Tolk.Ops.to_max_shapeto_max_shape s is s with each symbolic size replaced by its greatest value.val · tolk
- Talon.Time.Span.to_msto_ms d is like to_us in milliseconds.val · talon
- Talon.Time.to_msto_ms t is like to_us in milliseconds.val · talon
- Talon.Time.Span.to_nsto_ns d is d in nanoseconds.val · talon
- Talon.Time.to_nsto_ns t is the number of nanoseconds from 1970-01-01 00:00:00 of t's clock to t.val · talon
- Hugin_gg.Color.to_oklabto_oklab c is (l, a, b), the Oklab coordinates of c: the lightness l from 0.val · hugin
- Hugin_gg.Color.to_oklchto_oklch c is (l, c, h) for the Oklab coordinates (l, a, b) of c: the lightness l, the chroma c = Float.hypot a b and the hue h = Float.atan2 b a in radians, in [0;2π[.val · hugin
- Hugin_gg_kit.Pgon2.to_pathto_path p is the subpaths Ring2.to_path gives for the rings of p, in order.val · hugin
- Hugin_gg_kit.Ring2.to_pathto_path r is the closed subpath through the points of r in order, Hugin_gg.Path.polygon of its coordinates.val · hugin
- Tolk.Codegen.to_programto_program ~beam ast ren is the program of ast for ren: an Op.t.Program whose argument is its Ops.program_info for ren's target, and whose sources are:val · tolk
- Tolk.Codegen.to_programsto_programs ~beam kernels makes the program of each kernel (ast, ren) of kernels that read, run on the calling domain, does not find, as to_program ~beam ast ren does and keeps it, so that a later to_program returns it; kernels whose programs are all found spawn no domain.val · tolk
- Talon.Time.Span.to_sto_s d is like to_us in seconds.val · talon
- Talon.Time.to_sto_s t is like to_us in seconds.val · talon
- Tolk.Dtype.to_storage_scalarto_storage_scalar dt v is the value that stores v in storage_fmt: v rounded to Float16, the `Int of the bits that encode v for Bfloat16 and the 8-bit floats, and v itself otherwise.val · tolk
- Fehu.Value.to_stringto_string v is v formatted as a string via pp.val · fehu
- Hugin_kit.Number.to_stringto_string ~locale f x is x written in f with the strings of locale, which defaults to Locale.default, by the rules of How numbers are written.val · hugin
- Munin.Value.to_stringto_string v is s for `String s, and None otherwise.val · munin
- Nx.Ptree.Path.to_stringto_string p is p's steps joined by ".", as "blocks.3.fc.b": the name of the part at p in a saved value.val · nx
- Nx.to_stringto_string x is x printed by pp.val · nx
- Quill_markdown.to_stringto_string doc renders doc as a markdown string.val · quill
- Rig_bin.Line.to_stringto_string l is l's line, its newline included.val · rig
- Rig_program.to_stringto_string t is t as bytes, in a format with a version of its own.val · rig
- Tolk.Bigint.to_stringto_string n is n in decimal, with a - if negative.val · tolk
- Tolk.Graph.to_stringto_string sink is the graph under sink.val · tolk
- Ymir_fits.Fits.Header.to_stringto_string h is h's records, END, and spaces to a multiple of 2880 bytes.val · ymir
- Ymir_units.Unit.to_stringto_string u is u's canonical text.val · ymir
- Quill_markdown.to_string_with_outputsto_string_with_outputs doc renders doc as markdown with outputs.val · quill
- Nx_ragged.to_stringsto_strings r is the bytes of each row of r as a string: to_strings (of_strings ss) is ss.val · nx
- Nx.Rng.to_tensorto_tensor k is k's words: the inverse of of_tensor, and what the coercion of k to an array is.val · nx
- Talon.Column.to_tensorto_tensor dt c is c's values as stored, in O(1) after one read of the null count the first time: numbers, booleans as Nx.bit, the days or ticks of temporal values, the codes of a categorical (its dictionary is in its type), and (rows, …shape) for a tensor column.val · talon
- Talon.to_tensorto_tensor dt names t is the (rows t, List.length names) matrix whose column j is the column List.nth names j of t, each value converted to dt as Nx.cast converts it.val · talon
- Sowilo.to_uint8to_uint8 img is img scaled from [0, 1] to [0, 255] and cast to uint8.val · sowilo
- Talon.Time.Span.to_usto_us d is d in whole microseconds, rounded toward negative infinity: to_us (ns (-1)) is -1L.val · talon
- Talon.Time.to_usto_us t is the number of whole microseconds from 1970-01-01 00:00:00 of t's clock to t, rounded toward negative infinity: the stamp one nanosecond before the epoch gives -1L.val · talon
- Fehu.Info.to_valueto_value t is t as a Value.t.Dict.val · fehu
- Nx_io.Archive.to_valueto_value ~prefix s ~like a is like with each array replaced by a's entry of its name under prefix, placed where like's array lies (Nx.place).val · nx
- Tolk.Dtype.Value.to_zto_z v is v as an integer: 0 or 1 for a `Bool, and a float rounded towards zero.val · tolk
- Tolk.Shape.to_zto_z u is the value of the integer node u, as to_bool.val · tolk
- Quill_book.Render.toc_htmltoc_html project ~current ~root_path renders the sidebar table of contents with current highlighted as the active notebook.val · quill
- Quill_project.toc_itemtype · quill
- Brot.Post_processor.tokenA special token as (text, id).type · brot
- Brot.token_to_idtoken_to_id t token is the ID of token in t, if any.val · brot
- Brot.Encoding.tokenstokens enc is the string representation of each token.val · brot
- Jera.TolTolerances.module · jera
- tolkTolk is a port of tinygrad in OCaml — a minimal compiler for GPU tensor computation.library · tolk
- Tolkmodule · tolk
- Tolk_engineRunning compiled schedules on rig's devices.module · tolk
- Ymir_units.Unit.tonnetonne is 103 kg.val · ymir
- Nx.top_ktop_k ~k ~axis x is the first k entries along axis (default -1) of sort ~descending:true ~axis x: the greatest first, equal ones lowest position first, a NaN before every number.val · nx
- Talon.Kit.top_ktop_k k keys q is the first k rows of q in the order of keys:val · talon
- Kaun.Metric.top_k_accuracytop_k_accuracy ~k predictions labels is the fraction of examples whose label ranks among the k highest-scoring classes.val · kaun
- Tolk.Ops.toposorttoposort ~gate ~calls u is u and the nodes it reaches, each after its sources, in the order a depth-first walk of the sources, in order, finishes them.val · tolk
- Tolk.Ops.topovisittopovisit u f cache is f u, after f has been applied to each node u reaches, each after its sources, and stored in cache.val · tolk
- Rune.TotalWrite-only sums.module · rune
- TotalsA total is a write-only sum.guide · rune
- Ymir.Transform.tpvtpv ~stated pv maps intermediate coordinates (ξ, η), in an angle, to (Σ pv.(0).(k) t_k (ξ, η), Σ pv.(1).(k) t_k (η, ξ)) in degrees, over TPV's forty terms t_k: 1, x, y, r, then the monomials of each degree from 2 to 7 in decreasing powers of x, with r³, r⁵ and r⁷ after degrees 3, 5 and 7, r = √(x² + y²).val · ymir
- Nx.tracetrace ~offset x is the sum of diagonal ~offset x along its last axis: one value per matrix of the batch, +0 for an empty diagonal.val · nx
- Rig_amd_abi.Capability.traceThe type for the device's thread trace buffers, one set per GPU, which every program that traces writes (Thread_trace).type · rig
- TracingA function given to rune is ordinary OCaml code.guide · rune
- Tracking runsA run is one execution of a program.guide · munin
- Tolk.Bigint.trailing_zerostrailing_zeros n is the greatest k for which 2{^k} divides n, or max_int if n is zero.val · tolk
- Brot.train_bpetrain_bpe data trains a BPE tokenizer from data.val · brot
- Brot.train_unigramtrain_unigram data trains a Unigram tokenizer from data.val · brot
- Brot.train_word_leveltrain_word_level data trains a word-level tokenizer from data.val · brot
- Brot.train_wordpiecetrain_wordpiece data trains a WordPiece tokenizer from data.val · brot
- TrainingKaun has no trainer.guide · kaun
- Tolk.Setting.transcendentaltranscendental is how code generation decomposes transcendental functions into polynomial approximations, from TRANSCENDENTAL: from 2 on, all of them; below, those the target does not support.val · tolk
- Tolk.TranscendentalTranscendental functions as arithmetic.module · tolk
- Nx_array.transfertransfer ~src ~dst stores src's elements into dst, of the same shape.val · nx
- Rig_remote_abi.transferThe type for transfers: the same bytes at the same places in every run.type · rig
- Hugin_gg.Box2.transformtransform m b is the smallest box containing the images of b's four corners under m.val · hugin
- Hugin_gg.P2.transformtransform m p is the image of p under m.val · hugin
- Hugin_gg.Path.transformtransform m p is p with every point mapped through m.val · hugin
- Hugin_kit.Scale.transformThe type for the transforms of quantitative scales (Continuous scales).type · hugin
- Hugin_kit.Scale.transformtransform s is the transform of s.val · hugin
- Hugin_vg.Picture.transformtransform m p paints p with the plane mapped through m: what p paints at pt, transform m p paints at P2.transform m pt.val · hugin
- Norn.Dist.transformtransform b d is the distribution of b's image of a draw of d, with the density of d at the preimage less b's log-determinant there.val · norn
- Ymir.TransformMaps between planes and the sky.module · ymir
- Fehu.Buffer.transitionThe type for transitions.type · fehu
- Hugin_gg.Affine.translatetranslate dx dy moves points by (dx, dy).val · hugin
- Hugin_gg.Color.transparenttransparent is v ~alpha:0.val · hugin
- Rig_pci.Window.transportThe type for transports.type · rig
- Nx.transposetranspose ~axes x has x's axis List.nth axes i as its axis i; without axes, x's axes in reverse order.val · nx
- Nx.transposeThe type for the matrix a solve applies: aᵀ, or aᴴ.type · nx
- Tolk.Shape.transposetranspose u a b exchanges the axes a and b of u.val · tolk
- Tolk.Tc.transposetranspose tc is the core of Dᵀ = Bᵀ Aᵀ + Cᵀ, which a program computes by giving tc its operands in the other order: its A is tc's B and its B is tc's A, each with its M and N bits exchanged, and its C is tc's transposed.val · tolk
- Ymir.Cosmology.transversetransverse c d is the transverse comoving distance across a radial comoving separation d, in d's unit: d S(Ωk (d/DH)2), with S(x) = Σn xn/(2n+1)!, which is sinh √x/√x for x > 0 and sin √-x/√-x for x < 0.val · ymir
- Hugin_kit.Symbol.triangletriangle is an equilateral triangle pointing up.val · hugin
- Nx.triangleThe type for a square matrix's triangles, each with its diagonal: the elements on and below it, or on and above it.type · nx
- Nx_kernel.triangleThe type for a square matrix's triangles, each with its diagonal.type · nx
- Nx.trig_typeThe type for the four discrete cosine and sine transforms, I to IV.type · nx
- Nx.triltril ~k x is x where column - row <= k (default 0) and zero above, in each matrix of the batch.val · nx
- Talon.Expr.Str.trimtrim a is a without the white space at its start and end.val · talon
- Nx.Ptree.tripletriple a b c walks a triple's components with a, b and c at Index 0, Index 1 and Index 2, in that order.val · nx
- Ymir_units.Codata.triton_masstriton_mass r is mt in kg.val · ymir
- Nx.triutriu ~k x is x where column - row >= k (default 0) and zero below, as tril.val · nx
- Ymir.Frame.true_of_datetype · ymir
- Ymir.Frame.true_of_datetrue_of_date is the true equator and equinox of date: from GCRS by the IAU 2006/2000A bias, precession and nutation, the classical, equinox based form of cirs's pole (IERS Conventions 2010, 5.4.4).val · ymir
- Nx.trunctrunc x rounds toward zero.val · nx
- Nx_kernel.Prog.E.truncval · nx
- Tolk.Ops.Elementwise.truncval · tolk
- Tolk.Ops.truncval · tolk
- Tolk.Ops.Upat.truncval · tolk
- Brot.Encoding.truncatetruncate enc ~max_length limits enc to at most max_length tokens.val · brot
- Tolk.Dtype.truncatetruncate dt v is v as a value of dt:val · tolk
- Norn.Dist.truncatedtruncated ?low ?high d is d conditioned on lying between low and high, elementwise, a missing bound leaving its side as d's: the density of d over its mass F(high) - F(low) there, -inf outside.val · norn
- Nx.Rng.truncated_normaltruncated_normal lower upper is standard normal draws conditioned on [lower, upper], elementwise over their broadcast shape.val · nx
- Brot.truncationThe type for truncation configurations.type · brot
- Brot.truncationtruncation max_length is a truncation configuration limiting sequences to max_length tokens.val · brot
- Tolk.Shape.Sint.truthtruth c is c's value if known, and Shape.to_bool's otherwise.val · tolk
- Rig_nv.Lock.try_protecttry_protect l f is true once f () ran as in protect, or false, having run nothing, if another thread holds l or the calling thread holds a path's lock.val · rig
- Tolk.Heuristic_common.try_splittry_split k axis amount target is whether the split applied.val · tolk
- Jera.Ode.tsit5tsit5 is Tsitouras' (2011) method of order 5 with an embedded order 4: six evaluations per step, its last stage the next step's first.val · jera
- Ymir.Earth.tttt e is e's instants on TT.val · ymir
- Ymir.Time.tttype · ymir
- Ymir.Time.tttt is Terrestrial Time, TAI + 32.184 s.val · ymir
- Fehu.Space.Tuplemodule · fehu
- Tolk.Setting.tuple_ordertuple_order makes linearization order nodes of equal priority by their structure, from TUPLE_ORDER.val · tolk
- Hugin_kit.Scheme.turboturbo is a rainbow from dark blue through cyan, green, yellow and red to dark red (Mikhailov), with its published table of 256 colours rounded as those above.val · hugin
- Talon.TypeColumn types.module · talon
- Talon.Column.type_type_ c is the type of c's values.val · talon
- Quill.Kernel.type_attype_at k c ~pos is the type at byte pos of c's source, if any.val · quill
- Brot.Encoding.type_idstype_ids enc is the segment ID array.val · brot
- Quill.Kernel.type_infoThe type for type-at-position results.type · quill
- Tolk.Spec.type_verifytype_verify ~calls spec u checks every node of u's graph against spec, sources first (Ops.toposort ~calls u).val · tolk
- Talon.TzTime zones, decoded from TZif files.module · talon
- Hugin_kit.Scale.tz_offset_stz_offset_s s is the offset of the calendar of s, 0 if unset.val · hugin
- Hugin_kit.Time.tz_offset_sThe type for offsets from UTC, in seconds east of UTC: 3600 is UTC+01:00 and -18000 is UTC−05:00.type · hugin
u
- Rig_nv_abi.Template.u32u32 s v is a 4-byte word: v shifted right by s, in [0;63], then its low 32 bits.val · rig
- Rig_nv_abi.Template.u64u64 n v is an 8-byte word: v + n, modulo 2{^64}.val · rig
- Rig_mlx5_abi.UarAn access region (UAR) is a page of the NIC through which a process rings doorbells.module · rig
- Nx.uint16val · nx
- Talon.Type.uint16uint16 is Uint16.val · talon
- Nx.uint16_ttype · nx
- Nx.uint32val · nx
- Talon.Type.uint32uint32 is Uint32.val · talon
- Nx.uint32_elttype · nx
- Nx_array.Dtype.uint32_eltUnsigned 32-bit integers.type · nx
- Nx.uint32_ttype · nx
- Nx.uint4val · nx
- Nx.uint4_elttype · nx
- Nx_array.Dtype.uint4_eltUnsigned 4-bit integers.type · nx
- Nx.uint4_ttype · nx
- Nx.uint64val · nx
- Talon.Type.uint64uint64 is Uint64.val · talon
- Nx.uint64_elttype · nx
- Nx_array.Dtype.uint64_eltUnsigned 64-bit integers.type · nx
- Nx.uint64_ttype · nx
- Nx.uint8val · nx
- Talon.Type.uint8uint8 is Uint8.val · talon
- Nx.uint8_ttype · nx
- Tolk.Dtype.uintsuints is the unsigned integers, from Uint8 to Uint64.val · tolk
- Jera.Tol.ulpsulps k is rel (k *.val · jera
- Nx_kernel.Prog.unaryThe type for kinds of one operand, of its dtype.type · nx
- Tolk.Op.Set.unaryunary is the operations of one operand: Exp2, Log2, Sin, Sqrt, Recip, Neg and Trunc.val · tolk
- Tolk.Shape.unbindunbind v is (unbound v, x) for v bound to x.val · tolk
- Tolk.Shape.unbind_allunbind_all u is u with each bound variable unbound, and each variable with its value.val · tolk
- Rig_edge_primitives.unblockunblock () gives the calling thread back the mask it had before the outermost mask, right before a wait's blocking step (rig_unblock).val · rig
- Rig.Buffer.unborrowunborrow ~why b tries to end the borrow b early, and is false, changing nothing, if a claim or a hold through the borrow or one of its views keeps it, such as a host access or a submit through it on another domain, or a prepared submission that names it (Claim.hold).val · rig
- Tolk.Shape.unboundunbound v is v without its value and tag.val · tolk
- Tolk.Ops.unbound_outputsunbound_outputs c is those outputs, each ordered after c.val · tolk
- Ymir_units.Constant.uncertaintyuncertainty d k is k's standard uncertainty, which is never negative, as a scalar of d in unit k: zero for an exact constant, otherwise rounded once as quantity rounds.val · ymir
- Tolk.Shape.unchangedunchanged src u is true iff src is u's sources, node for node.val · tolk
- Norn_model.unconstrainunconstrain m p is the coordinates of the latent values p.val · norn
- Hugin_vg_instances.Instances.uncutuncut p is p without its clips and tags.val · hugin
- Nx.Placement.uncutuncut p is p with every axis whole on each of its devices, each mesh axis that cut one holding copies, and the value no sum.val · nx
- Nx.Ptree.Walk.underunder outer inner c x walks x with outer at c's path, and each position of outer's parameter with inner at that position's path, inner's paths after the position's as nest makes them.val · nx
- Quill.Session.undoundo s restores the previous document state.val · quill
- Tolk.Shape.unflattenunflatten u axis sizes splits axis into sizes.val · tolk
- Rig_edge_primitives.unholdunhold () counts the give of a lock that hold counted, after the give (rig_unhold).val · rig
- Brot.Pre_tokenizer.unicode_scriptsunicode_scripts splits on Unicode script boundaries.val · brot
- Kaun.Init.uniformuniform ~scale samples uniformly from [0;scale).val · kaun
- Norn.Dist.uniformuniform ~low ~high has density 1 / (high - low) on (low, high).val · norn
- Norn_model.Init.uniformuniform draws each coordinate uniformly in (-2, 2).val · norn
- Nx.Rng.uniformuniform dt shape is draws from [0, 1): multiples of 2{^-p}, each equally likely, p the largest such that every multiple of 2{^-p} in [0, 1) is a value of dt: dt's significand width, from float64's 53 down to float8_e5m2's 3; float4_e2m1fn's values below 1 are 0 and 0.5, so its p is 1.val · nx
- Tolk.UniformValues the same for every thread of a workgroup.module · tolk
- Ymir.Time.uniformThe type of a scale whose every day lasts 86400 of its seconds.type · ymir
- Nx.Rng.uniform_atuniform_at p dt shape is uniform dt shape at p.val · nx
- Brot.unigramunigram () is a Unigram tokenizer.val · brot
- Hugin_gg.Box2.unionunion a b is the smallest box containing a and b.val · hugin
- Nx_io.Archive.unionunion ts is the archive of the entries of all ts.val · nx
- Talon.Kit.unionunion rest q is q's rows, then rest's, written q |> Kit.union rest, with q's columns, then those that only rest has, in order; a column is null on the side that lacks it:val · talon
- Tolk.Op.Set.unionunion s0 s1 is the set of the operations in s0 or in s1.val · tolk
- Ymir_units.Vocabulary.unionunion a b is a's entries then b's.val · ymir
- Talon.Kit.union_allunion_all qs is the rows of qs, in order, with the columns of the first query, then each column that a later one adds, in order, null on the queries that lack it.val · talon
- Nx.uniqueunique keys groups the keys of keys (elements of a 1-d keys, rows of a 2-d one) that are equal in the sort order: every NaN is one group, -0.val · nx
- Nx_cpu.uniqueunique ~ids ~first ~counts ~count keys numbers the rows of keys, an [n; w] array whose elements are read as unsigned 64-bit words: two rows are one key iff their words are equal, and a w of 0 makes every row one key.val · nx
- Nx_kernel.S.uniqueunique ~ids ~first ~counts ~count keys numbers the rows of keys, an [n; w] array whose elements are read as unsigned 64-bit words: two rows are one key iff their words are equal, and a w of 0 makes every row one key.val · nx
- Nx_kernel.Shape.uniqueunique ~size keys is the shapes of a numbering's ids, of its first and counts, and of its count, for keys of [n; w]: [n], [size] and [].val · nx
- Nx_ragged.uniqueunique r groups the equal rows of r as Nx.unique groups keys, numbered in order of first appearance, with each group's first row and number of rows.val · nx
- Tolk.Ops.unique_numunique_num () is a slot number no other call returns.val · tolk
- Hugin_vg_vector.Vector.unitunit m is the linear part of m divided by Affine.stretch m, which scales no length by more than 1..val · hugin
- Nx.Ptree.unitunit is (), which has no part.val · nx
- Ymir_fits.Fits.Table.unitunit name hdu is the column's TUNITn, parsed as unit parses BUNIT.val · ymir
- Ymir_fits.Fits.unitunit hdu is BUNIT, parsed with Unit.parse scoped by <digest>#<EXTNAME> (HDU<i> when unnamed), so one file read twice gives equal units and two files give two; None if BUNIT is absent.val · ymir
- Ymir_fits.Fits.UnitFITS unit strings (FITS 4.0 §4.3).module · ymir
- Ymir_units.Constant.unitunit k is k's unit.val · ymir
- Ymir_units.Quantity.unitunit q is q's unit.val · ymir
- Ymir_units.UnitUnits as exact values.module · ymir
- Talon.Type.unit_The type for the units of temporal ticks.type · talon
- Ymir.UnitsAstronomy's units.module · ymir
- Brot.unk_tokenunk_token t is t's unknown token, if any.val · brot
- Hugin_kit.Scale.unknownunknown s is the colour of the missing values of s, if set.val · hugin
- Munin_sys.unknownunknown is a device of which nothing is known, every field None.val · munin
- Rig_amd.unloadunload d i releases what image, prepare and entry made for i.val · rig
- Rig_cuda.unloadunload d i releases what image, prepare and entry made for i.val · rig
- Rig_edge.Driver.unloadunload d i releases what image, prepare and entry made for i.val · rig
- Rig_metal.unloadunload d i releases what image, prepare and entry made for i.val · rig
- Rig_nv.unloadunload d i releases what image, prepare and entry made for i.val · rig
- Rig_remote_proxy.unloadunload d i releases what image, prepare and entry made for i.val · rig
- Rig_pci.Function.unmapunmap f w unmaps w.val · rig
- Rig_pci.Page_table.unmapunmap t ~va n unmaps the n bytes mapped from va, frees the tables that become empty, and flushes.val · rig
- Rig_edge_primitives.unmaskunmask () ends a mask (rig_unmask).val · rig
- Talon.Query.unnestunnest ?keep_empty os q is the rows of the frames of the lists that the outputs os compute (Lists), concatenated in q's row order, each row's positions in list order.val · talon
- Rig.Submission.unorderedunordered slot ~at is a ref through which the work reads and writes the buffer slot at any time, ordered only by device order, the exception to rule 3 of Order.val · rig
- Fehu.Space.unpackunpack s v is Ok x if v can be converted to a valid element of s, or Error msg otherwise.val · fehu
- Nx.unpackunpack dt p is the value in p if its dtype is dt.val · nx
- Rig_pci.Function.unpinunpin f a n releases one pin of the n bytes at a.val · rig
- Nx.unravel_indexunravel_index i s is the coordinates in the shape s, in C order, of each flat index of i: one value of i's shape per axis of s, [] for a 0-d s.val · nx
- Tolk.Postrange.Scheduler.unread_by_opaqueunread_by_opaque k is k's ranges that the sources of one of its opaque operations do not read, as j for f(&x[i]) beside a loop over j.val · tolk
- Tolk.Heuristic_host.unroll_roomunroll_room k is whether the lanes k has upcast leave an unroll room: fewer registers of them than the target has, where the target knows its registers.val · tolk
- Tolk.Heuristic_host.unroll_vectorunroll_vector k ~fits unrolled axis unrolls axis by the largest power of two that fills a vector beside the unrolled lanes and that fits admits.val · tolk
- Nx.Repr.unsafe_of_arrayunsafe_of_array s a is of_array s a for an a its caller gives up: the value owns a's memory, as it owns memory nx allocated for it, so a donation of the value may write its result there (donate).val · nx
- Nx.Repr.unsafe_of_shardsunsafe_of_shards p arrays is of_shards p arrays for arrays their caller gives up: the value owns their memory, as it owns memory nx allocated for it, so a donation of the value may write its result there on each device (donate).val · nx
- Rig_pci.Window.unsafe_transportunsafe_transport p is the transport whose struct rig_pci_transport is at address p of the process.val · rig
- Sowilo.unsharp_maskunsharp_mask ~sigma img sharpens by subtracting a Gaussian blur: img + amount * (img - gaussian_blur ~sigma img).val · sowilo
- Rig_edge_primitives.unshieldunshield () ends a shield as unmask ends a mask, re-arming the runtime at the outermost (rig_unshield).val · rig
- Nx.unsqueezeunsqueeze ~axes x is x with an axis of extent 1 at each position of axes, positions of the result: unsqueeze ~axes:[ 0; -1 ] of a [3] value is [1; 3; 1].val · nx
- Tolk.Shape.unsqueezeunsqueeze u axis is u with an axis of size 1 inserted at axis, which counts from the end of the new shape when negative.val · tolk
- Tolk.Symbolic.uop_given_validuop_given_valid valid u is u simplified with Shape.symbolic, knowing that the boolean valid holds.val · tolk
- Tolk.Uop_weakCommitting weak data types to widths.module · tolk
- Tolk.Ops.UpatPatterns.module · tolk
- Tolk.Heuristic_host.upcast_contiguousupcast_contiguous k upcasts the axis k's stores write at unit stride, when every access reading it does so at unit stride, by the largest power of two that fills a vector and fits the budget.val · tolk
- Tolk.Heuristic_common.upcast_maxupcast_max is the most values an upcast of one axis takes (Postrange.Scheduler.apply_opt's bound).val · tolk
- Tolk.Postrange.Scheduler.upcast_sizeupcast_size k is the product of the sizes of k's Ops.Axis_type.t.Upcast axes.val · tolk
- Tolk.Heuristic_host.upcast_vectorsupcast_vectors k upcasts the last output axes of k, a kernel without a reduce, while its lanes have room in a vector.val · tolk
- Tolk.Postrange.Scheduler.upcastable_dimsupcastable_dims ~reduce k is the positions of k's global, local and weak axes of a known size that no range's bound reads (read_by_bounds), no opaque operation reads (read_by_opaque) and no register store or loop's condition depends on (Rewriting), that are reduce axes if reduce is true and are not if it is false (default).val · tolk
- Quill.Doc.updateupdate id f d applies f to the cell identified by id.val · quill
- Quill.Session.update_sourceupdate_source id source s updates the source of cell id.val · quill
- Munin.Run.updated_atupdated_at r is when r last wrote: its log's modification time when read, from the clock of the file system that holds it.val · munin
- Talon.Expr.Str.upperupper a is a in uppercase by the same rules, from the full uppercase mappings: U+00DF becomes SS, and U+FB03 FFI.val · talon
- Tolk.Ops.uproduprod x ys is the product of x :: ys, or their conjunction if x is boolean.val · tolk
- Talon.Time.Span.usus n is n microseconds.val · talon
- Rig_nv_nvidia.usageThe type for what a GPU's RM reports of its use, by every process.type · rig
- Rig_nv_nvidia.usageusage ~window i is GPU i's usage, busy over the last window seconds.val · rig
- Rig_program.useThe type for how a ref's work uses its slot, as Rig.Submission.read, Rig.Submission.write and Rig.Submission.unordered.type · rig
- Munin.Session.use_artifactuse_artifact s a appends a use record: s's run used the version a.val · munin
- Tolk.Setting.use_tcuse_tc is how kernel optimization uses tensor cores, from TC.val · tolk
- Munin.History.usesuses h is the artifact versions the run used, as names and numbers, in log order.val · munin
- Tolk.Ops.usumusum x ys is the sum of x :: ys, or their disjunction if x is boolean.val · tolk
- Talon.Time.utcThe index of the UTC clock.type · talon
- Talon.Tz.utcutc is Coordinated Universal Time: the zone named "UTC", whose offset is 0 at every instant.val · talon
- Ymir.Time.utctype · ymir
- Ymir.Time.utcutc is Coordinated Universal Time.val · ymir
v
- Hugin.Mark.vv ~name ~reduce ~coord ~shape ~swatch bindings draw is the mark name of the channels bindings, where draw rows is its picture in a panel for that panel's rows, in points on the page.val · hugin
- Hugin.Theme.vv ~ink ~paper ~accent ~size ~fonts ~palette ~scheme ~locale () is the theme with:val · hugin
- Hugin_font.Run.vv ~ys ~clusters ~font ~size ~text ~glyphs ~xs () is the run of the glyphs of font at em size size rendering text, glyph i with its origin at (xs.(i), ys.(i)), where:val · hugin
- Hugin_gg.Box2.vv x y w h is the box with corner (x, y), width w and height h: it spans [x;x +.val · hugin
- Hugin_gg.Color.vv ~alpha r g b is the colour with encoded sRGB components r, g and b and opacity alpha, which defaults to 1..val · hugin
- Hugin_gg.P2.vv x y is the point (x, y).val · hugin
- Hugin_gg.Stroke.vv ~cap ~join ~miter_limit ~dash ~dash_offset width is the stroke style of line width width with:val · hugin
- Hugin_gg_kit.Field2.vv ~xs ~ys z is the field whose sample at row i and column j is the element (i, j) of the two-dimensional tensor z, at (xs.(j), ys.(i)), where:val · hugin
- Hugin_gg_kit.Pgon2.vv rs is the polygon of the rings rs, in order.val · hugin
- Hugin_gg_kit.Ring2.vv xs ys is the ring through the points (xs.(i), ys.(i)) in order.val · hugin
- Hugin_kit.Dash.vv lengths is the pattern of lengths, dash first.val · hugin
- Hugin_kit.Locale.vv ~decimal ~group ~grouping ~minus ~months () is the locale with:val · hugin
- Hugin_kit.Number.vv ~trim ~group n p is the format writing numbers in notation n to precision p, where:val · hugin
- Hugin_kit.Time.vv r n is the instant n units of r after the epoch, or -n units before it if n is negative.val · hugin
- Hugin_text.Text.Layout.vv ~width ~halign ~valign ~fonts ~size t is t set in fonts (Faces) at size size (Placing glyphs), broken into lines at most width wide (Lines) and aligned by halign and valign (Alignment).val · hugin
- Hugin_text.Text.vv s is the characters of the UTF-8 string s, neither bold nor italic, with size factor 1., shift 0.val · hugin
- Hugin_vg.Renderable.vv w h p is p on a page w points wide and h points high.val · hugin
- Jera.Piecewise.vv s ~breaks c is the series with pieces + 1 breaks and the coefficients c, each leaf of shape [pieces; degree + 1] @ value.val · jera
- Jera.Quad.Box.vv lo hi is the box with corners lo and hi, which broadcast together to lanes @ [d]: a point's coordinates are the last axis.val · jera
- Jera.Quad.Range.vv a b is [a, b], a and b broadcast together.val · jera
- Jera.Sde.Brownian.vv key dtype ~shape ~t0 ~t1 ~depth is a Brownian path of independent standard components of shape shape over [t0, t1], resolved to (t1 − t0) / 2^depth.val · jera
- Jera.Split.vv ~kick ~drift is the splitting K(kick.(0) h) D(drift.(0) h) ….val · jera
- Jera.Tol.vv ~rel ~abs is the tolerance abs + rel * |y|.val · jera
- Norn.Draws.vv u x is x as draws.val · norn
- Norn.Summary.vv u ?stats ?superchains d summarises the draws d of u.val · norn
- Norn_model.vv dtype latent observed gen is the model gen describes.val · norn
- Nx.Devices.vv is the set.val · nx
- Nx.Host.vv is the set.val · nx
- Nx.Mesh.vv s axes lays s's devices in row-major order over the named axes.val · nx
- Nx.Pattern.vv s is the pattern s.val · nx
- Nx.Ptree.Path.vv segs is the path of segs, from the root.val · nx
- Nx.Ptree.vv walk is the description whose values walk walks: a structure without a parameter, on one set, such as a record of parts walked with Walk.structure.val · nx
- Nx_array.Layout.vv ~offset ~strides s is the layout of shape s with strides and offset (defaults to 0), in canonical form.val · nx
- Nx_array.vv dt l b is the array of dt elements laid out by l over b's bytes.val · nx
- Nx_kernel.Prog.vv ~ins nodes ~outs is the program over operands of dtypes ins whose results are the nodes outs.val · nx
- Nx_ragged.vv ~offsets values is the ragged array of values cut at offsets.val · nx
- Nx_wide.vv ~lo hi is the number hi + lo, exactly.val · nx
- Rig_bin.Version.vv is the version, as "1.0.0~alpha3", which rig --version prints.val · rig
- Talon.Col.vv k name is the column name read as k: v (Kind.list Kind.int) "tokens".val · talon
- Talon.Column.vv ty vs is the column of type ty holding vs, without nulls.val · talon
- Talon.Error.vv ?file ?line ?column ?row_group ?bytes ?text msg is the error msg, found:val · talon
- Talon.Ext.vv ~name ~metadata ~ordered storage ~dec ~enc declares the extension type Type.ext ~name ~metadata storage with:val · talon
- Talon.Schema.vv columns is the schema of columns, in order.val · talon
- Talon.Source.vv ~name ~schema ?rows ?sorted ?pushdown parts is the source called name that yields rows of the columns schema, with:val · talon
- Talon.vv ?rows cs is the table of the columns cs, in order, as one batch of rows rows.val · talon
- Tolk.Compiler_worker.vv ~name program is the worker whose executable is program, named name in its files and its errors.val · tolk
- Tolk.Ops.Pattern_matcher.vv rules is the rewriter of the rules rules () returns, which it calls on its first rewrite.val · tolk
- Tolk.Ops.Upat.vv () matches the nodes of an operation in op, a type in dtype and argument arg, each when given.val · tolk
- Tolk.Ops.vv ~src ~arg op is the node op of src (default []) with argument arg (default No_arg).val · tolk
- Tolk.Postrange.Scheduler.vv ast ren schedules the kernel ast, the sink of a kernel (Ops.kernel_info), for ren, with the ranges its reductions and ends close flattened (Simplify.pm_flatten_range).val · tolk
- Tolk.Renderer.Compiler.vv ~cachekey ~disassemble compile is the compiler that compiles a source with compile.val · tolk
- Tolk.Renderer.vv target is the renderer for target with these fields.val · tolk
- Tolk.Tc.vv ~dtype_in ~dtype_out ~frag_a ~frag_b ~frag_c is the tensor core with these types and fragments.val · tolk
- Ymir.Healpix.vv f ~nside s is the pixelisation of f's sphere at nside, numbered by s.val · ymir
- Ymir.Observation.vv ?variance ?valid ?area g data is data, batch @ Grid.shape g, on g.val · ymir
- Ymir.Time.vv s ~day ~second is the instants second after 0h of the modified Julian dates day on s, their shapes broadcast.val · ymir
- Ymir_fits.Fits.vv h b is the HDU of any kind with header h and data unit b, as write writes it after the first HDU: DATASUM and CHECKSUM go, and a primary image becomes an IMAGE extension.val · ymir
- Ymir_units.Constant.vv ~name text u is the constant text in u, named name.val · ymir
- Ymir_units.Quantity.vv u x is x in u.val · ymir
- Ymir_units.Vocabulary.vv entries is the vocabulary of entries, each a symbol, whether it takes a prefix, and its unit, in the order spell prefers them.val · ymir
- Ymir_units.Codata.v2018v2018 is the 2018 adjustment.val · ymir
- Ymir_units.Codata.v2022v2022 is the 2022 adjustment.val · ymir
- Ymir_units.Codata.vacuum_permeabilityvacuum_permeability r is μ0, the vacuum magnetic permeability, in N A-2.val · ymir
- Ymir_units.Codata.vacuum_permittivityvacuum_permittivity r is ε0, the vacuum electric permittivity, in F m-1.val · ymir
- Norn.Dist.validvalid d is whether every element of every parameter of d is in its domain, a scalar over all of d's parameters: under Rune.vmap, one per lane.val · norn
- Tolk.Shape.validvalid u cond is u where cond holds and `Invalid elsewhere.val · tolk
- Ymir.Observation.validvalid o is where o's samples were measured, if known.val · ymir
- Talon.Column.validityvalidity c is true at row i iff row i of c holds a value.val · talon
- Ymir_fits.Fits.Image.validityvalidity ~window hdu is false where a pixel is undefined (BLANK or NaN), or None if none is.val · ymir
- Ymir_fits.Fits.Table.validityvalidity ~rows name hdu is false where an element is undefined, in raw's shape or the ragged elements'; None if none is.val · ymir
- Hugin_text.Text.Layout.valignThe type for vertical alignments (Alignment).type · hugin
- Fehu.ValueUniversal value type.module · fehu
- Hugin.Role.valuevalue ~name is the role name whose values are the quantities of its channel as they are, nan where missing: it reads no scale, contributes to no domain and yields no guide, so a channel given to it has neither a scale nor a title (Mark.v).val · hugin
- Munin.Metric.valuevalue s i is the value of s's ith sample.val · munin
- Munin.ValueValues of params, summaries and metadata.module · munin
- Nx_io.Gguf.valueThe type for metadata values, one case per GGUF value type.type · nx
- Rig.Point.valuevalue p is p's value, above the value its device's timeline started after (submitted).val · rig
- Rig_program.valueThe type for what a run knows: a word read when the step that needs it runs.type · rig
- Talon.Source.Pred.valuetype · talon
- Tolk.Dtype.valueThe type for values.type · tolk
- Tolk.Dtype.ValueArithmetic and comparison on values, as on numbers: a `Bool counts as 0 or 1, an operation on two integers gives an exact integer, and one that involves a float gives a float, in which an integer that rounds past the greatest double is the infinity of its sign.module · tolk
- Tolk.Ops.valuevalue c is the value of the constant c, read through a cast.val · tolk
- Tolk.Setting.valuevalue v is v's current value on the calling domain.val · tolk
- Ymir_fits.Fits.ValueValue fields.module · ymir
- Ymir_units.Quantity.valuevalue u q is q's payload in u: the payload itself when unit q is u, otherwise the payload times Unit.ratio d (unit q) u, with d the payload's dtype.val · ymir
- Rune.value_and_gradvalue_and_grad p f params is (f params, grad p f params), from one forward and one backward pass.val · rune
- Rune.value_and_grad'value_and_grad' f x is value_and_grad of one array.val · rune
- Rune.value_and_grad_auxvalue_and_grad_aux p x f params is (y, g, aux), where (y, aux) = f params and g is the gradient of y at params.val · rune
- Talon.Kit.value_countsvalue_counts c q is each distinct value of q's column c, null included, with its number of rows, most frequent first and ties in order of first appearance:val · talon
- Nx_ragged.valuesvalues r is r's values, including any rows outside its offsets.val · nx
- Talon.Column.valuesvalues k c is c's values read as k.val · talon
- Talon.Query.valuesvalues e q is e on each row of q, decoded to OCaml: values Expr.(const mk $ carrier $ option origin) q.val · talon
- Ymir_fits.Fits.Image.valuesvalues ~window dtype hdu is the physical values of hdu's image, NaN where undefined.val · ymir
- Ymir_fits.Fits.Table.valuesvalues ~rows dtype name hdu is a numeric Array column's physical values, NaN where undefined: on a scaled column TZERO + TSCAL × s computed as one fma in float64, then cast once.val · ymir
- Nx.varvar ~axes ~ddof x is the variance Σ (x - m)² / (n - ddof) of the n elements, m their mean; ddof defaults to 0, the population variance, and 1 gives the sample variance.val · nx
- Talon.Expr.varvar a is the sample variance of a's values, like std.val · talon
- Tolk.Ops.Upat.varvar ~dtype name matches any node of a type in dtype, named name.val · tolk
- Tolk.Ops.variablevariable ~dtype name lo hi is the scalar variable name of type dtype (default Dtype.t.Weak_int) ranging over lo to hi.val · tolk
- Tolk.Shape.variablesvariables u is the unbound scalar variables u reads, sorted by name and slot.val · tolk
- Norn.Gaussian.variancevariance u g is the diagonal of g's covariance, element by element.val · norn
- Ymir.Observation.variancevariance o is o's variance, 0 where invalid.val · ymir
- Kaun.Init.variance_scalingvariance_scaling ~scale ~mode ~distribution samples with target variance scale / n, with:val · kaun
- Hugin.variesvaries shape c a is true iff the values of c can differ along the axis a of a mark of shape shape (Mark.broadcast), a counting from the last axis if negative: a tensor or an array whose dimension aligned with a exceeds 1, or an index or a dim of the axis a if its length exceeds 1.val · hugin
- Tolk.Uniform.varyingvarying sink is v such that v u is false only when u, a value of the kernel sink, is the same for every thread of a workgroup that computes it.val · tolk
- Tolk.Shape.vconst_likevconst_like u c is the constant c of u's type repeated max_numel u times.val · tolk
- Nx.vdotvdot a b is the 0-d sum of the products of conjugate a and b, both flattened, accumulated as matmul.val · nx
- Fehu.Vec_envVectorized environments.module · fehu
- Nx.vecdotvecdot ~axis a b is the sum of the products of conjugate a and b along each operand's axis axis (default -1); their other axes broadcast to the result's shape.val · nx
- Hugin_vg_vector.VectorWhat the SVG and PDF renderers share: the walk of a picture, numbers written to an accuracy in points, the maps they write, and paths mapped to a frame and cut at a margin around the page.module · hugin
- Nx.vector_normvector_norm dt x is, in dt, the norm of order ord (default 2.) of the vector of the elements of x along axes (default all of them), dropping those axes unless keepdims.val · nx
- vegaVega provides gradient-based optimizers for OCaml.library · vega
- VegaGradient-descent optimizers.module · vega
- Ymir_fits.Fits.verifyverify hdu reads hdu once and checks its checksums (§4.4.2.7): DATASUM, the ones' complement sum of the data unit as big-endian 32-bit words, and CHECKSUM, which makes the whole HDU sum to −0.val · ymir
- Munin.Artifact.versionversion a is a's number, from 1.val · munin
- Munin.Store_format.versionversion is the format this library reads and writes, 1.val · munin
- Nx_io.Gguf.versionversion g is g's format version, 2 or 3.val · nx
- Rig_amd_abi.Gpu.versionThe type for versions, as (major, minor, stepping), such as (9, 4, 3) for the GC of an MI300.type · rig
- Rig_bin.Versionrig's version, the rig package's.module · rig
- Rig_remote_proxy.Wire.versionversion is the protocol's version, the handshake's: 12.val · rig
- Sowilo.vflipvflip img flips vertically (top to bottom).val · sowilo
- Norn.Gaussian.vivi u ?rank ?draws ~steps lp k start ascends the bound in steps stochastic gradient steps (ADVI; Kucukelbir et al.val · norn
- Hugin.ViewView values.module · hugin
- Rig.Buffer.viewview b ~first ~length is the length bytes of b from its byte first on, in b's memory.val · rig
- Rig_program.viewThe type for length bytes of memory memory from its byte offset.type · rig
- Tolk.Shape.view_asview_as u shape views the flat storage u as shape.val · tolk
- Tolk.Ops.view_stateview_state p is the state of storage the view p names, its second source: its storage's parameter, or the Op.t.After whose elements a read of the run takes.val · tolk
- Hugin_kit.Scheme.viridisviridis runs from dark purple through blue and green to yellow (van der Walt, Smith and Firing).val · hugin
- Kaun.Cache_index.visiblevisible index is mask as the mask of Nx.attention over [| batch; heads; seq; context |], k in place of context under a selection: at a pair it holds where mask index holds at the pair's lane, token and column, whatever the heads.val · kaun
- Nx.Ptree.visitThe type for the steps of a walk.type · nx
- Nx.Ptree.visitsvisits s x is the steps of s's walk of x, in walk order.val · nx
- Rune.vjpvjp p q f params is (f params, pullback), pullback mapping the result's cotangents to params' (Derivatives).val · rune
- Rune.vjp'vjp' f x is vjp of one array.val · rune
- Rune.vmapvmap ?axis s f is f mapped over axis 0 of its arguments' arrays, s being its signature, one structure per argument and one for the result.val · rune
- Rune.vmap'vmap' ?axis f x is vmap of a function of one array.val · rune
- Tolk.Ops.vmaxvmax u is an upper bound of every value u can take.val · tolk
- Tolk.Ops.vminvmin u is a lower bound of every value u can take.val · tolk
- Brot.vocabvocab t is t's vocabulary as (token, id) pairs, the model's followed by the added tokens it does not already hold.val · brot
- Brot.vocab_sizevocab_size t is the number of tokens in t's vocabulary, added tokens included.val · brot
- Ymir_fits.Fits.Unit.vocabularyvocabulary is FITS 4.0 Tables 3 and 4 at the standard's values (a parsec is 3.0857e16 m, a solar mass 1.9891e30 kg), with the prefixes of Table 5 on the symbols the standard allows them, and the archives' spellings Angstrom, angstrom, electron, ELECTRONS, DN and COUNTS.val · ymir
- Ymir_units.VocabularyNames for units.module · ymir
- Ymir_units.Unit.voltvolt is kg m2 s-3 A-1.val · ymir
- Norn.Dist.von_misesvon_mises ~loc ~concentration is the distribution of angles on [-π, π] with density exp (κ cos (x - loc)) / (2π I₀(κ)), κ the concentration, in [0, inf): uniform at κ = 0, near a normal of variance 1 / κ about loc for a large κ.val · norn
- Nx.Rng.von_misesvon_mises k is von Mises draws of mean direction 0 and concentration k, elementwise, as angles in [-π, π]: uniform on the circle at k = 0, near a normal of variance 1 / k for a large k.val · nx
w
- Hugin_gg.Box2.ww b is maxx b -.val · hugin
- Hugin_vg.Renderable.ww r is the width of the page of r, in points.val · hugin
- Rig.Buffer.waitwait b access returns once the work that a host access access of b must follow is done.val · rig
- Rig.Point.waitwait p returns once p's device has reached p's value.val · rig
- Rig_bin.Proc.waitwait ~until rs returns once a reader of rs has bytes or its end, a signal of signals came, or the time is until (defaults to never).val · rig
- Rig_bnxt.Qp.waitwait qp side ~us is poll once the oldest due entry of side has a completion, waiting for it at most us microseconds.val · rig
- Rig_mlx5.waitwait nic ~ms is the events of nic since the last wait, once there is one, or [] if none came within ms milliseconds.val · rig
- Rig_pci.Function.waitwait f ~us what cond calls cond, at least once, until it is true or at least us microseconds have passed on a monotonic clock.val · rig
- Rig_remote_proxy.Link.waitwait j ~ms is j's state once it is no longer Open, or after ms milliseconds.val · rig
- Rig_nv_abi.Method.wait_for_idlewait_for_idle holds the compute engine's later methods until the launches scheduled before it have completed.val · rig
- Rig_edge.waitsThe type for the producers a device's queues wait for.type · rig
- Hugin_vg_vector.Vector.walkwalk t ctx b p writes p into b with t.val · hugin
- Jera.Sde.Brownian.walkwalk c w walks the path w: its key at key (Nx.Rng.walk), the shape's dimensions reported at shape, the interval's ends at t0 and t1, and the depth reported at depth.val · jera
- Kaun.Attention.Cache.walkwalk c cache walks cache.keys then cache.values, at the paths "keys" and "values": the cache's Nx.Ptree.S instance.val · kaun
- Kaun.Attention.walkwalk c p walks the projections q, k, v then out, each by Linear.walk, at those paths: the layer's Nx.Ptree.S instance.val · kaun
- Kaun.Batch_norm.Stats.walkwalk c s walks s's statistics, mean then var, at those paths: the statistics' Nx.Ptree.S instance.val · kaun
- Kaun.Batch_norm.walkwalk c p walks p's parameters, gamma then beta, at those paths: the parameters' Nx.Ptree.S instance.val · kaun
- Kaun.Cache_index.walkwalk c index walks an index at c's path: Nx.Ptree.v walk is its structure, for the signature of a compiled step and for Nx.Ptree.map over its tensors.val · kaun
- Kaun.Conv.walkwalk c p walks p's parameters: w at "w", then b at "b", reporting whether b is present, so a layer with a bias and one without are distinct structures.val · kaun
- Kaun.Embedding.walkwalk c p walks p's parameters, table, at that path: the layer's Nx.Ptree.S instance.val · kaun
- Kaun.Layer_norm.walkwalk c p walks p's parameters, gamma then beta, at those paths: the layer's Nx.Ptree.S instance.val · kaun
- Kaun.Linear.walkwalk c p walks p's parameters: w at "w", then b at "b", reporting whether b is present, so a layer with a bias and one without are distinct structures.val · kaun
- Kaun.Rms_norm.walkwalk c p walks p's parameters, gamma, at that path: the layer's Nx.Ptree.S instance.val · kaun
- Nx.Ptree.S.walkwalk c x walks every part of x with c and is x rebuilt from what the walks return.val · nx
- Nx.Ptree.WalkThe functions a walk is written with, one per kind of part.module · nx
- Nx.Rng.walkwalk c k walks k, a key or a batch of keys, at c's path as one fixed array (Ptree.Walk.tensor), and rebuilds it as of_tensor does.val · nx
- Nx_quant.walkwalk c w walks w at c's path: its case, "mxfp4", "mxfp4_blocks", "q8_0", "q4_k" or "q6_k", then each part as a fixed array (Nx.Ptree.Walk.tensor) at its field's name, codes then scales, or blocks.val · nx
- Nx_wide.walkwalk c w walks w as two fixed arrays (Nx.Ptree.Walk.tensor) at c's path then "hi" and "lo": hi and lo for a value an operation or v made, and the words it was rebuilt from for one a walk rebuilt.val · nx
- Vega.Adafactor_state.walkwalk c x walks every part of x with c and is x rebuilt from what the walks return.val · vega
- Vega.Adagrad_state.walkwalk c x walks every part of x with c and is x rebuilt from what the walks return.val · vega
- Vega.Adam_state.walkwalk c x walks every part of x with c and is x rebuilt from what the walks return.val · vega
- Vega.Adan_state.walkwalk c x walks every part of x with c and is x rebuilt from what the walks return.val · vega
- Vega.Lion_state.walkwalk c x walks every part of x with c and is x rebuilt from what the walks return.val · vega
- Vega.Loss_scale.walkwalk c ls walks a loss scale at c's path: the fixed tensors scale then good_steps, at those fields, reporting nothing.val · vega
- Vega.Rmsprop_state.walkwalk c x walks every part of x with c and is x rebuilt from what the walks return.val · vega
- Vega.Sgd_state.walkwalk c x walks every part of x with c and is x rebuilt from what the walks return.val · vega
- Ymir.Cosmology.Structure.walkwalk c x walks every part of x with c and is x rebuilt from what the walks return.val · ymir
- Ymir.Cosmology.walkwalk c x reports codata's year with Walk.int, then walks the other fields in order, quantities with Quantity.walk.val · ymir
- Ymir.Direction.walkwalk c d walks d: the frame as Frame.walk reports it, then the vectors as a tensor of a fixed type, at c's path.val · ymir
- Ymir.Earth.walkwalk c e walks e: its TT instants at field "tt" as Time.walk walks them, then UT1's days at "ut1_day" and seconds at "ut1_second", as tensors of a fixed type.val · ymir
- Ymir.Frame.walkwalk c f reports name as one case at c's path and is f.val · ymir
- Ymir.Time.walkwalk c t walks t: the scale's name as one case at c's path, then the days at field "day" and the seconds at field "second", as tensors of a fixed type.val · ymir
- Ymir_units.Quantity.Structure.walkwalk c x walks every part of x with c and is x rebuilt from what the walks return.val · ymir
- Ymir_units.Quantity.walkwalk c q reports q's unit with one Walk.case holding its canonical text, then walks its payload with Walk.leaf, both at c's path.val · ymir
- Ymir_units.Unit.walkwalk c u is u, walked as one Walk.case holding its canonical text, and no tensor.val · ymir
- Talon.Col.wallwall name is v Kind.wall name, which binds every datetime type of wall clocks.val · talon
- Talon.Expr.Temporal.wallwall z a is the reading of z's wall clock at the instant a, in a's unit: a plus the offset of z at a (Tz.offset_s).val · talon
- Talon.Expr.wallwall t is the wall clock t.val · talon
- Talon.Kind.wallwall reads datetime columns of every unit whose clock is a wall clock (Instants and wall clocks).val · talon
- Talon.Time.wallThe type for wall clocks.type · talon
- Talon.Type.wallwall u is Datetime of wall clocks in the unit u.val · talon
- Norn.Ensemble.warmupwarmup u lp k ~steps s is s after steps transitions, whose draws are discarded.val · norn
- Norn.Hmc.warmupwarmup u lp k ~steps s is s after steps transitions that tune the step size, the length and the geometry, by Stan's schedule scaled to steps: an initial buffer, windows that double in length, and a final buffer.val · norn
- Norn.Nuts.warmupwarmup u lp k ~steps s is s after steps transitions that tune each chain's step size and geometry, by Stan's schedule scaled to steps: an initial buffer, windows that double in length, and a final buffer.val · norn
- Vega.Schedule.warmup_cosinewarmup_cosine ~init_value ~peak_value ~warmup_steps is cosine warmup from init_value to peak_value over warmup_steps.val · vega
- Vega.Schedule.warmup_cosine_decaywarmup_cosine_decay ~init_value ~peak_value ~warmup_steps ~decay_steps ?end_value () is linear warmup from init_value to peak_value over warmup_steps, then cosine decay to end_value over decay_steps.val · vega
- Hugin.Mark.warnwarn rows msg adds the warning msg, about the mark's data, to the warnings of the drawing (Drawing.warnings), under the mark's id.val · hugin
- Hugin.warningThe type for warnings: the node whose data a problem concerns, and a message for users.type · hugin
- Hugin.Drawing.warningswarnings d is the warnings of resolving, laying out and drawing, in the order of the figure.val · hugin
- Hugin.Layout.warningswarnings l is the warnings of resolving and laying out, in the order of the figure.val · hugin
- Hugin.Resolved.warningswarnings r is the warnings of resolving, in the order of the nodes they are about in the figure, a generated node in the place of the node it lies under, then those about values of the view that no key reads.val · hugin
- Ymir_units.Unit.wattwatt is kg m2 s-3.val · ymir
- Rig_amd_abi.Thread_trace.waveThe type for the waves of a trace.type · rig
- Rig_amd_abi.Thread_trace.waveswaves g trace is the waves that start and end in trace, one shader engine's bytes as g wrote them, in the order they end.val · rig
- Ymir.WcsFITS world coordinates.module · ymir
- Tolk.Dtype.weakweak dt is the weak data type of dt's kind: Weak_float for a float, Weak_int for an integer, and dt itself for Bool and Void.val · tolk
- Tolk.Shape.weakweak u is is_weak (Ops.dtype u).val · tolk
- Tolk.Dtype.weaksweaks is [Weak_int; Weak_float].val · tolk
- Ymir_units.Unit.weberweber is kg m2 s-2 A-1.val · ymir
- Hugin_kit.Time.weeksweeks k is the interval of k weeks starting on Monday.val · hugin
- Hugin_font.Font.weightweight f is the weight class of f from its OS/2 table, clamped to [1;1000]: 400 is regular and 700 bold.val · hugin
- Norn.WeightedWeighted draws.module · norn
- Nx.weighted_histogramweighted_histogram ~weights dims is histogram dims's cells, each holding the sum of its points' weights in weights' dtype.val · nx
- Ymir.Region.weightsweights r g is each cell's weight in [0, 1], batch @ Grid.shape g, the region's and the grid's batch axes broadcast.val · ymir
- Nx.wherewhere c x y is x's element where c is true and y's elsewhere.val · nx
- Nx_kernel.Prog.E.whereval · nx
- Talon.Sel.wherewhere p selects the columns (n, t) for which p n t is true, in schema order.val · talon
- Tolk.Ops.Elementwise.wherewhere c x y is x where c holds and y elsewhere.val · tolk
- Tolk.Ops.Upat.wherewhere c x y is x where c holds and y elsewhere.val · tolk
- Tolk.Ops.wherewhere c x y is x where c holds and y elsewhere.val · tolk
- Hugin_gg.Color.whitewhite is v 1.val · hugin
- Brot.Pre_tokenizer.whitespacewhitespace splits on whitespace using pattern \w+|[^\w\s]+.val · brot
- Brot.Pre_tokenizer.whitespace_splitwhitespace_split splits on any whitespace characters.val · brot
- Kaun.Cache_index.wholewhole ~batch ~seq () is the index of whole sequences at positions 0 to seq - 1: nothing is read and nothing is kept, so a layer given it leaves its cache as it was.val · kaun
- Tolk.Ops.wholewhole p is the parameter of the storage the Op.t.Param p views, for a view: a parameter of its own that a run of a buffer takes where a call stores into it or a loop reads it, so that the kernel's program does not depend on where the run starts.val · tolk
- Talon.Kit.widerwider ~by ~key ~value q is one row per group of q's rows that have the same by, in order of first appearance, holding by, then one column per category of key's categorical type, in dictionary order, holding value at the row of that group and category, null where there is none (Expr.pivot).val · talon
- Hugin.Role.widthwidth is a line width, in points.val · hugin
- Hugin_gg.Stroke.widthwidth s is the line width of s.val · hugin
- Rig_program.widthThe type for the widths of holes.type · rig
- Tolk.Jit.widthwidth ~beam () is the width jit_lower ~beam searches at: beam, or else the setting JITBEAM's, or else BEAM's.val · tolk
- Tolk.Ops.Upat.wildwild is v (), which matches any node.val · tolk
- Tolk.Compiler_cpu.win64win64 arch is true iff arch, as clang reads it, is x86_64's under Windows: its MACHINE-SYSTEM is x86_64-windows.val · tolk
- Kaun.Cache_index.windowwindow w index is index whose tokens see only the last w positions at or before their own, themselves included.val · kaun
- Nx_array.Move.windowThe type for windows along axis: window w's element j is the axis's element w·step + j·dilation, j < size.type · nx
- Rig_pci.WindowRanges of a machine's addresses that the process reads and writes.module · rig
- Talon.Expr.windowwindow ~on ~before ~after ~exclude_key r is, at each row, r over the row's window (Windows).val · talon
- Ymir.Grid.windowwindow ~start ~shape g is the block of shape cells of g from start, [...; 2] along g's own axes.val · ymir
- Ymir.Observation.windowwindow ~start ~shape o is o on Grid.window ~start ~shape (grid o): data, variance, validity and a per-cell area sliced with the grid, invalid beyond o's cells.val · ymir
- Rig_nv.windowswindows is 2{^40} - 8 GiB, the first of the GPU addresses at which a device's kernels see their local and shared memory, which end at 2{^40}.val · rig
- Rig_remote_proxy.WireThe protocol between the processes of a job.module · rig
- Rig.Claim.with_with_ ~read ~donate f takes claims c and is f c.val · rig
- Hugin_gg.Color.with_alphawith_alpha a c is c with opacity a.val · hugin
- Tolk.Renderer.with_compilerwith_compiler c r is r with its rendered source compiled by c.val · tolk
- Tolk.Ops.Pattern_matcher.with_ctxwith_ctx m is m, ignoring the context, for appending to matchers that read one.val · tolk
- Tolk.Shape.with_ctxwith_ctx m is Ops.Pattern_matcher.with_ctx m.val · tolk
- Hugin_kit.Scale.with_domainwith_domain d s is s with its domain set to d.val · hugin
- Norn.with_gradientwith_gradient u f is the density fun x -> fst (f x) whose gradient, row by row, is snd (f x): a tangent dx moves chain i's log density by the inner product of row i of the gradient with row i of dx, over every float tensor.val · norn
- Ymir_fits.Fits.with_headerwith_header h hdu is hdu with h's records, except the structural ones (write), which stay hdu's: its mandatory records first, each other one where h holds that keyword, or after h's records.val · ymir
- Nx.Rng.with_keywith_key k f runs f in a scope rooted at k: the sampler called nth in f without ~key, counting from 0, draws as it does with ~key:(fold_in k n).val · nx
- Talon_arrow.with_metadatawith_metadata ?column m f is f with the custom metadata m on the schema, or on the field column.val · talon
- Rig_bin.Address.with_portwith_port host port is HOST:PORT, host in brackets if it has colons.val · rig
- Tolk.Renderer.with_storagewith_storage u dt is the access u restated at dt on the storage it reaches through first sources: the Op.t.Param, Op.t.Buffer or Op.t.Alloc found from u by following first sources gets element type dt, and the nodes on the way are rebuilt on it.val · tolk
- Talon_arrow.with_typewith_type name t f is f with the column name read as t, the type sniff gives it, a declaration if it is undeclared, or, for a fixed_size_list of numbers of at least one element that reads as a tensor, a list of its elements, which may be null.val · talon
- Talon_csv.with_typewith_type name t f is f with the column name read as t.val · talon
- Talon_json.with_typewith_type name t f is f with the column name read as t.val · talon
- Talon_parquet.with_typewith_type name t f is f with the column name read as t, the type sniff gives it, a declaration if it is undeclared, or one of its alternatives.val · talon
- Brot.Encoding.with_type_idwith_type_id enc type_id is enc with every type_ids entry set to type_id.val · brot
- Munin.Session.withinwithin s f is f s, after which it finishes s as `Finished.val · munin
- Talon.Expr.withinwithin r o is the outputs o bound over r's fields, so that inside o a name reads the field of that name and no column of the frame, null where r is null or the field is.val · talon
- Tolk.Ops.without_afterwithout_after u is u without its Op.t.Afters.val · tolk
- Ymir.Cosmology.wmap1WMAP first-year (Spergel et al.val · ymir
- Ymir.Cosmology.wmap3WMAP three-year (Spergel et al.val · ymir
- Ymir.Cosmology.wmap5WMAP five-year (Komatsu et al.val · ymir
- Ymir.Cosmology.wmap7WMAP seven-year (Komatsu et al.val · ymir
- Ymir.Cosmology.wmap9WMAP nine-year (Hinshaw et al.val · ymir
- Tolk.Ops.wmmaThe type for the arguments of Op.t.Wmma.type · tolk
- Tolk.Ops.wmmawmma a b ~acc ~dims ~threads is the matrix multiply-accumulate of a and b into acc.val · tolk
- Tolk.Ops.wmma_axisThe type for an axis a tensor core folds: its range's role and identity, and its size.type · tolk
- Rig.wordword is 4, the bytes of a word.val · rig
- Rig_remote_proxy.Link.wordword l ~device v sends that the work of device up to v is done.val · rig
- Ymir_units.Vocabulary.wordThe type for words.type · ymir
- Brot.Encoding.word_idsword_ids enc maps each token to the index of the pretoken it came from, counting from 0 in each sequence.val · brot
- Brot.word_levelword_level () is a word-level tokenizer.val · brot
- Brot.Decoder.wordpiecewordpiece ~prefix ~cleanup () is a decoder for WordPiece tokens.val · brot
- Brot.wordpiecewordpiece () is a WordPiece tokenizer.val · brot
- Rig.Submission.wordswords b places the 32-bit words of b on the device's queue, such as indirect-buffer or GPFIFO entries.val · rig
- Rig_edge_primitives.wordswords () is (p, n) where p is the host address of rig's array of timeline words and n its bytes (rig_words and RIG_WORDS_BYTES), for a driver that maps the array whole.val · rig
- Tolk.WorkerParallel work.module · tolk
- Fehu.Env.wrapwrap ~observation_space ~action_space ~reset ~step inner builds a new environment that wraps inner.val · fehu
- Hugin_kit.Scale.wrapwrap s is the number of facet panels per row of s, if set.val · hugin
- Nx_array.writewrite a f is f b, where b is a bigarray of a's shape over a's own bytes: what f stores through b is a's elements.val · nx
- Rig.Buffer.writewrite b k f is f a, where a is b's bytes as elements of k, once every use of b's memory is done.val · rig
- Rig.Submission.writewrite slot ~at is a ref through which the work reads and writes the buffer slot, which must admit writes (Buffer.access).val · rig
- Rig_bin.Proc.writewrite fd s writes s on fd, ignoring the system's errors: a reader that is gone reads nothing.val · rig
- Rig_bnxt.Region.writewrite r at s writes s at offset at of r.val · rig
- Rig_edge.Io.writewrite d r ~at ~src ~len writes the len bytes of host memory at src at at in r.val · rig
- Rig_mlx5_abi.Entry.writewrite b at ~qp ~index e writes e in the size bytes of b at byte at, as entry number index of queue pair number qp.val · rig
- Rig_mlx5_abi.Receive.writewrite b at l writes the receive of up to l.bytes bytes into l in the size bytes of b at byte at.val · rig
- Rig_pci.Window.writewrite w off s is blit_string s 0 w off (String.length s).val · rig
- Ymir.Wcs.writewrite ~alt ~window t kw is the edits, in order, that make kw's description alt spell t, with window, (start, stop) per tensor axis, taking each CRPIX less its axis's start.val · ymir
- Ymir_fits.Fits.writewrite path hdus writes hdus to a new file beside path, syncs it and renames it over path, with mode 0640 on POSIX systems and writable on Windows, which has no modes.val · ymir
- Tolk.Shape.written_paramswritten_params body is the slots of the storage parameters the call body body writes, in increasing order: those a store of a Op.t.Sink or an Op.t.Store stores into, through indices, movements and Op.t.Afters, and, for a call the body makes, the parameters it passes where that call's body writes, at any depth; a call whose body does not show its writes writes every parameter it is passed.val · tolk
- Hugin_kit.Symbol.wyewye is a Y: an equilateral triangle with a square on each of its sides, one of them below it.val · hugin
x
- Hugin.Role.xx is the horizontal position.val · hugin
- Hugin_font.Run.xx r i is the x coordinate of the origin of glyph i of r.val · hugin
- Hugin_gg.P2.xx p is the x coordinate of p.val · hugin
- Hugin_gg_kit.Ring2.xx r i is the x coordinate of point i of r.val · hugin
- x-gpt2-tokenizerLoading a real GPT-2 tokenizer from HuggingFace model files.example · brot
- x-kaun-mnistMNIST training with kaun, tracked by muninexample · munin
- x-kaun-mnist-jitJitted MNIST training on a GPU, with readings of the deviceexample · munin
- x-pciNeeds Linux.example · rig
- Hugin.Role.x2x2 is the other end of a horizontal extent, on the scale of x.val · hugin
- Hugin_font.Font.x_heightx_height f is the height of flat lowercase letters above the baseline: the OS/2 value if f records a positive one, else the top of the ink of f's glyph for x if it has ink, else 0.5, as CSS assumes.val · hugin
- Tolk.Transcendental.xexp2xexp2 d is 2d: inf where it overflows d's type, 0 where it underflows past the subnormals, and NaN at NaN.val · tolk
- Tolk.Transcendental.xlog2xlog2 d is the base-2 logarithm of d, subnormals included: inf at inf, -inf at both zeros, and NaN below zero and at NaN.val · tolk
- Tolk.Transcendental.xpowxpow base exponent is base to the power exponent, computed as exp2 (exponent * log2 |base|).val · tolk
- Tolk.Transcendental.xpow_intxpow_int base exponent dt is base to the power exponent in the integer type dt, wrapping at its width as repeated products do.val · tolk
- Tolk.Transcendental.xsinxsin ~fast ~switch_over d is the sine of d, NaN at infinities and NaN.val · tolk
- Ymir.Direction.xyzxyz d is d's vectors, [...; 3]: unit vectors for values the constructors build.val · ymir
y
- Hugin.Role.yy is the vertical position.val · hugin
- Hugin_font.Run.yy r i is the y coordinate of the origin of glyph i of r.val · hugin
- Hugin_gg.P2.yy p is the y coordinate of p.val · hugin
- Hugin_gg_kit.Ring2.yy r i is the y coordinate of point i of r.val · hugin
- Hugin.Role.y2y2 is the other end of a vertical extent, on the scale of y.val · hugin
- Kaun.Rope.yarnyarn ...val · kaun
- Ymir_units.Codata.yearyear r is the year of r's adjustment.val · ymir
- Hugin_kit.Time.yearsyears k is the interval of k calendar years.val · hugin
- Hugin_kit.Scheme.ylgnylgn is Brewer's YlGn, yellow to green.val · hugin
- Hugin_kit.Scheme.ylgnbuylgnbu is Brewer's YlGnBu, yellow through green to blue.val · hugin
- Hugin_kit.Scheme.ylorbrylorbr is Brewer's YlOrBr, yellow through orange to brown.val · hugin
- Hugin_kit.Scheme.ylorrdylorrd is Brewer's YlOrRd, yellow through orange to red.val · hugin
- ymirAstronomy for OCaml, built on Nx.library · ymir
- YmirAstronomy over Nx tensors.module · ymir
- Ymir_fitsFITS files.module · ymir
- Ymir_spkSPK files: the positions of the Sun, the Moon and the planets from JPL's ephemerides.module · ymir
- Ymir_unitsExact physical units and quantities.module · ymir
- Ymir_units.Unit.yoctoyocto u is 10-24 u.val · ymir
- Jera.Split.yoshida4yoshida4 is Yoshida's (1990) composition of three leapfrogs: order 4.val · jera
- Jera.Split.yoshida6yoshida6 is Yoshida's (1990) solution A, seven leapfrogs: order 6.val · jera
- Jera.Split.yoshida8yoshida8 is Yoshida's (1990) solution D, fifteen leapfrogs: order 8.val · jera
- Ymir_units.Unit.yottayotta u is 1024 u.val · ymir
z
- Ymir_units.Unit.zeptozepto u is 10-21 u.val · ymir
- Nx_array.Dtype.zerozero dt is dt's additive identity: 0, and false for booleans.val · nx
- Tolk.Bigint.zerozero is 0.val · tolk
- Tolk.Renderer.Estimates.zerozero is the estimate of no work: every count is 0.val · tolk
- Kaun.Init.zeroszeros fills with 0.0.val · kaun
- Nx.zeroszeros dt s is the host value of shape s whose every element is zero.val · nx
- Nx.zeros_atzeros_at p dt s is zeros dt s at p.val · nx
- Nx.zeros_likezeros_like x is zeros (dtype x) (shape x).val · nx
- Ymir_units.Unit.zettazetta u is 1021 u.val · ymir
- Talon.Tz.zoneThe type for time zones.type · talon
- Hugin.View.zoomzoom ~at s is the key of the zoom of the continuous scale named like s, of the kind of s, in the scope that holds the node at (Scopes), the root by default.val · hugin
#
- Talon.Expr.($)f $ a applies f's function to a's value, once per row where no argument is null, and is null elsewhere: const mk $ carrier $ delay.val · talon
- Tolk.Dtype.Value.(%)v0 % v1 is v0 - (v0 // v1) * v1, of the sign of v1 when it is not zero: a float remainder is exact.val · tolk
- Tolk.Ops.Elementwise.O.(%)val · tolk
- Tolk.Ops.O.(%)val · tolk
- Tolk.Ops.Upat.(%)val · tolk
- Tolk.Ops.Upat.O.(%)val · tolk
- Tolk.Shape.Sint.(%)The remainder of //, with the sign of the divisor.val · tolk
- Nx.Infix.(&&)a && b is logical_and a b.val · nx
- Nx_kernel.Prog.E.(&&)x && y is logand x y, at the precedence of Stdlib.( && ), below the comparisons.val · nx
- Talon.Expr.(&&)a && b is Kleene's conjunction: false if either is false, else null if either is null.val · talon
- Talon.Join.(&&)c0 && c1 matches the pairs that both c0 and c1 match.val · talon
- Nx.Infix.(>$)x >$ c is greater_s x c.val · nx
- Ymir.Transform.(>>)t >> u applies t, then u.val · ymir
- Nx.Infix.(>)a > b is greater a b.val · nx
- Nx_kernel.Prog.E.(>)x > y is less y x.val · nx
- Talon.Expr.(>)a > b is b < a.val · talon
- Tolk.Bigint.(>)( > ) is gt.val · tolk
- Tolk.Dtype.Value.(>)v0 > v1 is v1 < v0.val · tolk
- Tolk.Ops.Elementwise.O.(>)val · tolk
- Tolk.Ops.O.(>)val · tolk
- Tolk.Ops.Upat.(>)val · tolk
- Tolk.Ops.Upat.O.(>)val · tolk
- Tolk.Shape.Sint.(>)val · tolk
- Nx.Infix.(>=$)x >=$ c is greater_equal_s x c.val · nx
- Nx.Infix.(>=)a >= b is greater_equal a b.val · nx
- Nx_kernel.Prog.E.(>=)x >= y is less_equal y x.val · nx
- Talon.Expr.(>=)a >= b is b <= a.val · talon
- Tolk.Bigint.(>=)( >= ) is geq.val · tolk
- Tolk.Dtype.Value.(>=)v0 >= v1 is v1 <= v0.val · tolk
- Tolk.Ops.Elementwise.O.(>=)val · tolk
- Tolk.Ops.O.(>=)val · tolk
- Tolk.Ops.Upat.(>=)val · tolk
- Tolk.Ops.Upat.O.(>=)val · tolk
- Tolk.Shape.Sint.(>=)val · tolk
- Nx.Infix.(<$)x <$ c is less_s x c.val · nx
- Nx.Infix.(<>$)x <>$ c is not_equal_s x c.val · nx
- Nx.Infix.(<>)a <> b is not_equal a b.val · nx
- Nx_kernel.Prog.E.(<>)x <> y is not_equal x y.val · nx
- Talon.Expr.(<>)a <> b is not (a = b).val · talon
- Tolk.Dtype.Value.(<>)v0 <> v1 is not (v0 = v1).val · tolk
- Tolk.Ops.Elementwise.O.(<>)val · tolk
- Tolk.Ops.O.(<>)val · tolk
- Tolk.Ops.Upat.(<>)val · tolk
- Tolk.Ops.Upat.O.(<>)val · tolk
- Tolk.Shape.Sint.(<>)val · tolk
- Nx.Infix.(<)a < b is less a b.val · nx
- Nx_kernel.Prog.E.(<)x < y is less x y.val · nx
- Talon.Expr.(<)a < b is true iff a comes before b in talon's total order, so x > float 15.val · talon
- Tolk.Bigint.(<)( < ) is lt.val · tolk
- Tolk.Dtype.Value.(<)v0 < v1 is true iff v0's magnitude is less than v1's.val · tolk
- Tolk.Ops.Elementwise.O.(<)val · tolk
- Tolk.Ops.O.(<)val · tolk
- Tolk.Ops.Upat.(<)val · tolk
- Tolk.Ops.Upat.O.(<)val · tolk
- Tolk.Shape.Sint.(<)val · tolk
- Nx.Infix.(<=$)x <=$ c is less_equal_s x c.val · nx
- Nx.Infix.(<=)a <= b is less_equal a b.val · nx
- Nx_kernel.Prog.E.(<=)x <= y is less_equal x y.val · nx
- Talon.Expr.(<=)a <= b is a < b || a = b.val · talon
- Tolk.Bigint.(<=)( <= ) is leq.val · tolk
- Tolk.Dtype.Value.(<=)v0 <= v1 is v0 < v1 || v0 = v1.val · tolk
- Tolk.Ops.Elementwise.O.(<=)val · tolk
- Tolk.Ops.O.(<=)val · tolk
- Tolk.Ops.Upat.(<=)val · tolk
- Tolk.Ops.Upat.O.(<=)val · tolk
- Tolk.Shape.Sint.(<=)val · tolk
- Nx.Infix.(*$)x *$ c is mul_s x c.val · nx
- Hugin_gg.Affine.(*)m * n is the map applying n first, then m.val · hugin
- Nx.Infix.(*)a * b is mul a b.val · nx
- Nx_kernel.Prog.E.(*)x * y is mul x y.val · nx
- Talon.Expr.(*)a * b is the product of a and b.val · talon
- Tolk.Bigint.(*)( * ) is mul.val · tolk
- Tolk.Dtype.Value.(*)v0 * v1 is the product of v0 and v1.val · tolk
- Tolk.Ops.Elementwise.O.(*)val · tolk
- Tolk.Ops.O.(*)val · tolk
- Tolk.Ops.Upat.(*)val · tolk
- Tolk.Ops.Upat.O.(*)val · tolk
- Tolk.Shape.Sint.(*)val · tolk
- Ymir_units.Unit.(*)u * w is the product of u and w.val · ymir
- Nx.Infix.(**$)x **$ c is pow_s x c.val · nx
- Nx.Infix.(**)a ** b is pow a b.val · nx
- Nx_kernel.Prog.E.(**)x ** y is pow x y.val · nx
- Talon.Expr.(**)a ** b is a to the power b.val · talon
- Ymir_units.Unit.(**)u ** n is u to the power n.val · ymir
- Talon.Expr.(*.)a *.val · talon
- Nx.Infix.(*@)a *@ b is matmul a b.val · nx
- Nx.Infix.(+$)x +$ c is add_s x c.val · nx
- Nx.Infix.(+)a + b is add a b.val · nx
- Nx_kernel.Prog.E.(+)x + y is add x y.val · nx
- Talon.Expr.(+)a + b is the sum of a and b, wrapping on overflow as nx's integers do.val · talon
- Talon.Sel.(+)s0 + s1 selects s0's columns, then s1's columns not in s0.val · talon
- Tolk.Bigint.(+)( + ) is add.val · tolk
- Tolk.Dtype.Value.(+)v0 + v1 is the sum of v0 and v1.val · tolk
- Tolk.Ops.Elementwise.O.(+)val · tolk
- Tolk.Ops.O.(+)val · tolk
- Tolk.Ops.Upat.(+)val · tolk
- Tolk.Ops.Upat.O.(+)val · tolk
- Tolk.Shape.Sint.(+)val · tolk
- Talon.Expr.(+.)a +.val · talon
- Nx.Infix.(-$)x -$ c is sub_s x c.val · nx
- Nx.Infix.(-)a - b is sub a b.val · nx
- Nx_kernel.Prog.E.(-)x - y is sub x y.val · nx
- Talon.Expr.(-)a - b is the difference of a and b.val · talon
- Talon.Sel.(-)s0 - s1 selects s0's columns that are not in s1, in s0's order.val · talon
- Tolk.Bigint.(-)( - ) is sub.val · tolk
- Tolk.Dtype.Value.(-)v0 - v1 is the difference of v0 and v1.val · tolk
- Tolk.Ops.Elementwise.O.(-)val · tolk
- Tolk.Ops.O.(-)val · tolk
- Tolk.Ops.Upat.(-)val · tolk
- Tolk.Ops.Upat.O.(-)val · tolk
- Tolk.Shape.Sint.(-)val · tolk
- Talon.Expr.(-.)a -.val · talon
- Nx.Infix.(.${})x.${idx} is slice idx x.val · nx
- Nx.Infix.(.%{})x.%{p} is get p x.val · nx
- Nx.Infix.(/$)x /$ c is div_s x c.val · nx
- Nx.Infix.(/)a / b is div a b.val · nx
- Nx_kernel.Prog.E.(/)x / y is idiv x y on integers and fdiv x y on other dtypes.val · nx
- Talon.Expr.(/)a / b is the quotient of a by b, truncated toward zero, and null where b is zero.val · talon
- Tolk.Bigint.(/)( / ) is div.val · tolk
- Tolk.Ops.Elementwise.O.(/)val · tolk
- Tolk.Ops.O.(/)val · tolk
- Tolk.Ops.Upat.(/)val · tolk
- Tolk.Ops.Upat.O.(/)val · tolk
- Ymir_units.Unit.(/)u / w is the quotient of u by w.val · ymir
- Talon.Expr.(/.)a /.val · talon
- Tolk.Dtype.Value.(//)v0 // v1 is the quotient of v0 by v1 rounded down.val · tolk
- Tolk.Ops.Elementwise.O.(//)val · tolk
- Tolk.Ops.O.(//)val · tolk
- Tolk.Ops.Upat.(//)val · tolk
- Tolk.Ops.Upat.O.(//)val · tolk
- Tolk.Shape.Sint.(//)Division rounding towards negative infinity.val · tolk
- Talon.Expr.(:=)name := e outputs e as the column name.val · talon
- Nx.Infix.(=$)x =$ c is equal_s x c.val · nx
- Nx.Infix.(=)a = b is equal a b.val · nx
- Nx_kernel.Prog.E.(=)x = y is equal x y.val · nx
- Talon.Expr.(=)a = b is true iff a and b are equal in talon's total order (Type.compare_value), so nan = nan and -0.val · talon
- Tolk.Bigint.(=)( = ) is equal.val · tolk
- Tolk.Dtype.Value.(=)v0 = v1 is true iff v0 and v1 have the same magnitude.val · tolk
- Nx.Ptree.(@->)a @-> f takes an argument of a, then is f.val · nx
- Nx.Infix.(land)a land b is bitwise_and a b.val · nx
- Nx_kernel.Prog.E.(land)x land y is logand x y.val · nx
- Tolk.Ops.Elementwise.O.(land)val · tolk
- Tolk.Ops.O.(land)val · tolk
- Tolk.Ops.Upat.(land)val · tolk
- Tolk.Ops.Upat.O.(land)val · tolk
- Nx.Infix.(lor)a lor b is bitwise_or a b.val · nx
- Nx_kernel.Prog.E.(lor)x lor y is logor x y.val · nx
- Tolk.Ops.Elementwise.O.(lor)val · tolk
- Tolk.Ops.O.(lor)val · tolk
- Tolk.Ops.Upat.(lor)val · tolk
- Tolk.Ops.Upat.O.(lor)val · tolk
- Tolk.Bigint.(lsl)( lsl ) is shift_left.val · tolk
- Tolk.Ops.Elementwise.O.(lsl)val · tolk
- Tolk.Ops.O.(lsl)val · tolk
- Tolk.Ops.Upat.(lsl)val · tolk
- Tolk.Ops.Upat.O.(lsl)val · tolk
- Tolk.Ops.Elementwise.O.(lsr)val · tolk
- Tolk.Ops.O.(lsr)val · tolk
- Tolk.Ops.Upat.(lsr)val · tolk
- Tolk.Ops.Upat.O.(lsr)val · tolk
- Nx.Infix.(lxor)a lxor b is bitwise_xor a b.val · nx
- Nx_kernel.Prog.E.(lxor)x lxor y is logxor x y.val · nx
- Tolk.Ops.Elementwise.O.(lxor)val · tolk
- Tolk.Ops.O.(lxor)val · tolk
- Tolk.Ops.Upat.(lxor)val · tolk
- Tolk.Ops.Upat.O.(lxor)val · tolk
- Nx_kernel.Prog.E.(mod)x mod y is rem x y.val · nx
- Talon.Expr.(mod)a mod b is the remainder of a by b, of a's sign, and null where b is zero.val · talon
- Nx.Infix.(||)a || b is logical_or a b, as ( && ).val · nx
- Nx_kernel.Prog.E.(||)x || y is logor x y, at the precedence of Stdlib.( || ).val · nx
- Talon.Expr.(||)a || b is Kleene's disjunction: true if either is true, else null if either is null.val · talon
- Nx.Infix.(~-)-x is neg x.val · nx
- Nx_kernel.Prog.E.(~-)~- x is neg x.val · nx
- Tolk.Dtype.Value.(~-)-v is the negation of v.val · tolk
- Tolk.Ops.Elementwise.O.(~-)val · tolk
- Tolk.Ops.O.(~-)val · tolk
- Tolk.Ops.Upat.(~-)val · tolk
- Tolk.Ops.Upat.O.(~-)val · tolk
- 01-addA kernel is a function that a device runs once for each thread of a grid.example · rig
- 01-basicStarts a run with params, logs a loss for 100 steps, writes a summary, and prints the run's id.example · munin
- 01-basic-optimizersYour first optimizer.example · vega
- 01-creating-arraysBuild arrays from scratch: constants, ranges, grids, and custom data.example · nx
- 01-distributionsA distribution describes one whole tensor value.example · norn
- 01-encode-decodeYour first tokenizer.example · brot
- 01-gradient-descentexample · rune
- 01-grayscaleConvert a color image to grayscale using Sowilo.to_grayscale and render the original and result side by side to grayscale.png with Hugin.example · sowilo
- 01-lineLines over a step axis, one per seed, coloured by a dim, in the default theme and in Theme.dark.example · hugin
- 01-random-agentexample · fehu
- 01-rootsRoot finders are elementwise: each element of the input is its own problem with its own status.example · jera
- 01-tablesBuild a table from OCaml arrays, read a column back, and run a first query.example · talon
- 01-unitsA unit is an exact value with one canonical text, and a quantity is a tensor in a unit.example · ymir
- 01-xorexample · kaun
- 02-bijectorsA bijector maps unconstrained coordinates onto a support and reports its log-determinant.example · norn
- 02-buffersA buffer is bytes of one device's memory, and the caller gives them their meaning.example · rig
- 02-constants-and-namesThe SI's defining constants are exact units.example · ymir
- 02-encoding-fieldsUnderstanding encodings.example · brot
- 02-flightsThe pipeline of the guide: read flights from a CSV file and carriers from a Parquet file, keep the late departures, average them per carrier, join each carrier's name and sort.example · talon
- 02-gaussian-blurApply a Gaussian blur with Sowilo.gaussian_blur to smooth a grayscale image.example · sowilo
- 02-infix-and-arithmeticElement-wise math with operators: inside Nx.Infix.( … ), array code reads like algebra.example · nx
- 02-metricsGoals, an x axis of another metric, series and statistics read backexample · munin
- 02-mnistexample · kaun
- 02-per-sample-gradsexample · rune
- 02-q-learningexample · fehu
- 02-scatterDots coloured by category and sized by a quantity, with two legends.example · hugin
- 02-systemsA system maps a value to a value of the same structure.example · jera
- 02-training-stepsFit a linear model y = x .example · vega
- 03-artifactsVersions, aliases, and which run produced and used each oneexample · munin
- 03-barsBars: a rect over a band scale, its length from zero.example · hugin
- 03-expressionsExpressions over a table of training losses: values per row, reductions per group, and reductions put back on the rows with over.example · talon
- 03-filesThe disk is a device whose buffers are files.example · rig
- 03-frames-and-directionsA direction is a batch of unit vectors in a celestial frame, and the frame is part of its type.example · ymir
- 03-hessianexample · rune
- 03-indexing-and-slicingSelect, slice, and mask: extract exactly the data you need.example · nx
- 03-learning-rate-schedulesExplore learning rate schedules.example · vega
- 03-median-blurApply a median filter with Sowilo.median_blur for noise removal.example · sowilo
- 03-minimizationGradient methods take rune's gradient of a plain objective.example · jera
- 03-mnist-cnnexample · kaun
- 03-normalizersText normalization before tokenization.example · brot
- 03-nutsA density over a record of your own, written by hand, sampled with NUTS over four chains, then diagnosed and summarised.example · norn
- 03-reinforceexample · fehu
- 04-dqnexample · fehu
- 04-gpt2example · kaun
- 04-heatmapA matrix as cells, its rows and columns read with dim, and a colour bar.example · hugin
- 04-joinsJoin trials with subjects on subject = id with each kind of join, assert that every trial has exactly one subject, then join trial windows with the events inside them.example · talon
- 04-linear-systemsA linear system is given by its operator's product, never its matrix.example · jera
- 04-mediaFiles a run logs at a step, found through its historyexample · munin
- 04-modelsnorn.model turns one function that draws every random variable into densities, starting points, simulations and predictions.example · norn
- 04-pre-tokenizersPre-tokenization: splitting text into fragments before vocabulary lookup.example · brot
- 04-reshaping-and-broadcastingChange array shapes and let broadcasting align dimensions automatically.example · nx
- 04-thresholdBinarize a grayscale image with Sowilo.threshold.example · sowilo
- 04-timelinesA device numbers the work submitted to it 1, 2, 3, ..., the values of its timeline.example · rig
- 04-transformsA transform maps points through a list of stages: pixel coordinates to a plane, a plane to the sky.example · ymir
- 05-algorithmsFive tokenization algorithms compared side-by-side.example · brot
- 05-facetsSmall multiples: one panel per category of fx, sharing every scale.example · hugin
- 05-fits-imagesA FITS file is a list of HDUs, each a header and a data unit.example · ymir
- 05-formatsRead CSV text with a sniffed format, change a column's type, write the table back, and read a Parquet file whose row groups the filter can skip.example · talon
- 05-hmc-and-ensembleTwo other samplers on a correlated Gaussian: HMC, which tunes one step size, trajectory length and geometry for all chains, and the ensemble stretch move, which needs no gradient and tunes nothing.example · norn
- 05-llamaexample · kaun
- 05-parameter-sweepRuns whose parent is a sweep, compared by their best errorexample · munin
- 05-quadratureIntegrals over ranges are elementwise.example · jera
- 05-reductions-and-statisticsSummarize data with reductions: means, variances, and aggregations along any axis.example · nx
- 05-sobelCompute horizontal and vertical edge gradients with Sowilo.sobel.example · sowilo
- 05-stampsEach allocation records the point of its last write and, for each device, the point of its last use.example · rig
- 06-cannyDetect edges with the Canny algorithm using Sowilo.canny.example · sowilo
- 06-confusion-matrixCells coloured by recall, their counts in a contrasting colour (map_range).example · hugin
- 06-diagnosticsChains that start in different modes of a bimodal target never meet.example · norn
- 06-fits-tablesA binary table holds rows of typed columns.example · ymir
- 06-gpt-ossexample · kaun
- 06-inspectRuns, status, provenance, metrics and artifacts read from a storeexample · munin
- 06-interpolationSplines, fits and their derivatives and integrals are one value, a piecewise Chebyshev series.example · jera
- 06-mistakesThe two ways a query fails.example · talon
- 06-random-numbersRandom numbers from keys: distributions, sampling, and shuffling.example · nx
- 06-special-tokensSpecial tokens and post-processing.example · brot
- 06-stepsCompiled work is made once and run many times.example · rig
- 07-attentionAttention maps faceted by layer and head, with one colour bar.example · hugin
- 07-evidenceThe evidence is the probability of the data under the model.example · norn
- 07-featuresTurn flights into a feature matrix and labels for a model.example · talon
- 07-fits-wcsWcs.read turns a header's celestial keywords into a transform from pixels to the sky, asking for each keyword's value as ymir.fits decodes it.example · ymir
- 07-linear-algebraSolve systems, decompose matrices, and fit models: linear algebra made practical.example · nx
- 07-morphologyApply morphological erosion and dilation with Sowilo.erode and Sowilo.dilate using a rectangular structuring element.example · sowilo
- 07-neural-odeexample · kaun
- 07-odesA problem is a field and an initial state of any structure.example · jera
- 07-padding-truncationPadding and truncation for batch processing.example · brot
- 07-reclamationNothing frees a buffer by hand.example · rig
- 07-system-monitorReadings of the machine and the process, sampled into a runexample · munin
- 08-claimsA function handed buffers may write over an input in place only if nothing else reads that memory.example · rig
- 08-dashboardThe dashboard over a sweep and a run that trains while it showsexample · munin
- 08-decodersDecoders convert token strings back to natural text.example · brot
- 08-embeddingAn embedding of 3,000 dots, and of 100,000 drawn as one image.example · hugin
- 08-grids-and-regionsA grid is an image's cells seen through a transform; a region is a shape placed on the grid's world.example · ymir
- 08-laplaceA Laplace approximation is the Gaussian whose precision is the curvature at the mode.example · norn
- 08-signal-processingAnalyze frequencies with FFT: decompose signals and filter noise.example · nx
- 08-stiff-and-daeOn a stiff field an explicit method spends its budget; Ode.kvaerno5 steps at the tolerance's pace.example · jera
- 08-text-and-timeText and temporal expressions over a table of recording sessions.example · talon
- 09-calibrationRanks of prior truths among posterior draws are uniform when the posterior is right.example · norn
- 09-image-gridImages in wrapped panels, captioned, with the mistakes framed.example · hugin
- 09-image-processingLoad, transform, and save images as arrays: convolutions, pooling, and pixel math.example · nx
- 09-lossA device whose work fails is lost, once and for good.example · rig
- 09-many-columnsWork on many columns at once with selectors, and use the compositions of Kit.example · talon
- 09-observationsAn observation is data on a grid with a variance and a mask of valid samples.example · ymir
- 09-sdesA problem is a drift, a diffusion and a Brownian path.example · jera
- 09-trainingTraining tokenizers from scratch.example · brot
- 10-bert-pipelineComplete BERT-style tokenizer pipeline.example · brot
- 10-cosmologyA cosmology is a record of tensors, and distances, volumes and times are functions of it and of redshifts.example · ymir
- 10-custom-sourceWrite a source of your own and fold a query over its batches.example · talon
- 10-hmmA hidden Markov model's states are discrete, so a gradient sampler cannot move them.example · norn
- 10-loss-landscapeFilled contours on a log scale and an optimiser's path.example · hugin
- 10-machine-learningFour classic ML algorithms built from Nx primitives: SVD, broadcasting, reductions, and scalar loops.example · nx
- 10-profilesA profile holds what every device did while a function ran.example · rig
- 10-splittingFor a separable Hamiltonian, a splitting composes the exact flows of its parts.example · jera
- 11-compile-and-batchEvery jera program runs unchanged under rune.example · jera
- 11-dashboardA training dashboard: loss on a log scale over epochs, and validation accuracy, sharing x.example · hugin
- 11-derivativesEvery ymir function is a formula of nx operations, so rune differentiates and compiles it.example · ymir
- 11-elfHost programs, GPU programs and firmware come as ELF objects.example · rig
- 11-nested-columnsColumns of lists, tensors, records and an extension type.example · talon
- 12-host-programsHost code is a C function compiled to an ELF object, loaded as an image on the host and launched as a GPU's kernel is.example · rig
- 12-paper-figureA two-column paper figure at 8 points, panels labelled (a) to (d).example · hugin
- 13-histogramTwo histograms over shared bins from Stats.histogram, drawn as rects from their edges.example · hugin
- 13-poolOne pool of threads per process computes on the host, shared by nx's kernels, host programs and a library's own C code.example · rig
- 14-areaA band between the lowest and highest of eight seeds (area with y2), under their mean.example · hugin
- 14-machinesA job is one process, the controller, and an agent on each other machine, which opens that machine's devices and runs the work the controller submits.example · rig
- 15-errorbarsError bars as a composition: a rule from y to y2 under each dot.example · hugin
- 16-stylesSeries told apart by colour and dash from one scale, a dotted rule, a frame and a legend inside the panel.example · hugin
- 17-calibrationA calibration plot over the dashed diagonal abline y = x, its axes in percent.example · hugin