| 1 | ;;; GNU Guix --- Functional package management for GNU |
| 2 | ;;; Copyright © 2015-2026 Ricardo Wurmus <[email protected]> |
| 3 | ;;; Copyright © 2016, 2020-2024 Efraim Flashner <[email protected]> |
| 4 | ;;; Copyright © 2016, 2017, 2020 Marius Bakke <[email protected]> |
| 5 | ;;; Copyright © 2016 Hartmut Goebel <[email protected]> |
| 6 | ;;; Copyright © 2018, 2019, 2020 Tobias Geerinckx-Rice <[email protected]> |
| 7 | ;;; Copyright © 2018 Mark Meyer <[email protected]> |
| 8 | ;;; Copyright © 2018 Ben Woodcroft <[email protected]> |
| 9 | ;;; Copyright © 2018 Fis Trivial <[email protected]> |
| 10 | ;;; Copyright © 2018 Julien Lepiller <[email protected]> |
| 11 | ;;; Copyright © 2018 Björn Höfling <[email protected]> |
| 12 | ;;; Copyright © 2019 Nicolas Goaziou <[email protected]> |
| 13 | ;;; Copyright © 2019, 2020 Guillaume Le Vaillant <[email protected]> |
| 14 | ;;; Copyright © 2019 Brett Gilio <[email protected]> |
| 15 | ;;; Copyright © 2020 Konrad Hinsen <[email protected]> |
| 16 | ;;; Copyright © 2020 Edouard Klein <[email protected]> |
| 17 | ;;; Copyright © 2020-2025 Vinicius Monego <[email protected]> |
| 18 | ;;; Copyright © 2020, 2021, 2022, 2023 Maxim Cournoyer <[email protected]> |
| 19 | ;;; Copyright © 2022-2025 Nicolas Graves <[email protected]> |
| 20 | ;;; Copyright © 2022 Kiran Shila <[email protected]> |
| 21 | ;;; Copyright © 2022 Wiktor Zelazny <[email protected]> |
| 22 | ;;; Copyright © 2023 zamfofex <[email protected]> |
| 23 | ;;; Copyright © 2023 Navid Afkhami <[email protected]> |
| 24 | ;;; Copyright © 2023 Zheng Junjie <[email protected]> |
| 25 | ;;; Copyright © 2023 Troy Figiel <[email protected]> |
| 26 | ;;; Copyright © 2024-2026 Sharlatan Hellseher <[email protected]> |
| 27 | ;;; Copyright © 2024 David Pflug <[email protected]> |
| 28 | ;;; Copyright © 2024 Timothee Mathieu <[email protected]> |
| 29 | ;;; Copyright © 2024 Spencer King <[email protected]> |
| 30 | ;;; Copyright © 2024, 2025 David Elsing <[email protected]> |
| 31 | ;;; Copyright © 2024, 2025 Andy Tai <[email protected]> |
| 32 | ;;; Copyright © 2025 Lapearldot <[email protected]> |
| 33 | ;;; Copyright © 2025, 2026 Cayetano Santos <[email protected]> |
| 34 | ;;; Copyright © 2025 Janneke Nieuwenhuizen <[email protected]> |
| 35 | ;;; Copyright © 2025 Romain Garbage <[email protected]> |
| 36 | ;;; Copyright © 2026 Nguyễn Gia Phong <[email protected]> |
| 37 | ;;; Copyright © 2026 Goran Vukoman <[email protected]> |
| 38 | ;;; Copyright © 2026 bdunahu <[email protected]> |
| 39 | ;;; |
| 40 | ;;; This file is part of GNU Guix. |
| 41 | ;;; |
| 42 | ;;; GNU Guix is free software; you can redistribute it and/or modify it |
| 43 | ;;; under the terms of the GNU General Public License as published by |
| 44 | ;;; the Free Software Foundation; either version 3 of the License, or (at |
| 45 | ;;; your option) any later version. |
| 46 | ;;; |
| 47 | ;;; GNU Guix is distributed in the hope that it will be useful, but |
| 48 | ;;; WITHOUT ANY WARRANTY; without even the implied warranty of |
| 49 | ;;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 50 | ;;; GNU General Public License for more details. |
| 51 | ;;; |
| 52 | ;;; You should have received a copy of the GNU General Public License |
| 53 | ;;; along with GNU Guix. If not, see <http://www.gnu.org/licenses/>. |
| 54 | |
| 55 | (define-module (gnu packages machine-learning) |
| 56 | #:use-module (gnu packages algebra) |
| 57 | #:use-module (gnu packages assembly) |
| 58 | #:use-module (gnu packages audio) |
| 59 | #:use-module (gnu packages autotools) |
| 60 | #:use-module (gnu packages base) |
| 61 | #:use-module (gnu packages bash) |
| 62 | #:use-module (gnu packages bdw-gc) |
| 63 | #:use-module (gnu packages boost) |
| 64 | #:use-module (gnu packages build-tools) |
| 65 | #:use-module (gnu packages c) |
| 66 | #:use-module (gnu packages calendar) |
| 67 | #:use-module (gnu packages check) |
| 68 | #:use-module (gnu packages cmake) |
| 69 | #:use-module (gnu packages compression) |
| 70 | #:use-module (gnu packages cpp) |
| 71 | #:use-module (gnu packages cran) |
| 72 | #:use-module (gnu packages curl) |
| 73 | #:use-module (gnu packages databases) |
| 74 | #:use-module (gnu packages dejagnu) |
| 75 | #:use-module (gnu packages digest) |
| 76 | #:use-module (gnu packages dns) |
| 77 | #:use-module (gnu packages documentation) |
| 78 | #:use-module (gnu packages freedesktop) |
| 79 | #:use-module (gnu packages gcc) |
| 80 | #:use-module (gnu packages gettext) |
| 81 | #:use-module (gnu packages gl) |
| 82 | #:use-module (gnu packages glib) |
| 83 | #:use-module (gnu packages graphviz) |
| 84 | #:use-module (gnu packages gstreamer) |
| 85 | #:use-module (gnu packages guile) |
| 86 | #:use-module (gnu packages haskell-xyz) |
| 87 | #:use-module (gnu packages image) |
| 88 | #:use-module (gnu packages image-processing) |
| 89 | #:use-module (gnu packages imagemagick) |
| 90 | #:use-module (gnu packages jupyter) |
| 91 | #:use-module (gnu packages language) |
| 92 | #:use-module (gnu packages libedit) |
| 93 | #:use-module (gnu packages libevent) |
| 94 | #:use-module (gnu packages libffi) |
| 95 | #:use-module (gnu packages linux) |
| 96 | #:use-module (gnu packages llvm) |
| 97 | #:use-module (gnu packages logging) |
| 98 | #:use-module (gnu packages maths) |
| 99 | #:use-module (gnu packages monitoring) |
| 100 | #:use-module (gnu packages mpi) |
| 101 | #:use-module (gnu packages nss) |
| 102 | #:use-module (gnu packages ocaml) |
| 103 | #:use-module (gnu packages onc-rpc) |
| 104 | #:use-module (gnu packages oneapi) |
| 105 | #:use-module (gnu packages opencl) |
| 106 | #:use-module (gnu packages parallel) |
| 107 | #:use-module (gnu packages pdf) |
| 108 | #:use-module (gnu packages perl) |
| 109 | #:use-module (gnu packages pkg-config) |
| 110 | #:use-module (gnu packages pretty-print) |
| 111 | #:use-module (gnu packages protobuf) |
| 112 | #:use-module (gnu packages pulseaudio) |
| 113 | #:use-module (gnu packages python) |
| 114 | #:use-module (gnu packages python-build) |
| 115 | #:use-module (gnu packages python-check) |
| 116 | #:use-module (gnu packages python-crypto) |
| 117 | #:use-module (gnu packages python-science) |
| 118 | #:use-module (gnu packages python-web) |
| 119 | #:use-module (gnu packages python-xyz) |
| 120 | #:use-module (gnu packages qt) |
| 121 | #:use-module (gnu packages rdf) |
| 122 | #:use-module (gnu packages regex) |
| 123 | #:use-module (gnu packages rocm) |
| 124 | #:use-module (gnu packages rpc) |
| 125 | #:use-module (gnu packages rust) |
| 126 | #:use-module (gnu packages sdl) |
| 127 | #:use-module (gnu packages serialization) |
| 128 | #:use-module (gnu packages sphinx) |
| 129 | #:use-module (gnu packages sqlite) |
| 130 | #:use-module (gnu packages ssh) |
| 131 | #:use-module (gnu packages statistics) |
| 132 | #:use-module (gnu packages swig) |
| 133 | #:use-module (gnu packages time) |
| 134 | #:use-module (gnu packages tls) |
| 135 | #:use-module (gnu packages valgrind) |
| 136 | #:use-module (gnu packages version-control) |
| 137 | #:use-module (gnu packages video) |
| 138 | #:use-module (gnu packages vulkan) |
| 139 | #:use-module (gnu packages web) |
| 140 | #:use-module (gnu packages xdisorg) |
| 141 | #:use-module (gnu packages xml) |
| 142 | #:use-module (gnu packages xorg) |
| 143 | #:use-module (gnu packages) |
| 144 | #:use-module (guix build-system cargo) |
| 145 | #:use-module (guix build-system cmake) |
| 146 | #:use-module (guix build-system copy) |
| 147 | #:use-module (guix build-system gnu) |
| 148 | #:use-module (guix build-system ocaml) |
| 149 | #:use-module (guix build-system pyproject) |
| 150 | #:use-module (guix build-system r) |
| 151 | #:use-module (guix build-system trivial) |
| 152 | #:use-module (guix download) |
| 153 | #:use-module (guix gexp) |
| 154 | #:use-module (guix git-download) |
| 155 | #:use-module ((guix licenses) #:prefix license:) |
| 156 | #:use-module (guix packages) |
| 157 | #:use-module (guix svn-download) |
| 158 | #:use-module (guix utils) |
| 159 | #:use-module (ice-9 match)) |
| 160 | |
| 161 | (define-public dlpack |
| 162 | (package |
| 163 | (name "dlpack") |
| 164 | (version "1.2") |
| 165 | (source |
| 166 | (origin |
| 167 | (method git-fetch) |
| 168 | (uri (git-reference |
| 169 | (url "https://github.com/dmlc/dlpack") |
| 170 | (commit (string-append "v" version)))) |
| 171 | (file-name (git-file-name name version)) |
| 172 | (sha256 |
| 173 | (base32 "1zm28s4das0isa5rrh5a1c6iv1k1mcjqgnihbva74s78d52a7hpn")))) |
| 174 | (build-system cmake-build-system) |
| 175 | (arguments (list #:tests? #f)) ;No tests. |
| 176 | (home-page "https://dmlc.github.io/dlpack/latest/") |
| 177 | (synopsis "In memory tensor structure") |
| 178 | (description |
| 179 | "DLPack is an in-memory tensor structure for sharing tensors among |
| 180 | frameworks.") |
| 181 | (license license:asl2.0))) |
| 182 | |
| 183 | (define-public fasttext |
| 184 | (package |
| 185 | (name "fasttext") |
| 186 | (version "0.9.2") |
| 187 | (source |
| 188 | (origin |
| 189 | (method git-fetch) |
| 190 | (uri (git-reference |
| 191 | (url "https://github.com/facebookresearch/fastText") |
| 192 | (commit (string-append "v" version)))) |
| 193 | (file-name (git-file-name name version)) |
| 194 | (sha256 |
| 195 | (base32 "07cz2ghfq6amcljaxpdr5chbd64ph513y8zqmibfx2xwfp74xkhn")))) |
| 196 | (build-system cmake-build-system) |
| 197 | ;; Tests require downloading of test data. |
| 198 | (arguments |
| 199 | (list |
| 200 | #:tests? #false |
| 201 | #:phases |
| 202 | #~(modify-phases %standard-phases |
| 203 | (add-after 'unpack 'gcc-14-fix |
| 204 | (lambda _ |
| 205 | (substitute* "src/args.cc" |
| 206 | (("#include <unordered_map>" all) |
| 207 | (string-append all "\n#include <cstdint>")))))))) |
| 208 | (home-page "https://github.com/facebookresearch/fastText") |
| 209 | (synopsis "Library for fast text representation and classification") |
| 210 | (description "fastText is a library for efficient learning of word |
| 211 | representations and sentence classification.") |
| 212 | (license license:expat))) |
| 213 | |
| 214 | (define-public ggml |
| 215 | (package |
| 216 | (name "ggml") |
| 217 | (version "0.25.2") ;sync with llama.cpp |
| 218 | (source |
| 219 | (origin |
| 220 | (method git-fetch) |
| 221 | (uri (git-reference |
| 222 | (url "https://github.com/ggml-org/ggml") |
| 223 | (commit (string-append "v" version)))) |
| 224 | (file-name (git-file-name name version)) |
| 225 | (sha256 |
| 226 | (base32 "0ww5wk2d6ldy2l5fxh4nr259mnwdpypp316h383as46cbvywyh60")))) |
| 227 | (build-system cmake-build-system) |
| 228 | (arguments |
| 229 | (list |
| 230 | #:configure-flags |
| 231 | #~(list #$(string-append "-DGGML_BUILD_COMMIT=" version) |
| 232 | "-DBUILD_SHARED_LIBS_DEFAULT=ON" |
| 233 | "-DGGML_BACKEND_DL=ON" |
| 234 | (string-append "-DGGML_BACKEND_DIR=" |
| 235 | #$output "/lib/backends") |
| 236 | "-DGGML_VULKAN=ON" |
| 237 | "-DGGML_BUILD_TESTS=ON" |
| 238 | "-DGGML_OPENCL=ON" |
| 239 | ;; Defaults to ON upstream; ggml-opencl.cpp's own |
| 240 | ;; ggml_opencl_is_device_supported() unconditionally |
| 241 | ;; rejects every non-Adreno (non-Qualcomm) device when |
| 242 | ;; this is compiled in, regardless of how capable the |
| 243 | ;; device actually is. |
| 244 | "-DGGML_OPENCL_USE_ADRENO_KERNELS=OFF" |
| 245 | "-DGGML_NATIVE_DEFAULT=OFF" ;no '-march=native' |
| 246 | "-DGGML_BLAS_DEFAULT=ON" |
| 247 | "-DGGML_BLAS_VENDOR_DEFAULT=OpenBLAS"))) |
| 248 | (inputs |
| 249 | (list glslang |
| 250 | openblas |
| 251 | openblas-ilp64 |
| 252 | opencl-icd-loader |
| 253 | shaderc |
| 254 | spirv-headers |
| 255 | spirv-tools |
| 256 | vulkan-headers |
| 257 | vulkan-loader)) |
| 258 | (native-inputs |
| 259 | (list opencl-headers pkg-config python-minimal-wrapper)) |
| 260 | (properties '((tunable? . #true))) |
| 261 | (home-page "https://github.com/ggml-org/ggml") |
| 262 | (synopsis "Tensor library for machine learning") |
| 263 | (description "@code{ggml} is a @acronym{ML, Machine Learning} library |
| 264 | written in C and C++ with a focus on transformer inference, similar to ML |
| 265 | libraries such as PyTorch and TensorFlow.") |
| 266 | (license license:expat))) |
| 267 | |
| 268 | (define-public ggml-for-whisper |
| 269 | (package |
| 270 | (inherit ggml) |
| 271 | (name "ggml-for-whisper") |
| 272 | (version "0.15.1") ;sync whisper.cpp |
| 273 | (source |
| 274 | (origin |
| 275 | (method git-fetch) |
| 276 | (uri (git-reference |
| 277 | (url "https://github.com/ggml-org/ggml") |
| 278 | (commit (string-append "v" version)))) |
| 279 | (file-name (git-file-name name version)) |
| 280 | (sha256 |
| 281 | (base32 "06bw4dbnzmy9wz3k6g4a2vf7yjn7xpqzam4ra6mxv3lkvhz0fhsg")))))) |
| 282 | |
| 283 | (define-public git-ai |
| 284 | (package |
| 285 | (name "git-ai") |
| 286 | (version "1.1.18") |
| 287 | (source (origin |
| 288 | (method git-fetch) |
| 289 | (uri (git-reference |
| 290 | (url "https://github.com/git-ai-project/git-ai") |
| 291 | (commit (string-append "v" version)))) |
| 292 | (file-name (git-file-name name version)) |
| 293 | (sha256 |
| 294 | (base32 "0yf864shgi5b35s0f6frds0lf9jwwn8qw4pxiniplhnxadrbpx2f")))) |
| 295 | (build-system cargo-build-system) |
| 296 | (arguments |
| 297 | (list #:install-source? #f |
| 298 | #:cargo-test-flags |
| 299 | ''("--" |
| 300 | ;;This test requires network access. |
| 301 | "--skip=test_load_ai_touched_files_for_specific_commits" |
| 302 | ;;This test fails in the build environment. |
| 303 | "--skip=exit_status_was_interrupted_on_sigint") |
| 304 | #:phases |
| 305 | #~(modify-phases %standard-phases |
| 306 | (add-after 'unpack 'patch-/bin/sh |
| 307 | (lambda _ |
| 308 | (substitute* (list "tests/integration/continue_session.rs" |
| 309 | "tests/integration/diff.rs" |
| 310 | "tests/integration/stats.rs" |
| 311 | "tests/integration/status_ignore.rs" |
| 312 | "tests/integration/git_ai_hooks.rs" |
| 313 | "tests/integration/hook_forwarding.rs" |
| 314 | "tests/integration/hook_modes.rs" |
| 315 | "tests/integration/rebase.rs") |
| 316 | (("#!/bin/sh") (string-append "#!" (which "sh")))))) |
| 317 | (add-after 'unpack 'set-home |
| 318 | (lambda _ |
| 319 | (setenv "HOME" (getenv "TMPDIR")))) |
| 320 | (add-before 'check 'setup-git-path |
| 321 | (lambda* (#:key inputs #:allow-other-keys) |
| 322 | (let ((git (assoc-ref inputs "git-minimal")) |
| 323 | (home (getenv "HOME"))) |
| 324 | (mkdir-p (string-append home "/.git-ai")) |
| 325 | (let ((port (open-file |
| 326 | (string-append home "/.git-ai/config.json") |
| 327 | "a"))) |
| 328 | (format port "{\"git_path\": \"~a/bin/git\"}" git) |
| 329 | (close port)) |
| 330 | |
| 331 | (substitute* "src/config.rs" |
| 332 | (("\"/usr/bin/git\"") |
| 333 | (string-append "\"" git "/bin/git\"")))))) |
| 334 | (add-before 'check 'git-init |
| 335 | (lambda _ |
| 336 | ;;The tests expect a git repo with at least one commit. |
| 337 | (invoke "git" "init") |
| 338 | (invoke "git" "config" "--global" "user.email" "[email protected]") |
| 339 | (invoke "git" "config" "--global" "user.name" "User Name") |
| 340 | (invoke "git" "add" ".") |
| 341 | (invoke "git" "commit" "-m" "Commit all.")))))) |
| 342 | (native-inputs (list |
| 343 | git-minimal/pinned ;for tests |
| 344 | pkg-config)) |
| 345 | (inputs |
| 346 | (cons* |
| 347 | libgit2 libssh2 openssl sqlite zlib (cargo-inputs 'git-ai))) |
| 348 | (home-page "https://usegitai.com/") |
| 349 | (synopsis "Git extension for tracking LLM-generated code") |
| 350 | (description "@code{git-ai} is an extension for git that tracks which |
| 351 | lines of code were created by an LLM along with metadata about how and why.") |
| 352 | (license license:asl2.0))) |
| 353 | |
| 354 | (define-public python-apricot-select |
| 355 | (package |
| 356 | (name "python-apricot-select") |
| 357 | ;; 0.6.1 was released in 2021 |
| 358 | (properties '((commit . "962f597a57fcb880a3b19befa7a3eebccc6b5228") |
| 359 | (revision . "0"))) |
| 360 | (version (git-version "0.6.1" |
| 361 | (assoc-ref properties 'revision) |
| 362 | (assoc-ref properties 'commit))) |
| 363 | (source |
| 364 | (origin |
| 365 | (method git-fetch) |
| 366 | (uri (git-reference |
| 367 | (url "https://github.com/jmschrei/apricot") |
| 368 | (commit (assoc-ref properties 'commit)))) |
| 369 | (file-name (git-file-name name version)) |
| 370 | (sha256 |
| 371 | (base32 "16hj76nzdr4pbx7wy5f3237f9c1d0yizmz1skix0rwlvjpj3rc9x")))) |
| 372 | (build-system pyproject-build-system) |
| 373 | (arguments |
| 374 | ;; See: <https://github.com/jmschrei/apricot/issues/19>. |
| 375 | (list #:tests? #f)) ;tests are very compue havy |
| 376 | (native-inputs |
| 377 | (list python-pytest |
| 378 | python-pytest-xdist |
| 379 | python-setuptools |
| 380 | python-scikit-learn)) |
| 381 | (propagated-inputs |
| 382 | (list python-numba |
| 383 | python-numpy |
| 384 | python-scipy |
| 385 | python-tqdm)) |
| 386 | (home-page "https://github.com/jmschrei/apricot") |
| 387 | (synopsis "Submodular selection of representative sets for ML models") |
| 388 | (description |
| 389 | "@code{apricot} implements submodular optimization for the purpose of |
| 390 | summarizing massive data sets into minimally redundant subsets that are still |
| 391 | representative of the original data. These subsets are useful for both |
| 392 | visualizing the modalities in the data and for training accurate machine |
| 393 | learning models with just a fraction of the examples and compute.") |
| 394 | (license license:expat))) |
| 395 | |
| 396 | (define-public python-autograd-gamma |
| 397 | (package |
| 398 | (name "python-autograd-gamma") |
| 399 | (version "0.5.0") |
| 400 | (source |
| 401 | (origin |
| 402 | (method git-fetch) |
| 403 | (uri (git-reference |
| 404 | (url "https://github.com/CamDavidsonPilon/autograd-gamma") |
| 405 | ;; The 5.0.0 version is not tagged, but can only be this commit. |
| 406 | (commit "5876e0320865c592ed4a370c6a6ef282cf163480"))) |
| 407 | (file-name (git-file-name name version)) |
| 408 | (sha256 |
| 409 | (base32 "0hpjnpw7bsy1xja4pnxx5fvv7xly72lp3hdazhr9zbmn6s7vcs19")))) |
| 410 | (build-system pyproject-build-system) |
| 411 | (native-inputs |
| 412 | (list python-pytest |
| 413 | python-setuptools)) |
| 414 | (propagated-inputs |
| 415 | (list python-autograd |
| 416 | python-scipy)) |
| 417 | (home-page "https://github.com/CamDavidsonPilon/autograd-gamma") |
| 418 | (synopsis |
| 419 | "Autograd-compatible approximations to the gamma family of functions") |
| 420 | (description |
| 421 | "This package provides Autograd-compatible approximations to the gamma |
| 422 | family of functions.") |
| 423 | (license license:expat))) |
| 424 | |
| 425 | (define-public python-burr |
| 426 | (package |
| 427 | (name "python-burr") |
| 428 | (version "0.42.0") |
| 429 | (source |
| 430 | (origin |
| 431 | (method git-fetch) |
| 432 | (uri (git-reference |
| 433 | (url "https://github.com/apache/burr/") |
| 434 | (commit (string-append "burr-" version)))) |
| 435 | (file-name (git-file-name name version)) |
| 436 | (sha256 |
| 437 | (base32 "0a2sf72qy9jkamg6r7d37mv7ayp698h031md7arfyq7v8q9q121n")))) |
| 438 | (build-system pyproject-build-system) |
| 439 | (arguments |
| 440 | (list |
| 441 | #:phases |
| 442 | '(modify-phases %standard-phases |
| 443 | ;; Avoid "async def functions are not natively supported." with 70+ |
| 444 | ;; test failures. This hack triggers a bunch of warnings, because |
| 445 | ;; now also synchronous tests are marked with @pytest.mark.asyncio, |
| 446 | ;; but at least the tests run successfully. |
| 447 | (add-after 'unpack 'patch-tests |
| 448 | (lambda _ |
| 449 | (substitute* "tests/common/test_async_utils.py" |
| 450 | (("from typing.*" m) |
| 451 | (string-append m "\ |
| 452 | import asyncio |
| 453 | import pytest |
| 454 | pytestmark = pytest.mark.asyncio\n"))) |
| 455 | (substitute* '("tests/visibility/test_tracing.py" |
| 456 | "tests/core/test_parallelism.py" |
| 457 | "tests/core/test_action.py" |
| 458 | "tests/core/test_application.py" |
| 459 | "tests/test_end_to_end.py") |
| 460 | (("^import asyncio.*" m) |
| 461 | (string-append m "\ |
| 462 | import pytest |
| 463 | pytestmark = pytest.mark.asyncio\n")))))) |
| 464 | ;; There are integration tests and examples comparing Burr to and |
| 465 | ;; interfacing with alternatives, such as the overly complex langchain. |
| 466 | ;; We should not add them here just to run these tests. |
| 467 | #:test-flags |
| 468 | #~(list "--ignore-glob=tests/integration**" |
| 469 | "--ignore-glob=examples/**" |
| 470 | "--ignore-glob=burr/examples/**" |
| 471 | "--ignore=tests/core/test_graphviz_display.py" |
| 472 | "--ignore=tests/core/test_persistence.py"))) |
| 473 | (native-inputs |
| 474 | (list python-flit-core |
| 475 | python-anyio |
| 476 | python-pytest |
| 477 | python-pytest-asyncio |
| 478 | python-setuptools)) |
| 479 | ;; Burr has very few hard dependencies. Including these optional packages |
| 480 | ;; makes it more useful, though. |
| 481 | (propagated-inputs |
| 482 | (list python-pydantic ;optional |
| 483 | python-pydantic-settings ;optional |
| 484 | python-loguru |
| 485 | python-requests)) |
| 486 | (home-page "https://burr.apache.org/") |
| 487 | (synopsis "Build applications that make decisions") |
| 488 | (description |
| 489 | "Apache Burr makes it easy to develop applications that make |
| 490 | decisions (chatbots, agents, simulations, etc...) from simple Python building |
| 491 | blocks. Apache Burr works well for any application that uses LLMs, and can |
| 492 | integrate with any of your favorite frameworks. Burr includes a UI that can |
| 493 | track/monitor/trace your system in real time, along with pluggable |
| 494 | persisters (e.g. for memory) to save and load application state.") |
| 495 | (license license:asl2.0))) |
| 496 | |
| 497 | (define-public python-faster-whisper |
| 498 | (package |
| 499 | (name "python-faster-whisper") |
| 500 | (version "1.2.0") |
| 501 | (source |
| 502 | (origin |
| 503 | (method git-fetch) |
| 504 | (uri (git-reference |
| 505 | (url "https://github.com/SYSTRAN/faster-whisper") |
| 506 | (commit (string-append "v" version)))) |
| 507 | (file-name (git-file-name name version)) |
| 508 | (sha256 |
| 509 | (base32 "0wmc9aszpnbr60l5v97np0rc3zsz4pldjald7kmnr668rhpakzcj")))) |
| 510 | (build-system pyproject-build-system) |
| 511 | ;; XXX: Currently tests requires model download, which we'd rather avoid |
| 512 | ;; in Guix unless we're sure about the FOSS weights. To test in source : |
| 513 | ;; guix shell -D python-faster-whisper -- pytest |
| 514 | (arguments (list #:tests? #f)) |
| 515 | (propagated-inputs (list (list onnxruntime "python") |
| 516 | python-av |
| 517 | python-ctranslate2 |
| 518 | python-huggingface-hub |
| 519 | python-tokenizers |
| 520 | python-tqdm)) |
| 521 | (native-inputs (list python-numpy |
| 522 | python-pytest |
| 523 | python-setuptools)) |
| 524 | (home-page "https://github.com/SYSTRAN/faster-whisper") |
| 525 | (synopsis "Whisper transcription reimplementation") |
| 526 | (description |
| 527 | "This package provides a reimplementation of OpenAI's Whisper model using |
| 528 | CTranslate2, which is a inference engine for transformer models.") |
| 529 | (license license:expat))) |
| 530 | |
| 531 | (define-public python-fasttext |
| 532 | (package |
| 533 | (inherit fasttext) |
| 534 | (name "python-fasttext") |
| 535 | (build-system pyproject-build-system) |
| 536 | (propagated-inputs (list python-numpy python-scipy)) |
| 537 | (inputs (list fasttext)) |
| 538 | (native-inputs (list pybind11-2 python-setuptools python-wheel)))) |
| 539 | |
| 540 | (define-public python-funsor |
| 541 | (package |
| 542 | (name "python-funsor") |
| 543 | (version "0.4.7") |
| 544 | (source |
| 545 | (origin |
| 546 | (method git-fetch) |
| 547 | (uri (git-reference |
| 548 | (url "https://github.com/pyro-ppl/funsor") |
| 549 | (commit version))) |
| 550 | (file-name (git-file-name name version)) |
| 551 | (sha256 |
| 552 | (base32 "1ragxy19qfaczn3wfpamg0giqyiscxvm366r3md255lfafzwj96i")))) |
| 553 | (build-system pyproject-build-system) |
| 554 | (arguments |
| 555 | (list |
| 556 | ;; tests: 7665 passed, 3949 skipped, 71 xfailed, 2 xpassed, 8 warnings |
| 557 | #:test-flags |
| 558 | #~(list "--numprocesses" (number->string (min 8 (parallel-job-count))) |
| 559 | ;; XXX: The project has no updates since 2023 and probably not |
| 560 | ;; compatible with current Python science stack (NumPy, SciPy, |
| 561 | ;; Pandas...). |
| 562 | ;; |
| 563 | ;; See open issues: |
| 564 | ;; <https://github.com/pyro-ppl/funsor/issues/609> |
| 565 | ;; <https://github.com/pyro-ppl/funsor/pull/610> |
| 566 | ;; |
| 567 | ;; TypeError: argument of type 'property' is not iterable |
| 568 | "--ignore=test/examples/test_sensor_fusion.py" |
| 569 | "--ignore=test/torch/test_provenance.py" |
| 570 | ;; The most of the tests fail with error: ValueError: Cannot |
| 571 | ;; convert to Funsor: tensor.. |
| 572 | #$@(map (lambda (test) (string-append "--deselect=test/" |
| 573 | test)) |
| 574 | (list "pyro/test_convert.py::test_dist_to_funsor_bernoulli" |
| 575 | "pyro/test_convert.py::test_dist_to_funsor_categorical" |
| 576 | "pyro/test_convert.py::test_dist_to_funsor_independent" |
| 577 | "pyro/test_convert.py::test_dist_to_funsor_masked" |
| 578 | "pyro/test_convert.py::test_dist_to_funsor_mvn" |
| 579 | "pyro/test_convert.py::test_dist_to_funsor_normal" |
| 580 | "pyro/test_convert.py::test_funsor_to_cat_and_mvn" |
| 581 | "pyro/test_convert.py::test_funsor_to_mvn" |
| 582 | "pyro/test_convert.py::test_matrix_and_mvn_to_funsor" |
| 583 | "pyro/test_convert.py::test_matrix_and_mvn_to_funsor_diag" |
| 584 | "pyro/test_convert.py::test_mvn_to_funsor" |
| 585 | "pyro/test_convert.py::test_tensor_funsor_tensor" |
| 586 | "pyro/test_distribution.py::test_categorical_log_prob" |
| 587 | "pyro/test_distribution.py::test_categorical_sample" |
| 588 | "pyro/test_hmm.py::test_discrete_categorical_log_prob" |
| 589 | "pyro/test_hmm.py::test_discrete_diag_normal_log_prob" |
| 590 | "pyro/test_hmm.py::test_discrete_mvn_log_prob" |
| 591 | "pyro/test_hmm.py::test_discrete_normal_log_prob" |
| 592 | "pyro/test_hmm.py::test_gaussian_hmm_log_prob" |
| 593 | "pyro/test_hmm.py::test_gaussian_hmm_log_prob_null_dynamics" |
| 594 | "pyro/test_hmm.py::test_gaussian_mrf_log_prob" |
| 595 | "pyro/test_hmm.py::test_switching_linear_hmm_log_prob" |
| 596 | "pyro/test_hmm.py::test_switching_linear_hmm_log_prob_alternating" |
| 597 | "pyro/test_hmm.py::test_switching_linear_hmm_shape" |
| 598 | "pyroapi/test_pyroapi.py::test_generate_data" |
| 599 | "pyroapi/test_pyroapi.py::test_generate_data_plate" |
| 600 | "pyroapi/test_pyroapi.py::test_local_param_ok" |
| 601 | "pyroapi/test_pyroapi.py::test_mean_field_ok" |
| 602 | "pyroapi/test_pyroapi.py::test_nested_plate_plate_ok" |
| 603 | "pyroapi/test_pyroapi.py::test_nonempty_model_empty_guide_ok" |
| 604 | "pyroapi/test_pyroapi.py::test_optimizer" |
| 605 | "pyroapi/test_pyroapi.py::test_plate_ok" |
| 606 | "test_tensor.py::test_function_of_numeric_array"))))) |
| 607 | (native-inputs |
| 608 | (list python-pandas |
| 609 | python-pillow |
| 610 | python-pyro-api |
| 611 | python-pytest |
| 612 | python-pytest-xdist |
| 613 | python-requests |
| 614 | python-scipy |
| 615 | python-setuptools |
| 616 | python-torchvision)) |
| 617 | (propagated-inputs |
| 618 | (list python-makefun |
| 619 | python-multipledispatch |
| 620 | python-numpy |
| 621 | python-opt-einsum |
| 622 | python-typing-extensions |
| 623 | ;; [optional] |
| 624 | ;; python-jax |
| 625 | ;; python-jaxlib |
| 626 | ;; python-numpyro |
| 627 | python-pyro-ppl |
| 628 | python-pytorch)) |
| 629 | (home-page "https://github.com/pyro-ppl/funsor") |
| 630 | (synopsis "Tensor-like library for functions and distributions") |
| 631 | (description |
| 632 | "This package provides a tensor-like library for functions and |
| 633 | distributions.") |
| 634 | (license license:asl2.0))) |
| 635 | |
| 636 | (define-public fann |
| 637 | ;; The last release is 14 years old. |
| 638 | (let ((commit "1783cbf6239a597c4d29f694e227e22b8d4f4bf6") |
| 639 | (revision "2")) |
| 640 | (package |
| 641 | (name "fann") |
| 642 | (version (git-version "2.2.0" revision commit)) |
| 643 | (source |
| 644 | (origin |
| 645 | (method git-fetch) |
| 646 | (uri (git-reference |
| 647 | (url "https://github.com/libfann/fann") |
| 648 | (commit commit))) |
| 649 | (modules '((guix build utils))) |
| 650 | (snippet #~(delete-file-recursively "lib/googletest")) |
| 651 | (file-name (git-file-name name version)) |
| 652 | (sha256 |
| 653 | (base32 |
| 654 | "0jlcxl0czlr3982mak3935mb08i2f368f0jsxca91ppgfd596ldr")))) |
| 655 | (build-system cmake-build-system) |
| 656 | (arguments |
| 657 | (list |
| 658 | #:phases |
| 659 | #~(modify-phases %standard-phases |
| 660 | (add-after 'unpack 'remove-googletest |
| 661 | (lambda _ |
| 662 | (substitute* "CMakeLists.txt" |
| 663 | (("ADD_SUBDIRECTORY\\( lib/googletest \\)") |
| 664 | "")))) |
| 665 | (replace 'check |
| 666 | (lambda* (#:key tests? #:allow-other-keys) |
| 667 | (when tests? |
| 668 | (with-directory-excursion "tests" |
| 669 | (invoke "./fann_tests")))))))) |
| 670 | (native-inputs (list googletest-1.13)) |
| 671 | (home-page "https://leenissen.dk/") |
| 672 | (synopsis "Fast Artificial Neural Network") |
| 673 | (description |
| 674 | "FANN is a neural network library, which implements multilayer |
| 675 | artificial neural networks in C with support for both fully connected and |
| 676 | sparsely connected networks.") |
| 677 | (license license:lgpl2.1)))) |
| 678 | |
| 679 | (define-public libsvm |
| 680 | (package |
| 681 | (name "libsvm") |
| 682 | (version "337") |
| 683 | (source |
| 684 | (origin |
| 685 | (method git-fetch) |
| 686 | (uri (git-reference |
| 687 | (url "https://github.com/cjlin1/libsvm/") |
| 688 | (commit (string-append "v" version)))) |
| 689 | (file-name (git-file-name name version)) |
| 690 | (sha256 |
| 691 | (base32 "0rd4balgr62jisml6q682klkcxn8f8mjc5rc69rl2bzvdbrk3xia")))) |
| 692 | (build-system gnu-build-system) |
| 693 | (arguments |
| 694 | (list |
| 695 | #:tests? #f ;no "check" target |
| 696 | #:phases |
| 697 | #~(modify-phases %standard-phases |
| 698 | (delete 'configure) |
| 699 | (add-after 'build 'build-lib |
| 700 | (lambda _ |
| 701 | (invoke "make" "lib"))) |
| 702 | (replace 'install |
| 703 | ;no "install" target |
| 704 | (lambda _ |
| 705 | (for-each (lambda (file) |
| 706 | (install-file file |
| 707 | (string-append #$output "/bin"))) |
| 708 | '("svm-train" "svm-predict" "svm-scale")) |
| 709 | (install-file "libsvm.so.4" |
| 710 | (string-append #$output "/lib")) |
| 711 | (symlink (string-append #$output "/lib/libsvm.so.4") |
| 712 | (string-append #$output "/lib/libsvm.so")) |
| 713 | (install-file "svm.h" |
| 714 | (string-append #$output "/include/" |
| 715 | #$name))))))) |
| 716 | (native-inputs (list python-minimal-wrapper)) |
| 717 | (home-page "https://www.csie.ntu.edu.tw/~cjlin/libsvm/") |
| 718 | (synopsis "Library for Support Vector Machines") |
| 719 | (description |
| 720 | "LIBSVM is a machine learning library for support vector |
| 721 | classification, (C-SVC, nu-SVC), regression (epsilon-SVR, nu-SVR) and |
| 722 | distribution estimation (one-class SVM). It supports multi-class |
| 723 | classification.") |
| 724 | (license license:bsd-3))) |
| 725 | |
| 726 | (define-public python-gpy |
| 727 | (package |
| 728 | (name "python-gpy") |
| 729 | (version "1.14.2") |
| 730 | (source |
| 731 | (origin |
| 732 | (method git-fetch) |
| 733 | (uri (git-reference |
| 734 | (url "https://github.com/SheffieldML/GPy") |
| 735 | (commit (string-append "v" version)))) |
| 736 | (file-name (git-file-name name version)) |
| 737 | (sha256 |
| 738 | (base32 "1l0yxidhsf9s87fcc7v1l38dywq9ps6nr1d84r1nmblp9dgmibr2")))) |
| 739 | (build-system pyproject-build-system) |
| 740 | (arguments |
| 741 | (list |
| 742 | #:phases |
| 743 | #~(modify-phases %standard-phases |
| 744 | (add-after 'unpack 'compatibility |
| 745 | (lambda _ |
| 746 | ;; This file uses Python 2 statements |
| 747 | (delete-file "GPy/testing/mpi_test__.py")))))) |
| 748 | (native-inputs |
| 749 | (list python-cython |
| 750 | python-pods |
| 751 | python-pytest |
| 752 | python-setuptools)) |
| 753 | (propagated-inputs |
| 754 | (list python-numpy |
| 755 | python-paramz |
| 756 | python-scipy |
| 757 | python-six ;hard dependency |
| 758 | ;; [optional] |
| 759 | python-matplotlib |
| 760 | python-plotly)) |
| 761 | (home-page "https://sheffieldml.github.io/GPy/") |
| 762 | (synopsis "The Gaussian Process Toolbox") |
| 763 | (description |
| 764 | "@command{GPy} is a Gaussian Process (GP) framework written in Python, |
| 765 | from the Sheffield machine learning group. GPy implements a range of machine |
| 766 | learning algorithms based on GPs.") |
| 767 | (license license:bsd-3))) |
| 768 | |
| 769 | (define-public python-libsvm |
| 770 | (package |
| 771 | (inherit libsvm) |
| 772 | (name "python-libsvm") |
| 773 | (build-system pyproject-build-system) |
| 774 | (arguments |
| 775 | (list |
| 776 | #:tests? #f ;no "check" target |
| 777 | #:phases |
| 778 | #~(modify-phases %standard-phases |
| 779 | (add-before 'build 'chdir |
| 780 | (lambda _ |
| 781 | (chdir "python")))))) |
| 782 | (native-inputs (list python-setuptools)) |
| 783 | (propagated-inputs (list libsvm python-scipy)) |
| 784 | (synopsis "Python bindings of libSVM"))) |
| 785 | |
| 786 | (define-public python-mcfit |
| 787 | ;; PyPI variant fails during compile-bytecode phase: "IndentationError: |
| 788 | ;; expected an indented block after function definition on line 10 |
| 789 | ;; (mkfit.py, line 15)". |
| 790 | ;; GitHub provides no release tags, use the latest commit from the master |
| 791 | ;; branch. |
| 792 | ;; See: <https://github.com/eelregit/mcfit/issues/5>. |
| 793 | (let ((commit "be3a5cf9c474e16875126adcd35ba785fb781ebb") |
| 794 | (revision "0")) |
| 795 | (package |
| 796 | (name "python-mcfit") |
| 797 | (version (git-version "0.0.22" revision commit)) |
| 798 | (source |
| 799 | (origin |
| 800 | (method git-fetch) |
| 801 | (uri (git-reference |
| 802 | (url "https://github.com/eelregit/mcfit") |
| 803 | (commit commit))) |
| 804 | (sha256 |
| 805 | (base32 "120ybwhrqpn9a9q96m6l8pw8a7cdz705vzqn3s87wnffa8nslbsi")))) |
| 806 | (build-system pyproject-build-system) |
| 807 | (native-inputs |
| 808 | (list python-pytest |
| 809 | python-setuptools |
| 810 | python-wheel)) |
| 811 | (propagated-inputs |
| 812 | (list python-mpmath |
| 813 | python-numpy |
| 814 | python-scipy)) |
| 815 | (home-page "https://github.com/eelregit/mcfit") |
| 816 | (synopsis "Multiplicatively convolutional fast integral transforms") |
| 817 | (description |
| 818 | "This package provides multiplicatively convolutional fast integral |
| 819 | transforms.") |
| 820 | (license license:gpl3+)))) |
| 821 | |
| 822 | (define-public python-minisbd |
| 823 | (package |
| 824 | (name "python-minisbd") |
| 825 | (version "0.9.5") |
| 826 | (source |
| 827 | (origin |
| 828 | (method git-fetch) |
| 829 | (uri (git-reference |
| 830 | (url "https://github.com/LibreTranslate/MiniSBD") |
| 831 | (commit (string-append "v" version)))) |
| 832 | (file-name (git-file-name name version)) |
| 833 | (sha256 |
| 834 | (base32 "1dfflfx39vq1fijrzmfq43hva2ackbx7pj6ixpxbq48p1v9r715a")))) |
| 835 | (build-system pyproject-build-system) |
| 836 | (arguments (list #:tests? #f)) ;no tests |
| 837 | (native-inputs |
| 838 | (list python-hatchling)) |
| 839 | (propagated-inputs |
| 840 | (list python-filelock |
| 841 | python-numpy |
| 842 | (list onnxruntime "python"))) |
| 843 | (home-page "https://github.com/LibreTranslate/MiniSBD") |
| 844 | (synopsis "Sentence boundary detection Python library") |
| 845 | (description |
| 846 | "This package provides a Python library for fast sentence boundary |
| 847 | detection. It uses 8bit quantized ONNX models for inference, thus making it |
| 848 | fast and lightweight.") |
| 849 | (license license:agpl3+))) |
| 850 | |
| 851 | (define-public python-ml-collections |
| 852 | (package |
| 853 | (name "python-ml-collections") |
| 854 | (version "1.1.0") |
| 855 | (source |
| 856 | (origin |
| 857 | (method git-fetch) |
| 858 | (uri (git-reference |
| 859 | (url "https://github.com/google/ml_collections") |
| 860 | (commit (string-append "v" version)))) |
| 861 | (sha256 |
| 862 | (base32 "1lv7vs84v4zwyrqg2zdlkrx3x6w08j9lyz74m8vk55ysl4399pqv")))) |
| 863 | (build-system pyproject-build-system) |
| 864 | (arguments |
| 865 | (list |
| 866 | ;; tests: 373 passed, 1 skipped, 81 warnings |
| 867 | #:test-flags |
| 868 | #~(list "--ignore=docs/" |
| 869 | "--ignore=ml_collections/config_dict/examples/examples_test.py"))) |
| 870 | (native-inputs |
| 871 | (list python-flit-core |
| 872 | python-pytest)) |
| 873 | (propagated-inputs |
| 874 | (list python-absl-py |
| 875 | python-pyyaml)) |
| 876 | (home-page "https://github.com/google/ml_collections") |
| 877 | (synopsis "Python collections designed for Machine Learning usecases") |
| 878 | (description |
| 879 | "ML Collections is a library of Python collections designed for Machine |
| 880 | Learning usecases.") |
| 881 | (license license:asl2.0))) |
| 882 | |
| 883 | (define-public python-ml-dtypes |
| 884 | (package |
| 885 | (name "python-ml-dtypes") |
| 886 | (version "0.5.4") |
| 887 | (source |
| 888 | (origin |
| 889 | (method git-fetch) |
| 890 | (uri (git-reference |
| 891 | (url "https://github.com/jax-ml/ml_dtypes") |
| 892 | (commit (string-append "v" version)))) |
| 893 | (file-name (git-file-name name version)) |
| 894 | (sha256 |
| 895 | (base32 "0q6y0m1jh0a2hj0x887gp361zbrsfz93csr8hm3pkamg354plczc")) |
| 896 | (modules '((guix build utils))) |
| 897 | (snippet |
| 898 | ;; Do not use bundled eigen. |
| 899 | '(delete-file-recursively "third_party/eigen")))) |
| 900 | (build-system pyproject-build-system) |
| 901 | (arguments |
| 902 | (list |
| 903 | ;; tests: 2087 passed, 17 skipped, 2 deselected, 2720 subtests passed |
| 904 | #:test-flags |
| 905 | ;; Arrays are not equal |
| 906 | #~(list (string-append "--deselect=ml_dtypes/tests/" |
| 907 | "custom_float_test.py::CustomFloatNumPyTest" |
| 908 | "::testArange_float8_e4m3b11fnuz") |
| 909 | (string-append "--deselect=ml_dtypes/tests/custom_float_test.py" |
| 910 | "::CustomFloatNumPyTest" |
| 911 | "::testArange_float8_e4m3b11fnuz_multi_threaded")) |
| 912 | #:phases |
| 913 | #~(modify-phases %standard-phases |
| 914 | (add-after 'unpack 'use-eigen-package |
| 915 | (lambda* (#:key inputs #:allow-other-keys) |
| 916 | (substitute* "setup.py" |
| 917 | (("third_party/eigen") |
| 918 | (search-input-directory inputs "include/eigen3"))))) |
| 919 | (add-after 'install 'symlink-lib |
| 920 | (lambda* (#:key inputs outputs #:allow-other-keys) |
| 921 | (for-each |
| 922 | (lambda (file) |
| 923 | (symlink file |
| 924 | (string-append #$output "/lib/ml_dtypes.so"))) |
| 925 | (find-files (site-packages inputs outputs) "\\.so$"))))))) |
| 926 | (inputs (list eigen-for-python-ml-dtypes)) |
| 927 | (propagated-inputs (list python-numpy)) |
| 928 | (native-inputs (list pybind11 |
| 929 | python-absl-py |
| 930 | python-pytest |
| 931 | python-setuptools)) |
| 932 | (home-page "https://github.com/jax-ml/ml_dtypes") |
| 933 | (synopsis "NumPy dtype extensions used in machine learning") |
| 934 | (description |
| 935 | "This package is a stand-alone implementation of several |
| 936 | NumPy @code{dtype} extensions used in machine learning libraries, including: |
| 937 | |
| 938 | @itemize |
| 939 | @item @code{bfloat16}: an alternative to the standard @code{float16} format |
| 940 | @item @code{float8_*}: several experimental 8-bit floating point |
| 941 | representations including: |
| 942 | @itemize |
| 943 | @item @code{float8_e4m3b11fnuz} |
| 944 | @item @code{float8_e4m3fn} |
| 945 | @item @code{float8_e4m3fnuz} |
| 946 | @item @code{float8_e5m2} |
| 947 | @item @code{float8_e5m2fnuz} |
| 948 | @end itemize |
| 949 | @item @code{int4} and @code{uint4}: low precision integer types. |
| 950 | @end itemize |
| 951 | ") |
| 952 | (license license:asl2.0))) |
| 953 | |
| 954 | (define-public llama-cpp |
| 955 | ;; llama.cpp itself now tags real semantic-version releases (unlike the |
| 956 | ;; old "b<N>" raw-build-number scheme this package used to track), but |
| 957 | ;; those releases carry no binary assets of their own -- only a |
| 958 | ;; "nightly-tag.txt" pointing at the "b<N>"-tagged nightly build they |
| 959 | ;; correspond to, which is where the prebuilt web UI tarball (below) |
| 960 | ;; actually lives. |
| 961 | (let ((ui-build "11146")) ;from that tag's nightly-tag.txt |
| 962 | (package |
| 963 | (name "llama-cpp") |
| 964 | (version "0.5.0") |
| 965 | (source |
| 966 | (origin |
| 967 | (method git-fetch) |
| 968 | (uri (git-reference |
| 969 | (url "https://github.com/ggml-org/llama.cpp") |
| 970 | (commit (string-append "v" version)))) |
| 971 | (file-name (git-file-name name version)) |
| 972 | (sha256 |
| 973 | (base32 "0q13scf0ygxrj4zjwcdcpi1j2ncxzqzcbmfcg068vw4c9avrbm1j")))) |
| 974 | (build-system cmake-build-system) |
| 975 | (arguments |
| 976 | (list |
| 977 | ;; ctest includes a fast-growing set of network-dependent tests |
| 978 | ;; (test-download-model and everything that depends on its |
| 979 | ;; FIXTURES_REQUIRED, several test-recurrent-state-rollback-* |
| 980 | ;; variants, test-jinja-py, ...); the previous approach of |
| 981 | ;; patching each one out by name (llama-cpp-disable-tests.patch) |
| 982 | ;; already stopped applying cleanly after upstream added more of |
| 983 | ;; them, so just skip ctest entirely instead of re-patching it |
| 984 | ;; again every version bump. |
| 985 | #:tests? #f |
| 986 | #:configure-flags |
| 987 | #~(list "-DBUILD_SHARED_LIBS=ON" |
| 988 | "-DLLAMA_USE_SYSTEM_GGML=ON") |
| 989 | #:phases |
| 990 | #~(modify-phases %standard-phases |
| 991 | ;; See : |
| 992 | ;; llama.cpp/issues/23105#issuecomment-4531264818 |
| 993 | (add-before 'build 'get-ui |
| 994 | (lambda _ |
| 995 | (with-directory-excursion "tools/ui" |
| 996 | (copy-file #$(this-package-input "ui.tar.gz") "ui.tar.gz") |
| 997 | (invoke "tar" "xvf" "ui.tar.gz") |
| 998 | (rename-file #$(string-append "llama-b" ui-build) "dist")))) |
| 999 | (add-after 'install 'remove-tests |
| 1000 | (lambda _ |
| 1001 | (for-each delete-file |
| 1002 | (find-files (string-append #$output "/bin") |
| 1003 | "^test-"))))))) |
| 1004 | (inputs |
| 1005 | (list ggml openssl |
| 1006 | (origin |
| 1007 | (method url-fetch) |
| 1008 | (uri |
| 1009 | (format #f "~a/releases/download/b~a/llama-b~a-ui.tar.gz" |
| 1010 | "https://github.com/ggml-org/llama.cpp" |
| 1011 | ui-build ui-build)) |
| 1012 | (file-name "ui.tar.gz") |
| 1013 | (sha256 |
| 1014 | (base32 "1wm69rrlhr505q1ww0vs01y9y8lai8pj3lnx8yq80y873lxyfaw3"))))) |
| 1015 | (properties '((tunable? . #true))) ;use AVX512, FMA, etc. when available |
| 1016 | (home-page "https://github.com/ggml-org/llama.cpp") |
| 1017 | (synopsis "Port of Facebook's LLaMA model in C/C++") |
| 1018 | (description "This package provides a port to Facebook's LLaMA collection |
| 1019 | of foundation language models. It requires models parameters to be downloaded |
| 1020 | independently to be able to run a LLaMA model.") |
| 1021 | (license license:expat)))) |
| 1022 | |
| 1023 | (define-public llama-cpp-scripts |
| 1024 | (package |
| 1025 | (inherit llama-cpp) |
| 1026 | (name "llama-cpp-scripts") |
| 1027 | (version "0.0.0") ;from pyproject.toml |
| 1028 | (build-system pyproject-build-system) |
| 1029 | (arguments |
| 1030 | (list |
| 1031 | #:test-flags |
| 1032 | #~(list "--ignore=scripts/snapdragon" |
| 1033 | "--ignore=tools/server/tests"))) |
| 1034 | (inputs '()) |
| 1035 | (propagated-inputs |
| 1036 | (list python-numpy |
| 1037 | python-pytorch |
| 1038 | python-protobuf |
| 1039 | python-sentencepiece |
| 1040 | python-transformers |
| 1041 | python-gguf)) |
| 1042 | (native-inputs (list python-poetry-core python-pytest)) |
| 1043 | (properties '()) |
| 1044 | (synopsis "Scripts that ship with @code{llama.cpp}") |
| 1045 | (description "This package includes the Python scripts included in |
| 1046 | @code{llama.cpp}. These scripts are necessary for format conversion, shader |
| 1047 | generation, etc.") |
| 1048 | (license license:expat))) |
| 1049 | |
| 1050 | (define-public whisper-cpp |
| 1051 | (package |
| 1052 | (name "whisper-cpp") |
| 1053 | (version "1.9.1") ;keep in sync with ggml-for-whisper |
| 1054 | (source (origin |
| 1055 | (method git-fetch) |
| 1056 | (uri (git-reference |
| 1057 | (url "https://github.com/ggml-org/whisper.cpp") |
| 1058 | (commit (string-append "v" version)))) |
| 1059 | (file-name (git-file-name name version)) |
| 1060 | (sha256 |
| 1061 | (base32 |
| 1062 | "1mpmmqm83qy39a1zqphy059x2crzwll2j34ypgwwghy2ji6nc2xm")))) |
| 1063 | (build-system cmake-build-system) |
| 1064 | (arguments |
| 1065 | (list |
| 1066 | #:configure-flags |
| 1067 | #~(list "-DBUILD_SHARED_LIBS=ON" |
| 1068 | "-DWHISPER_SDL2=TRUE" |
| 1069 | "-DWHISPER_COMMON_FFMPEG=TRUE" |
| 1070 | "-DWHISPER_USE_SYSTEM_GGML=ON") |
| 1071 | #:phases |
| 1072 | #~(modify-phases %standard-phases |
| 1073 | #$@(if (target-32bit?) |
| 1074 | '((add-after 'unpack 'skip-failing-tests |
| 1075 | (lambda _ |
| 1076 | ;; 32-bit system |
| 1077 | ;; large model does not fit in RAM in 32-bit system, |
| 1078 | ;; disable large model test |
| 1079 | (substitute* "tests/CMakeLists.txt" |
| 1080 | (("LABELS \"large\"") |
| 1081 | "DISABLED true"))))) |
| 1082 | '()) |
| 1083 | (add-after 'unpack 'fix-parakeet-test |
| 1084 | (lambda _ |
| 1085 | ;; Unlike whisper-cli, test-parakeet does not load the |
| 1086 | ;; dynamically loadable ggml backends, so it aborts with |
| 1087 | ;; "GGML_ASSERT(device) failed" since our ggml is built |
| 1088 | ;; with -DGGML_BACKEND_DL=ON. |
| 1089 | (substitute* "tests/test-parakeet.cpp" |
| 1090 | (("#include \"parakeet\\.h\"") |
| 1091 | "#include \"parakeet.h\"\n#include \"ggml-backend.h\"") |
| 1092 | (("int main\\(\\) \\{") |
| 1093 | "int main() {\n ggml_backend_load_all();")))) |
| 1094 | (add-after 'unpack 'skip-failing-vad-tests |
| 1095 | (lambda _ |
| 1096 | (substitute* "tests/CMakeLists.txt" |
| 1097 | ;; error: failed to read audio data as wav (Unknown error) |
| 1098 | (("\\$\\{VAD_TEST\\} PROPERTIES LABELS \"unit\"") |
| 1099 | "${VAD_TEST} PROPERTIES DISABLED true") |
| 1100 | ;; error: failed to read audio data as wav (Unknown error) |
| 1101 | (("\\$\\{VAD_TEST\\} PROPERTIES LABELS \"base;en\"") |
| 1102 | "${VAD_TEST} PROPERTIES DISABLED true"))))))) |
| 1103 | (native-inputs |
| 1104 | (list git-minimal/pinned pkg-config)) |
| 1105 | (inputs |
| 1106 | (list ffmpeg ggml-for-whisper sdl2)) |
| 1107 | (synopsis "OpenAI's Whisper model in C/C++") |
| 1108 | (description |
| 1109 | "This package is a high-performance inference of OpenAI's |
| 1110 | Whisper automatic speech recognition (ASR) model, implemented in plain C/C++ |
| 1111 | without dependencies, with |
| 1112 | @itemize |
| 1113 | @item AVX intrinsics support for x86 architectures |
| 1114 | @item VSX intrinsics support for POWER architectures |
| 1115 | @item Mixed F16 / F32 precision |
| 1116 | @item 4-bit and 5-bit integer quantization support |
| 1117 | @item Zero memory allocations at runtime |
| 1118 | @item Support for CPU-only inference |
| 1119 | @item Efficient GPU support for NVIDIA |
| 1120 | @item OpenVINO Support |
| 1121 | @item C-style API |
| 1122 | @end itemize") |
| 1123 | (properties '((tunable? . #true))) ;use AVX512, FMA, etc. when available |
| 1124 | (home-page "https://github.com/ggml-org/whisper.cpp/") |
| 1125 | (license license:expat))) |
| 1126 | |
| 1127 | (define-public mcl |
| 1128 | (package |
| 1129 | (name "mcl") |
| 1130 | (version "14.137") |
| 1131 | (source (origin |
| 1132 | (method url-fetch) |
| 1133 | (uri (string-append |
| 1134 | "http://micans.org/mcl/src/mcl-" |
| 1135 | (string-replace-substring version "." "-") |
| 1136 | ".tar.gz")) |
| 1137 | (sha256 |
| 1138 | (base32 |
| 1139 | "15xlax3z31lsn62vlg94hkm75nm40q4679amnfg13jm8m2bnhy5m")))) |
| 1140 | (build-system gnu-build-system) |
| 1141 | (arguments |
| 1142 | (list |
| 1143 | #:configure-flags |
| 1144 | #~(list "--enable-blast" |
| 1145 | (string-append "CFLAGS=-fcommon -g -O2" |
| 1146 | " -Wno-error=implicit-function-declaration")))) |
| 1147 | (inputs |
| 1148 | (list perl)) |
| 1149 | (home-page "https://micans.org/mcl/") |
| 1150 | (synopsis "Clustering algorithm for graphs") |
| 1151 | (description |
| 1152 | "The MCL algorithm is short for the @dfn{Markov Cluster Algorithm}, a |
| 1153 | fast and scalable unsupervised cluster algorithm for graphs (also known as |
| 1154 | networks) based on simulation of (stochastic) flow in graphs.") |
| 1155 | ;; In the LICENCE file and web page it says "The software is licensed |
| 1156 | ;; under the GNU General Public License, version 3.", but in several of |
| 1157 | ;; the source code files it suggests GPL3 or later. |
| 1158 | ;; http://listserver.ebi.ac.uk/pipermail/mcl-users/2016/000376.html |
| 1159 | (license license:gpl3))) |
| 1160 | |
| 1161 | (define-public ocaml-mcl |
| 1162 | (package |
| 1163 | (name "ocaml-mcl") |
| 1164 | (version "12-068oasis4") |
| 1165 | (source |
| 1166 | (origin |
| 1167 | (method git-fetch) |
| 1168 | (uri (git-reference |
| 1169 | (url "https://github.com/fhcrc/mcl") |
| 1170 | (commit version))) |
| 1171 | (file-name (git-file-name name version)) |
| 1172 | (sha256 |
| 1173 | (base32 |
| 1174 | "0009dc3h2jp3qg5val452wngpqnbfyhbcxylghq0mrjqxx0jdq5p")))) |
| 1175 | (build-system ocaml-build-system) |
| 1176 | (arguments |
| 1177 | `(#:phases |
| 1178 | (modify-phases %standard-phases |
| 1179 | (add-before 'configure 'patch-paths |
| 1180 | (lambda _ |
| 1181 | (substitute* "setup.ml" |
| 1182 | (("LDFLAGS=-fPIC") |
| 1183 | (string-append "LDFLAGS=-fPIC\"; \"SHELL=" (which "sh"))) |
| 1184 | (("-std=c89") "-std=gnu99 -fcommon") |
| 1185 | |
| 1186 | ;; This is a mutable string, which is no longer supported. Use |
| 1187 | ;; a byte buffer instead. |
| 1188 | (("String.make \\(String.length s\\)") |
| 1189 | "Bytes.make (String.length s)") |
| 1190 | |
| 1191 | ;; These two belong together. |
| 1192 | (("OASISString.replace_chars") |
| 1193 | "Bytes.to_string (OASISString.replace_chars") |
| 1194 | ((" s;") |
| 1195 | " s);")) |
| 1196 | (substitute* "myocamlbuild.ml" |
| 1197 | (("std=c89") "std=gnu99 -fcommon")) |
| 1198 | ;; Since we build with a more recent OCaml, we have to use C99 or |
| 1199 | ;; later. This causes problems with the old C code. |
| 1200 | (substitute* "src/impala/matrix.c" |
| 1201 | (("restrict") "restrict_")) |
| 1202 | #t))))) |
| 1203 | (native-inputs |
| 1204 | (list ocamlbuild)) |
| 1205 | (home-page "https://github.com/fhcrc/mcl") |
| 1206 | (synopsis "OCaml wrappers around MCL") |
| 1207 | (description |
| 1208 | "This package provides OCaml bindings for the MCL graph clustering |
| 1209 | algorithm.") |
| 1210 | (license license:gpl3))) |
| 1211 | |
| 1212 | (define-public openmm |
| 1213 | (package |
| 1214 | (name "openmm") |
| 1215 | (version "8.5.2") |
| 1216 | (source |
| 1217 | (origin |
| 1218 | (method git-fetch) |
| 1219 | (uri (git-reference |
| 1220 | (url "https://github.com/openmm/openmm") |
| 1221 | (commit version))) |
| 1222 | (file-name (git-file-name name version)) |
| 1223 | (sha256 |
| 1224 | (base32 |
| 1225 | "0y62jg9jl4hrwgcpak6gm7pyvr509msx72a5yp2bqlqir2fa0qzn")))) |
| 1226 | (build-system cmake-build-system) |
| 1227 | (arguments |
| 1228 | (list |
| 1229 | #:configure-flags |
| 1230 | '(list "-DOPENMM_BUILD_SHARED_LIB=TRUE" |
| 1231 | "-DOPENMM_BUILD_REFERENCE_TESTS=FALSE" ;takes too long on ci |
| 1232 | "-DOPENMM_BUILD_C_AND_FORTRAN_WRAPPERS=TRUE" |
| 1233 | "-DOPENMM_BUILD_PYTHON_WRAPPERS=TRUE" |
| 1234 | "-DOPENMM_BUILD_CUDA_LIB=FALSE") |
| 1235 | #:phases |
| 1236 | #~(modify-phases %standard-phases |
| 1237 | (add-after 'unpack 'patch-wrappers-python-install |
| 1238 | (lambda _ |
| 1239 | (substitute* "wrappers/python/CMakeLists.txt" |
| 1240 | (("install --root=\\\\\\$ENV\\{DESTDIR\\}/") |
| 1241 | (string-append "install --prefix=" #$output |
| 1242 | " --root=/ --single-version-externally-managed"))))) |
| 1243 | (add-after 'unpack 'remove-shaky-tests |
| 1244 | (lambda _ |
| 1245 | ;; These tests may unpredictably fail in CI, even passing |
| 1246 | ;; locally. |
| 1247 | (for-each delete-file |
| 1248 | (append '("./serialization/tests/TestSerializeIntegrator.cpp") |
| 1249 | (find-files "./platforms" |
| 1250 | (string-append "Test.*" |
| 1251 | "(BrownianIntegrat|" |
| 1252 | "LangevinIntegrator)" |
| 1253 | "\\.cpp$")))))) |
| 1254 | (add-after 'install 'install-python |
| 1255 | (lambda _ |
| 1256 | (invoke "make" "PythonInstall")))))) |
| 1257 | (inputs |
| 1258 | (list python-wrapper)) |
| 1259 | (propagated-inputs |
| 1260 | (list python-numpy)) |
| 1261 | (native-inputs |
| 1262 | (list doxygen |
| 1263 | gfortran |
| 1264 | opencl-headers |
| 1265 | python-cython |
| 1266 | python-setuptools |
| 1267 | swig-4.4)) |
| 1268 | (home-page "https://github.com/openmm/openmm/") |
| 1269 | (synopsis "Toolkit for molecular simulation") |
| 1270 | (description |
| 1271 | "OpenMM is a toolkit for molecular simulation. It can be used either as |
| 1272 | a stand-alone application for running simulations, or as a library you call |
| 1273 | from your own code.") |
| 1274 | ;; All licenses are listed in <docs-source/licenses/Licenses.txt>. |
| 1275 | (license (list license:cc-by3.0 ;PdbxReader |
| 1276 | license:expat ;API, Reference Platform, CPU Platform, VkFFT |
| 1277 | license:gpl3 ;Hilbert Curve |
| 1278 | license:lgpl3)))) ;CUDA and OpenCL Platforms |
| 1279 | |
| 1280 | (define-public r-rcppml/devel |
| 1281 | (let ((commit "2beac6580174f8f3ef4ac6fdb2ca2b65705d265a") |
| 1282 | (revision "3")) |
| 1283 | (package |
| 1284 | (name "r-rcppml-devel") |
| 1285 | (version (git-version "0.5.6" revision commit)) |
| 1286 | (source (origin |
| 1287 | (method git-fetch) |
| 1288 | (uri (git-reference |
| 1289 | (url "https://github.com/zdebruine/RcppML") |
| 1290 | (commit commit))) |
| 1291 | (file-name (git-file-name name version)) |
| 1292 | (sha256 |
| 1293 | (base32 |
| 1294 | "0nv4w2a31bhccblnp5jqylpf9dbbwv07mpkxma40gp8r12k6sv0r")))) |
| 1295 | (properties `((upstream-name . "RcppML"))) |
| 1296 | (build-system r-build-system) |
| 1297 | ;; Reported here: https://github.com/zdebruine/RcppML/issues/71 |
| 1298 | (arguments (list #:tests? #false)) |
| 1299 | (propagated-inputs (list r-matrix r-rcpp)) |
| 1300 | (native-inputs (list r-knitr r-testthat)) |
| 1301 | (home-page "https://github.com/zdebruine/RcppML") |
| 1302 | (synopsis "Rcpp machine learning Library") |
| 1303 | (description |
| 1304 | "This package provides fast machine learning algorithms including |
| 1305 | matrix factorization and divisive clustering for large sparse and dense |
| 1306 | matrices.") |
| 1307 | (license license:gpl3+)))) |
| 1308 | |
| 1309 | (define-public openfst |
| 1310 | (package |
| 1311 | (name "openfst") |
| 1312 | (version "1.8.4") |
| 1313 | (source (origin |
| 1314 | (method url-fetch) |
| 1315 | (uri (string-append "https://www.openfst.org/twiki/pub/FST/" |
| 1316 | "FstDownload/openfst-" version ".tar.gz")) |
| 1317 | (sha256 |
| 1318 | (base32 |
| 1319 | "05l057mx1cmbm2jm99mrg75qgz4ca5r78n002mkpxl4j7mpvpsx8")))) |
| 1320 | (build-system gnu-build-system) |
| 1321 | (arguments '(#:configure-flags '("--enable-ngram-fsts"))) |
| 1322 | (home-page "https://www.openfst.org") |
| 1323 | (synopsis "Library for weighted finite-state transducers") |
| 1324 | (description "OpenFst is a library for constructing, combining, |
| 1325 | optimizing, and searching weighted finite-state transducers (FSTs).") |
| 1326 | (license license:asl2.0))) |
| 1327 | |
| 1328 | (define openfst-for-vosk |
| 1329 | (package |
| 1330 | (inherit openfst) |
| 1331 | (version "1.8.0") |
| 1332 | (source |
| 1333 | (origin |
| 1334 | (method url-fetch) |
| 1335 | (uri (string-append "http://www.openfst.org/twiki/pub/FST/" |
| 1336 | "FstDownload/openfst-" version ".tar.gz")) |
| 1337 | (sha256 |
| 1338 | (base32 "0h2lfhhihg63b804hrcljnkggijbjmp84i5g8q735wb09y9z2c4p")) |
| 1339 | (patches |
| 1340 | (search-patches "openfst-for-vosk-fix-unique-ptr.patch")))) |
| 1341 | (arguments |
| 1342 | '(#:configure-flags |
| 1343 | '("--enable-shared" "--enable-far" "--enable-ngram-fsts" |
| 1344 | "--enable-lookahead-fsts" "--with-pic" "--disable-bin"))))) |
| 1345 | |
| 1346 | (define-public sentencepiece |
| 1347 | (package |
| 1348 | (name "sentencepiece") |
| 1349 | (version "0.2.1") |
| 1350 | (source |
| 1351 | (origin |
| 1352 | (method git-fetch) |
| 1353 | (uri |
| 1354 | (git-reference |
| 1355 | (url "https://github.com/google/sentencepiece/") |
| 1356 | (commit (string-append "v" version)))) |
| 1357 | (file-name (git-file-name name version)) |
| 1358 | (sha256 |
| 1359 | (base32 "1yzsaam3kk9fqzxfc3nm2pskwq1bdgkrijmcmzmg5x0338rn0hmb")))) |
| 1360 | (build-system cmake-build-system) |
| 1361 | (arguments (list #:tests? #f)) ;no tests |
| 1362 | (native-inputs (list gperftools)) |
| 1363 | (home-page "https://github.com/google/sentencepiece") |
| 1364 | (synopsis "Unsupervised tokenizer for Neural Network-based text generation") |
| 1365 | (description |
| 1366 | "SentencePiece is an unsupervised text tokenizer and detokenizer mainly |
| 1367 | for Neural Network-based text generation systems where the vocabulary size is |
| 1368 | predetermined prior to the neural model training. SentencePiece implements |
| 1369 | subword units---e.g., byte-pair-encoding (BPE) and unigram language |
| 1370 | model---with the extension of direct training from raw sentences. |
| 1371 | SentencePiece allows us to make a purely end-to-end system that does not |
| 1372 | depend on language-specific pre- or post-processing.") |
| 1373 | (license license:asl2.0))) |
| 1374 | |
| 1375 | (define-public stable-diffusion-cpp |
| 1376 | (package |
| 1377 | (name "stable-diffusion-cpp") |
| 1378 | (version "827-97d2990") |
| 1379 | (source |
| 1380 | (origin |
| 1381 | (method git-fetch) |
| 1382 | (uri (git-reference |
| 1383 | (url "https://github.com/leejet/stable-diffusion.cpp") |
| 1384 | (commit (string-append "master-" version)) |
| 1385 | ;; For libwebm 1.0.0, libwepb 1.6.0, and custom fork of ggml. |
| 1386 | (recursive? #t))) |
| 1387 | (file-name (git-file-name name version)) |
| 1388 | (sha256 |
| 1389 | (base32 "1p9b8jvqsxd0dn8ccn0zhxs9nwqx3l20qbp9ybg2dirpja4i0xfm")))) |
| 1390 | (build-system cmake-build-system) |
| 1391 | (arguments |
| 1392 | (list |
| 1393 | #:tests? #f ;no tests |
| 1394 | ;; see docs/build.md |
| 1395 | #:configure-flags |
| 1396 | #~(list "-DSD_BUILD_SHARED_LIBS=ON") |
| 1397 | #:phases |
| 1398 | #~(modify-phases %standard-phases |
| 1399 | (add-before 'configure 'set-CXXFLAGS |
| 1400 | (lambda _ |
| 1401 | (setenv "CXXFLAGS" "-DGGML_MAX_NAME=128")))))) |
| 1402 | (properties '((tunable? . #true))) ;use AVX512, FMA, etc. when available |
| 1403 | (home-page "https://github.com/leejet/stable-diffusion.cpp") |
| 1404 | (synopsis "Diffusion model inference in pure C/C++") |
| 1405 | (description |
| 1406 | "This package provides a Plain C/C++ implementation based |
| 1407 | on ggml, working in the same way as llama.cpp. It supports diffusion models |
| 1408 | such as SD, Flux, Wan, Qwen and Image,Z-Image.") |
| 1409 | (license license:expat))) |
| 1410 | |
| 1411 | (define-public python-hopcroftkarp |
| 1412 | ;; This commit fixes a broken import, but has not been released to PyPI. |
| 1413 | (let ((commit "2846e1dd3265d95d2bddb0cf4190b830cbb4efe6") |
| 1414 | (revision "1")) |
| 1415 | (package |
| 1416 | (name "python-hopcroftkarp") |
| 1417 | (version (git-version "1.2.5" revision commit)) |
| 1418 | (source |
| 1419 | (origin |
| 1420 | (method git-fetch) |
| 1421 | (uri (git-reference |
| 1422 | (url "https://github.com/sofiatolaosebikan/hopcroftkarp") |
| 1423 | (commit commit))) |
| 1424 | (file-name (git-file-name name version)) |
| 1425 | (sha256 |
| 1426 | (base32 "018ilrp41fcclmb5lsml3aijwbmhbq3m7wy65hr1fryj0avic8fr")))) |
| 1427 | (build-system pyproject-build-system) |
| 1428 | (arguments (list #:test-backend #~'unittest)) |
| 1429 | (native-inputs (list python-setuptools python-wheel)) |
| 1430 | (home-page "https://github.com/sofiatolaosebikan/hopcroftkarp") |
| 1431 | (synopsis "Implementation of the Hopcroft-Karp algorithm") |
| 1432 | (description |
| 1433 | "This package implements the Hopcroft-Karp algorithm, producing a maximum |
| 1434 | cardinality matching from a bipartite graph.") |
| 1435 | (license license:gpl3)))) |
| 1436 | |
| 1437 | (define-public python-paramz |
| 1438 | (package |
| 1439 | (name "python-paramz") |
| 1440 | (version "0.10.0") |
| 1441 | (source |
| 1442 | (origin |
| 1443 | (method git-fetch) |
| 1444 | (uri (git-reference |
| 1445 | (url "https://github.com/sods/paramz") |
| 1446 | (commit (string-append "v" version)))) |
| 1447 | (file-name (git-file-name name version)) |
| 1448 | (sha256 |
| 1449 | (base32 "0h7m7fzb88cx4f8qq2p34z3jv3f48b46pc11is6592zlmrdp67lf")))) |
| 1450 | (build-system pyproject-build-system) |
| 1451 | (arguments |
| 1452 | (list |
| 1453 | ;; tests: 114 passed, 4 skipped, 14 warnings |
| 1454 | #:test-flags |
| 1455 | ;; Compute instance test which times out. |
| 1456 | #~(list (string-append "--deselect=paramz/tests/model_tests.py" |
| 1457 | "::TestModel::test_optimize_restarts_parallel")))) |
| 1458 | (native-inputs |
| 1459 | (list python-pytest |
| 1460 | python-setuptools)) |
| 1461 | (propagated-inputs |
| 1462 | (list python-decorator |
| 1463 | python-numpy |
| 1464 | python-scipy |
| 1465 | python-six)) ;hard dependency |
| 1466 | (home-page "https://github.com/sods/paramz") |
| 1467 | (synopsis "The Parameterization Framework") |
| 1468 | (description |
| 1469 | "@command{paramz} is a lightweight parameterization framework |
| 1470 | for parameterized model creation and handling. Its features include: |
| 1471 | |
| 1472 | @itemize |
| 1473 | @item Easy model creation with parameters. |
| 1474 | @item Fast optimized access of parameters for optimization routines. |
| 1475 | @item Memory efficient storage of parameters (only one copy in memory). |
| 1476 | @item Renaming of parameters. |
| 1477 | @item Intuitive printing of models and parameters. |
| 1478 | @item Gradient saving directly inside parameters. |
| 1479 | @item Gradient checking of parameters. |
| 1480 | @item Optimization of parameters. |
| 1481 | @item Jupyter notebook integration. |
| 1482 | @item Efficient storage of models, for reloading. |
| 1483 | @item Efficient caching. |
| 1484 | @end itemize") |
| 1485 | (license license:bsd-3))) |
| 1486 | |
| 1487 | (define-public python-persim |
| 1488 | (package |
| 1489 | (name "python-persim") |
| 1490 | (version "0.3.8") |
| 1491 | (source |
| 1492 | (origin |
| 1493 | (method url-fetch) |
| 1494 | (uri (pypi-uri "persim" version)) |
| 1495 | (sha256 |
| 1496 | (base32 "03d4kgakpgj54c3pl9dkqrkbmj6w13gmczkds5jagf3n85c1hgg1")))) |
| 1497 | (build-system pyproject-build-system) |
| 1498 | (native-inputs |
| 1499 | (list python-pytest python-setuptools)) |
| 1500 | (propagated-inputs |
| 1501 | (list python-deprecated |
| 1502 | python-hopcroftkarp |
| 1503 | python-joblib |
| 1504 | python-matplotlib |
| 1505 | python-numpy |
| 1506 | python-scikit-learn |
| 1507 | python-scipy)) |
| 1508 | (home-page "https://persim.scikit-tda.org") |
| 1509 | (synopsis "Tools for analyzing persistence diagrams in Python") |
| 1510 | (description |
| 1511 | "This package includes a variety of tools used to analyze persistence diagrams. |
| 1512 | It currently houses implementations of |
| 1513 | @itemize |
| 1514 | @item Persistence images |
| 1515 | @item Persistence landscapes |
| 1516 | @item Bottleneck distance |
| 1517 | @item Modified Gromov–Hausdorff distance |
| 1518 | @item Sliced Wasserstein kernel |
| 1519 | @item Heat kernel |
| 1520 | @item Diagram plotting |
| 1521 | @end itemize |
| 1522 | ") |
| 1523 | (license license:expat))) ; MIT License |
| 1524 | |
| 1525 | (define-public python-pomegranate |
| 1526 | (package |
| 1527 | (name "python-pomegranate") |
| 1528 | (version "1.1.0") |
| 1529 | (source |
| 1530 | (origin |
| 1531 | (method git-fetch) |
| 1532 | (uri (git-reference |
| 1533 | (url "https://github.com/jmschrei/pomegranate") |
| 1534 | (commit (string-append "v" version)))) |
| 1535 | (file-name (git-file-name name version)) |
| 1536 | (sha256 |
| 1537 | (base32 "0arx5wik4ywna7k7g5y3ggwz98fl1hgap0278pph3c77ap8afqd7")))) |
| 1538 | (build-system pyproject-build-system) |
| 1539 | (arguments |
| 1540 | (list |
| 1541 | ;; tests: 830 passed, 19 deselected, 529 warnings |
| 1542 | #:test-flags |
| 1543 | #~(list #$@(map (lambda (test) (string-append "--deselect=" |
| 1544 | "tests/distributions/" |
| 1545 | test)) |
| 1546 | ;; _pickle.UnpicklingError: Weights only load failed. |
| 1547 | (list "test_bernoulli.py::test_serialization" |
| 1548 | "test_categorical.py::test_serialization" |
| 1549 | "test_exponential.py::test_serialization" |
| 1550 | "test_gamma.py::test_serialization" |
| 1551 | "test_independent_component.py::test_serialization" |
| 1552 | "test_normal_diagonal.py::test_serialization" |
| 1553 | "test_normal_full.py::test_serialization" |
| 1554 | "test_poisson.py::test_serialization" |
| 1555 | "test_student_t.py::test_serialization" |
| 1556 | "test_uniform.py::test_serialization" |
| 1557 | ;; Arrays are not almost equal to 3 decimals |
| 1558 | "test_categorical.py::test_sample" |
| 1559 | "test_independent_component.py::test_sample" |
| 1560 | "test_joint_categorical.py::test_sample" |
| 1561 | "test_joint_categorical.py::test_sample")) |
| 1562 | #$@(map (lambda (test) (string-append "--deselect=" |
| 1563 | "tests/" |
| 1564 | test)) |
| 1565 | ;; _pickle.UnpicklingError: Weights only load failed. |
| 1566 | (list "test_bayes_classifier.py::test_serialization" |
| 1567 | "test_gmm.py::test_serialization" |
| 1568 | "test_kmeans.py::test_serialization" |
| 1569 | ;; Arrays are not almost equal to 3 decimals |
| 1570 | "test_bayesian_network.py::test_sample" |
| 1571 | "test_gmm.py::test_sample" |
| 1572 | "test_markov_chain.py::test_sample" |
| 1573 | "test_markov_chain.py::test_sample"))))) |
| 1574 | (native-inputs |
| 1575 | (list python-cython |
| 1576 | python-pytest |
| 1577 | python-setuptools)) |
| 1578 | (propagated-inputs |
| 1579 | (list python-apricot-select |
| 1580 | python-networkx |
| 1581 | python-numpy |
| 1582 | python-scikit-learn |
| 1583 | python-scipy |
| 1584 | python-pytorch)) |
| 1585 | (home-page "https://pypi.python.org/pypi/pomegranate/") |
| 1586 | (synopsis "Graphical models library for Python") |
| 1587 | (description |
| 1588 | "Pomegranate is a graphical models library for Python, implemented in |
| 1589 | Cython for speed.") |
| 1590 | (license license:expat))) |
| 1591 | |
| 1592 | (define-public python-pot |
| 1593 | (package |
| 1594 | (name "python-pot") |
| 1595 | (version "0.9.7.post1") |
| 1596 | (source |
| 1597 | (origin |
| 1598 | (method git-fetch) |
| 1599 | (uri (git-reference |
| 1600 | (url "https://github.com/PythonOT/POT") |
| 1601 | (commit version))) |
| 1602 | (file-name (git-file-name name version)) |
| 1603 | (sha256 |
| 1604 | (base32 "1pcsvxsmilz3xwa8lm5cgllbjjrx9q92nxm2vmb8jlwhh56wan0g")))) |
| 1605 | (build-system pyproject-build-system) |
| 1606 | ;; tests: 1885 passed, 60 skipped, 6 xfailed, 9288 warnings |
| 1607 | (arguments |
| 1608 | (list |
| 1609 | #:phases |
| 1610 | #~(modify-phases %standard-phases |
| 1611 | (add-after 'unpack 'use-system-eigen |
| 1612 | (lambda* (#:key inputs #:allow-other-keys) |
| 1613 | (substitute* "setup.py" |
| 1614 | (("os\\.path\\.join\\(ROOT, \"deps/eigen\"\\)") |
| 1615 | (format #f "~s" |
| 1616 | (search-input-directory inputs |
| 1617 | "include/eigen3"))))))))) |
| 1618 | (native-inputs |
| 1619 | (list python-cython |
| 1620 | python-pytest |
| 1621 | python-setuptools)) |
| 1622 | (inputs |
| 1623 | (list eigen)) |
| 1624 | (propagated-inputs |
| 1625 | (list python-numpy |
| 1626 | python-scipy |
| 1627 | ;; [optional] |
| 1628 | python-autograd |
| 1629 | python-cvxopt |
| 1630 | ;; python-jax |
| 1631 | ;; python-jaxlib |
| 1632 | python-matplotlib |
| 1633 | python-pymanopt |
| 1634 | python-pytorch |
| 1635 | python-pytorch-geometric |
| 1636 | python-scikit-learn)) |
| 1637 | (home-page "https://github.com/PythonOT/POT") |
| 1638 | (synopsis "Python Optimal Transport Library") |
| 1639 | (description |
| 1640 | "This Python library provides several solvers for optimization problems |
| 1641 | related to Optimal Transport for signal, image processing and machine |
| 1642 | learning.") |
| 1643 | (license license:expat))) |
| 1644 | |
| 1645 | (define-public python-pymanopt |
| 1646 | (package |
| 1647 | (name "python-pymanopt") |
| 1648 | (version "2.2.1") |
| 1649 | (source |
| 1650 | (origin |
| 1651 | (method url-fetch) |
| 1652 | (uri (pypi-uri "pymanopt" version)) |
| 1653 | (sha256 |
| 1654 | (base32 "1nm1yz5hbj1valqq23r8c1g9rhfdndfswlqv6xrnvc3f8fd95167")))) |
| 1655 | (build-system pyproject-build-system) |
| 1656 | (arguments |
| 1657 | (list |
| 1658 | ;; Tests require jax and tensorflow, which are optional. |
| 1659 | #:tests? #false |
| 1660 | #:phases |
| 1661 | '(modify-phases %standard-phases |
| 1662 | ;; This is probably a bad idea. We don't have scipy 1.13 just yet. |
| 1663 | (add-after 'unpack 'relax-requirements |
| 1664 | (lambda _ |
| 1665 | (substitute* "pyproject.toml" |
| 1666 | ((",!=1.12\\.\\*") ""))))))) |
| 1667 | (propagated-inputs (list python-numpy python-scipy)) |
| 1668 | (native-inputs (list python-autograd |
| 1669 | python-matplotlib |
| 1670 | python-pytest |
| 1671 | python-setuptools |
| 1672 | python-setuptools-scm |
| 1673 | python-wheel)) |
| 1674 | (home-page "https://pymanopt.org/") |
| 1675 | (synopsis "Toolbox for optimization on Riemannian manifolds") |
| 1676 | (description |
| 1677 | "This package is a toolbox for optimization on Riemannian manifolds with |
| 1678 | support for automatic differentiation.") |
| 1679 | (license license:bsd-3))) |
| 1680 | |
| 1681 | (define-public python-ripser |
| 1682 | (package |
| 1683 | (name "python-ripser") |
| 1684 | (version "0.6.14") |
| 1685 | (source |
| 1686 | (origin |
| 1687 | (method git-fetch) |
| 1688 | (uri (git-reference |
| 1689 | (url "https://github.com/scikit-tda/ripser.py") |
| 1690 | (commit (string-append "v" version)))) |
| 1691 | (file-name (git-file-name name version)) |
| 1692 | (sha256 |
| 1693 | (base32 |
| 1694 | "00na7dpvpc5vwjwgh9ndqrqlswgrzgqf1s7rxfv1vy5wn63fz3m7")))) |
| 1695 | (build-system pyproject-build-system) |
| 1696 | (propagated-inputs (list python-numpy python-persim python-scikit-learn |
| 1697 | python-scipy)) |
| 1698 | (native-inputs |
| 1699 | (list python-cython |
| 1700 | python-pytest |
| 1701 | python-setuptools)) |
| 1702 | (home-page "https://ripser.scikit-tda.org") |
| 1703 | (synopsis "Persistent homology library for Python") |
| 1704 | (description |
| 1705 | "This package implements a variety of persistent homology algorithms. It |
| 1706 | provides an interface for |
| 1707 | @itemize |
| 1708 | @item computing persistence cohomology of sparse and dense data sets |
| 1709 | @item visualizing persistence diagrams |
| 1710 | @item computing lowerstar filtrations on images |
| 1711 | @item computing representative cochains |
| 1712 | @end itemize |
| 1713 | ") |
| 1714 | (license license:expat))) ; MIT License |
| 1715 | |
| 1716 | (define-public python-sacrebleu |
| 1717 | (package |
| 1718 | (name "python-sacrebleu") |
| 1719 | (version "2.6.0") |
| 1720 | (source (origin |
| 1721 | (method git-fetch) |
| 1722 | (uri (git-reference |
| 1723 | (url "https://github.com/mjpost/sacrebleu") |
| 1724 | (commit (string-append "v" version)))) |
| 1725 | (file-name (git-file-name name version)) |
| 1726 | (sha256 |
| 1727 | (base32 |
| 1728 | "0d1s5f4cv668cjpz7zjxl0mjn7jhl6b0y9mx3cf78frzszglvya7")))) |
| 1729 | (build-system pyproject-build-system) |
| 1730 | (arguments |
| 1731 | (list |
| 1732 | #:test-flags |
| 1733 | ;; These all need internet access. |
| 1734 | '(list "-k" "not test_api_get_source \ |
| 1735 | and not test_api_get_reference \ |
| 1736 | and not test_maybe_download \ |
| 1737 | and not test_process_to_text \ |
| 1738 | and not test_get_files_and_fieldnames \ |
| 1739 | and not test_source_and_references \ |
| 1740 | and not test_wmt22_references") |
| 1741 | #:phases |
| 1742 | '(modify-phases %standard-phases |
| 1743 | ;; Needed for tests. |
| 1744 | (add-before 'check 'set-HOME |
| 1745 | (lambda _ (setenv "HOME" "/tmp")))))) |
| 1746 | (propagated-inputs (list python-colorama |
| 1747 | python-lxml |
| 1748 | python-numpy |
| 1749 | python-portalocker |
| 1750 | python-regex |
| 1751 | python-tabulate)) |
| 1752 | (native-inputs (list python-pytest python-setuptools python-setuptools-scm)) |
| 1753 | (home-page "https://github.com/mjpost/sacrebleu") |
| 1754 | (synopsis |
| 1755 | "Compute shareable, comparable, and reproducible BLEU, chrF, and TER scores") |
| 1756 | (description |
| 1757 | "This is a package for hassle-free computation of shareable, comparable, |
| 1758 | and reproducible BLEU, chrF, and TER scores for natural language processing.") |
| 1759 | (license license:asl2.0))) |
| 1760 | |
| 1761 | (define-public python-safetensors |
| 1762 | (package |
| 1763 | (name "python-safetensors") |
| 1764 | (version "0.4.3") |
| 1765 | (source |
| 1766 | (origin |
| 1767 | (method url-fetch) |
| 1768 | (uri (pypi-uri "safetensors" version)) |
| 1769 | (sha256 |
| 1770 | (base32 "1hhiwy67jarm70l0k26fs1cjhzkgzrh79q14bklj2yp0qi8gr19g")) |
| 1771 | (modules '((guix build utils) |
| 1772 | (ice-9 ftw))) |
| 1773 | (snippet |
| 1774 | #~(begin ;Only keep bindings. |
| 1775 | (for-each |
| 1776 | (lambda (file) |
| 1777 | (unless (member file '("." ".." "bindings" "PKG-INFO")) |
| 1778 | (delete-file-recursively file))) |
| 1779 | (scandir ".")) |
| 1780 | (for-each |
| 1781 | (lambda (file) |
| 1782 | (unless (member file '("." "..")) |
| 1783 | (rename-file (string-append "bindings/python/" file) |
| 1784 | file))) |
| 1785 | (scandir "bindings/python")) |
| 1786 | (substitute* "Cargo.toml" |
| 1787 | (("^path = .*") "")))))) |
| 1788 | (build-system cargo-build-system) |
| 1789 | (arguments |
| 1790 | (list |
| 1791 | #:install-source? #f |
| 1792 | #:phases |
| 1793 | #~(modify-phases %standard-phases |
| 1794 | (add-before 'check 'install-rust-library |
| 1795 | (lambda _ |
| 1796 | (copy-file "target/release/libsafetensors_rust.so" |
| 1797 | "py_src/safetensors/_safetensors_rust.so"))) |
| 1798 | (replace 'check |
| 1799 | (lambda* (#:key tests? #:allow-other-keys) |
| 1800 | (when tests? |
| 1801 | (setenv "PYTHONPATH" (string-append (getcwd) "/py_src")) |
| 1802 | (invoke "python3" |
| 1803 | "-m" "pytest" |
| 1804 | "-n" "auto" |
| 1805 | "--dist=loadfile" |
| 1806 | "-s" "-v" "./tests/" |
| 1807 | ;; Missing jax and tensorflow dependency |
| 1808 | "--ignore=./tests/test_flax_comparison.py" |
| 1809 | "--ignore=./tests/test_tf_comparison.py")))) |
| 1810 | (add-after 'install 'install-python |
| 1811 | (lambda _ |
| 1812 | (let* ((pversion #$(version-major+minor |
| 1813 | (package-version python))) |
| 1814 | (lib (string-append #$output "/lib/python" pversion |
| 1815 | "/site-packages/")) |
| 1816 | (info (string-append lib "safetensors-" |
| 1817 | #$(package-version this-package) |
| 1818 | ".dist-info"))) |
| 1819 | (mkdir-p info) |
| 1820 | (copy-file "PKG-INFO" (string-append info "/METADATA")) |
| 1821 | (copy-recursively |
| 1822 | "py_src/safetensors" |
| 1823 | (string-append lib "safetensors")))))))) |
| 1824 | (inputs |
| 1825 | (cargo-inputs 'python-safetensors)) |
| 1826 | (native-inputs |
| 1827 | (list python-h5py |
| 1828 | python-minimal |
| 1829 | python-numpy |
| 1830 | python-pytest |
| 1831 | python-pytest-xdist |
| 1832 | python-pytorch)) |
| 1833 | (home-page "https://huggingface.co/docs/safetensors") |
| 1834 | (synopsis "Simple and safe way to store and distribute tensors") |
| 1835 | (description |
| 1836 | "This package provides a fast (zero-copy) and safe (dedicated) format for |
| 1837 | storing tensors safely.") |
| 1838 | (license license:asl2.0))) |
| 1839 | |
| 1840 | (define-public python-sentencepiece |
| 1841 | (package |
| 1842 | (inherit sentencepiece) |
| 1843 | (name "python-sentencepiece") |
| 1844 | (build-system pyproject-build-system) |
| 1845 | (arguments |
| 1846 | (list |
| 1847 | #:phases |
| 1848 | #~(modify-phases %standard-phases |
| 1849 | (add-after 'unpack 'chdir |
| 1850 | (lambda _ |
| 1851 | (chdir "python")))))) |
| 1852 | (native-inputs |
| 1853 | (list pkg-config protobuf python-pytest python-setuptools)) |
| 1854 | (propagated-inputs (list sentencepiece)) |
| 1855 | (synopsis "SentencePiece python wrapper") |
| 1856 | (description "This package provides a Python wrapper for the SentencePiece |
| 1857 | unsupervised text tokenizer."))) |
| 1858 | |
| 1859 | (define-public python-sentence-transformers |
| 1860 | (package |
| 1861 | (name "python-sentence-transformers") |
| 1862 | (version "5.5.0") |
| 1863 | (source |
| 1864 | (origin |
| 1865 | (method git-fetch) |
| 1866 | (uri (git-reference |
| 1867 | (url "https://github.com/UKPLab/sentence-transformers/") |
| 1868 | (commit (string-append "v" version)))) |
| 1869 | (file-name (git-file-name name version)) |
| 1870 | (sha256 |
| 1871 | (base32 |
| 1872 | "0x6b6gvsasdmm40gr5ld5gwxggpb2j285jv4scvws1bp2s193k9i")))) |
| 1873 | (build-system pyproject-build-system) |
| 1874 | (arguments |
| 1875 | (list #:tests? #f)) ;network access is required |
| 1876 | (propagated-inputs (list python-huggingface-hub |
| 1877 | python-numpy |
| 1878 | python-pillow |
| 1879 | python-pytorch |
| 1880 | python-typing-extensions |
| 1881 | python-scikit-learn |
| 1882 | python-scipy |
| 1883 | python-tqdm |
| 1884 | python-transformers)) |
| 1885 | (native-inputs (list python-setuptools)) |
| 1886 | (home-page "https://www.SBERT.net") |
| 1887 | (synopsis "Multilingual text embeddings") |
| 1888 | (description "This framework provides an easy method to compute dense |
| 1889 | vector representations for sentences, paragraphs, and images. The models are |
| 1890 | based on transformer networks like BERT / RoBERTa / XLM-RoBERTa and achieve |
| 1891 | state-of-the-art performance in various tasks. Text is embedded in vector |
| 1892 | space such that similar text are closer and can efficiently be found using |
| 1893 | cosine similarity. |
| 1894 | |
| 1895 | This package provides easy access to pretrained models for more than 100 |
| 1896 | languages, fine-tuned for various use-cases. |
| 1897 | |
| 1898 | Further, this framework allows an easy fine-tuning of custom embeddings |
| 1899 | models, to achieve maximal performance on your specific task.") |
| 1900 | (license license:asl2.0))) |
| 1901 | |
| 1902 | (define-public python-spacy-legacy |
| 1903 | (package |
| 1904 | (name "python-spacy-legacy") |
| 1905 | (version "3.0.12") |
| 1906 | (source (origin |
| 1907 | (method url-fetch) |
| 1908 | (uri (pypi-uri "spacy-legacy" version)) |
| 1909 | (sha256 |
| 1910 | (base32 |
| 1911 | "0x57aw1qgjrzgapsv1cwymqlck2anqm1bisvryhpq7bfkc66wzdk")))) |
| 1912 | (build-system pyproject-build-system) |
| 1913 | ;; This package depends on spacy, which depends on this package. |
| 1914 | (arguments (list #:tests? #false)) |
| 1915 | (native-inputs (list python-pytest python-setuptools python-wheel)) |
| 1916 | (home-page "https://spacy.io") |
| 1917 | (synopsis "Legacy registered functions for spaCy backwards compatibility") |
| 1918 | (description |
| 1919 | "This package contains legacy registered functions for spaCy backwards |
| 1920 | compatibility.") |
| 1921 | (license license:expat))) |
| 1922 | |
| 1923 | (define-public python-spacy-loggers |
| 1924 | (package |
| 1925 | (name "python-spacy-loggers") |
| 1926 | (version "1.0.4") |
| 1927 | (source (origin |
| 1928 | (method url-fetch) |
| 1929 | (uri (pypi-uri "spacy-loggers" version)) |
| 1930 | (sha256 |
| 1931 | (base32 |
| 1932 | "0a5zjfxd0mci7pwda24ihg55whalpmjby4bvpgar2013f6zq7yg6")))) |
| 1933 | (build-system pyproject-build-system) |
| 1934 | ;; This package depends on spacy, which depends on this package. |
| 1935 | (arguments (list #:tests? #false)) |
| 1936 | (native-inputs (list python-pytest python-setuptools python-wheel)) |
| 1937 | (home-page "https://github.com/explosion/spacy-loggers") |
| 1938 | (synopsis "Logging utilities for SpaCy") |
| 1939 | (description "This package provides logging utilities for the SpaCy |
| 1940 | natural language processing framework.") |
| 1941 | (license license:expat))) |
| 1942 | |
| 1943 | (define-public python-spacy |
| 1944 | (package |
| 1945 | (name "python-spacy") |
| 1946 | (version "3.8.14") |
| 1947 | (source |
| 1948 | (origin |
| 1949 | (method git-fetch) |
| 1950 | (uri (git-reference |
| 1951 | (url "https://github.com/explosion/spaCy") |
| 1952 | (commit (string-append "release-v" version)))) |
| 1953 | (file-name (git-file-name name version)) |
| 1954 | (sha256 |
| 1955 | (base32 "0inb1x7gmf1j7kcr35v472hia4lkn1qr234snqkz27rqglzhvmy3")))) |
| 1956 | (build-system pyproject-build-system) |
| 1957 | (arguments |
| 1958 | (list |
| 1959 | ;; tests: 3605 passed, 1232 skipped, 6 deselected, 24 xfailed, 7 |
| 1960 | ;; xpassed, 5 subtests passed |
| 1961 | #:test-flags |
| 1962 | #~(list #$@(map (lambda (test) |
| 1963 | (string-append "--deselect=tests/" test)) |
| 1964 | (list |
| 1965 | ;; Fail to compare setup.cfg and requirements.txt |
| 1966 | ;; versions. |
| 1967 | "package/test_requirements.py::test_build_dependencies" |
| 1968 | ;; Result |
| 1969 | ;; DeprecationWarning('datetime.datetime.utcfromtimestamp() |
| 1970 | ;; is deprecated and scheduled for removal in a future |
| 1971 | ;; version. Use timezone-aware objects to represent |
| 1972 | ;; datetimes in UTC: |
| 1973 | ;; datetime.datetime.fromtimestamp(timestamp, |
| 1974 | ;; datetime.UTC).' |
| 1975 | "test_cli_app.py::test_project_assets" |
| 1976 | "test_cli_app.py::test_project_push_pull" |
| 1977 | ;; AssertionError: Registry 'loggers' missing entries. |
| 1978 | "test_registry_population.py::test_registry_entries" |
| 1979 | ;; Tests require network access. |
| 1980 | "test_cli.py::test_download_compatibility" |
| 1981 | "test_cli.py::test_validate_compatibility_table"))) |
| 1982 | #:phases |
| 1983 | #~(modify-phases %standard-phases |
| 1984 | (add-before 'check 'remove-local-spacy |
| 1985 | (lambda _ |
| 1986 | (copy-recursively "spacy/tests" "tests") |
| 1987 | ;; This would otherwise interfere with finding the installed |
| 1988 | ;; spacy when running tests. |
| 1989 | (delete-file-recursively "spacy")))))) |
| 1990 | (propagated-inputs (list python-catalogue |
| 1991 | python-cymem |
| 1992 | python-jinja2 |
| 1993 | python-murmurhash |
| 1994 | python-numpy |
| 1995 | python-preshed |
| 1996 | python-pydantic |
| 1997 | python-requests |
| 1998 | python-spacy-legacy |
| 1999 | python-spacy-loggers |
| 2000 | python-srsly |
| 2001 | python-thinc |
| 2002 | python-tqdm |
| 2003 | python-typer |
| 2004 | python-wasabi |
| 2005 | python-weasel)) |
| 2006 | (native-inputs |
| 2007 | (list python-cython |
| 2008 | python-pytest |
| 2009 | python-mock |
| 2010 | python-setuptools)) |
| 2011 | (home-page "https://spacy.io") |
| 2012 | (synopsis "Natural Language Processing (NLP) in Python") |
| 2013 | (description |
| 2014 | "SpaCy is a library for advanced Natural Language Processing in Python |
| 2015 | and Cython. It comes with pretrained pipelines and currently supports |
| 2016 | tokenization and training for 70+ languages. It features state-of-the-art |
| 2017 | speed and neural network models for tagging, parsing, named entity |
| 2018 | recognition, text classification and more, multi-task learning with pretrained |
| 2019 | transformers like BERT, as well as a production-ready training system and easy |
| 2020 | model packaging, deployment and workflow management.") |
| 2021 | (license license:expat))) |
| 2022 | |
| 2023 | (define-public python-stanza |
| 2024 | (package |
| 2025 | (name "python-stanza") |
| 2026 | (version "1.10.1") |
| 2027 | (source |
| 2028 | (origin |
| 2029 | (method git-fetch) |
| 2030 | (uri (git-reference |
| 2031 | (url "https://github.com/stanfordnlp/stanza") |
| 2032 | (commit (string-append "v" version)))) |
| 2033 | (sha256 |
| 2034 | (base32 "0zcpzmbv0aafircl12m3x5999hxpg2hzm1xxv97pz09y4v589snj")))) |
| 2035 | (build-system pyproject-build-system) |
| 2036 | (arguments |
| 2037 | (list |
| 2038 | #:phases |
| 2039 | #~(modify-phases %standard-phases |
| 2040 | (add-before 'check 'set-up-check |
| 2041 | (lambda _ |
| 2042 | ;; Cherry pick from stanza/tests/setup.py, |
| 2043 | ;; which downloads many datasets |
| 2044 | (mkdir-p "stanza_test/out") |
| 2045 | (copy-file "stanza/tests/data/example_french.json" |
| 2046 | "stanza_test/out/example_french.json")))) |
| 2047 | ;; tests: 288 passed, 361 deselected, 1 warning |
| 2048 | #:test-flags |
| 2049 | #~(list |
| 2050 | "-k" |
| 2051 | (string-join |
| 2052 | '("not CoreNLP" "EnglishPipeline" "FrenchPipeline" |
| 2053 | "SentimentPipeline" "TestTrainer" |
| 2054 | "amt_annotator" "arabic_pos" "bert" |
| 2055 | "charlm" "conllu" "convert_units" |
| 2056 | "data_objects" "defaultdict_config" |
| 2057 | "depparse" "dictionary" "download" |
| 2058 | "ensemble" "example" "finetune" "install" |
| 2059 | "langid" "lemmatizer" "long_paragraph" "long_tokens" |
| 2060 | "model" "morphology" "multilingual" "mwt" |
| 2061 | "pipeline_" "pretrain" "process_doc" |
| 2062 | "read_snippets" "register" "reload" |
| 2063 | "requirements" "resources" "retag" |
| 2064 | "score" "semgrex" "serialized" "server_" "ssurgeon" |
| 2065 | "tagger" "test_core" "test_one_sentence" "test_tokenizer" |
| 2066 | "text_processing" "tokenize_files" "tokensregex" |
| 2067 | "train_pipeline" "training" "tsurgeon") |
| 2068 | " and not ") ;exclude tests requiring datasets |
| 2069 | "stanza/tests"))) |
| 2070 | (native-inputs (list python-pytest |
| 2071 | python-setuptools |
| 2072 | python-transformers)) |
| 2073 | (propagated-inputs (list python-emoji |
| 2074 | python-networkx |
| 2075 | python-numpy |
| 2076 | python-protobuf |
| 2077 | python-pytorch |
| 2078 | python-requests |
| 2079 | python-tqdm)) |
| 2080 | (home-page "https://stanfordnlp.github.io/stanza/") |
| 2081 | (synopsis "Stanford NLP Python library for many human languages") |
| 2082 | (description |
| 2083 | "Stanza is a collection of accurate and efficient tools |
| 2084 | for the linguistic analysis of many human languages. Starting from raw text, |
| 2085 | Stanza divides it into sentences and words, and then can recognize |
| 2086 | parts of speech and entities, do syntactic analysis, and more.") |
| 2087 | (license license:asl2.0))) |
| 2088 | |
| 2089 | (define-public onnx |
| 2090 | (package |
| 2091 | (name "onnx") |
| 2092 | (version "1.17.0") |
| 2093 | (source (origin |
| 2094 | (method git-fetch) |
| 2095 | (uri (git-reference |
| 2096 | (url "https://github.com/onnx/onnx") |
| 2097 | (commit (string-append "v" version)))) |
| 2098 | (sha256 |
| 2099 | (base32 |
| 2100 | "1i6bh4z2xzz1maykr0xmrwfybm6i3g38vnx7hsls8hr58rdr30zn")) |
| 2101 | (file-name (git-file-name name version)) |
| 2102 | (patches (search-patches |
| 2103 | "onnx-shared-libraries.patch" |
| 2104 | "onnx-skip-model-downloads.patch")) |
| 2105 | (modules '((guix build utils))) |
| 2106 | (snippet '(delete-file-recursively "third_party")))) |
| 2107 | (build-system pyproject-build-system) |
| 2108 | (arguments |
| 2109 | (list |
| 2110 | ;; python-nbval depends transitively on Rust. |
| 2111 | #:tests? |
| 2112 | (->bool (member (or (%current-target-system) |
| 2113 | (%current-system)) |
| 2114 | (package-transitive-supported-systems python-nbval))) |
| 2115 | ;; tests: 4610 passed, 3044 skipped, 7 deselected, 13 warnings, 98 |
| 2116 | ;; subtests passed |
| 2117 | #:test-flags |
| 2118 | ;; XXX: Onnx is on old version and might need to be updated, see: |
| 2119 | ;; <https://codeberg.org/guix/guix/issues/9443>. |
| 2120 | #~(list #$@(map (lambda (ls) (string-append "--deselect=onnx/" |
| 2121 | (string-join ls "::"))) |
| 2122 | '(("test/test_backend_reference.py" |
| 2123 | "OnnxBackendNodeModelTest" |
| 2124 | "test_affine_grid_2d_align_corners_expanded_cpu") |
| 2125 | ("test/test_backend_reference.py" |
| 2126 | "OnnxBackendNodeModelTest" |
| 2127 | "test_affine_grid_2d_expanded_cpu") |
| 2128 | ("test/test_backend_reference.py" |
| 2129 | "OnnxBackendNodeModelTest" |
| 2130 | "test_affine_grid_3d_align_corners_expanded_cpu") |
| 2131 | ("test/test_backend_reference.py" |
| 2132 | "OnnxBackendNodeModelTest" |
| 2133 | "test_affine_grid_3d_expanded_cpu") |
| 2134 | ("test/tools_test.py" "TestToolsFunctions" |
| 2135 | "test_replace_range") |
| 2136 | ("test/tools_test.py" |
| 2137 | "TestToolsFunctions" "test_replace_range_function") |
| 2138 | ("test/tools_test.py" |
| 2139 | "TestToolsFunctions" "test_replace_range_graph")))) |
| 2140 | #:phases |
| 2141 | #~(modify-phases %standard-phases |
| 2142 | (add-before 'build 'pass-cmake-arguments |
| 2143 | (lambda* (#:key outputs tests? #:allow-other-keys) |
| 2144 | ;; For derived package use |
| 2145 | (substitute* "CMakeLists.txt" |
| 2146 | (("set\\(ONNX_ROOT.*") "") |
| 2147 | (("\\$\\{ROOT_DIR\\}(/tools.*)" _ rest) |
| 2148 | (string-append "${PROJECT_SOURCE_DIR}" rest))) |
| 2149 | ;; Pass options to the CMake-based build process. |
| 2150 | (define out |
| 2151 | (assoc-ref outputs "out")) |
| 2152 | |
| 2153 | (define args |
| 2154 | ;; Copy arguments from 'cmake-build-system', plus ask |
| 2155 | ;; for shared libraries. |
| 2156 | (list "-DCMAKE_BUILD_TYPE=RelWithDebInfo" |
| 2157 | (string-append "-DCMAKE_INSTALL_PREFIX=" out) |
| 2158 | "-DCMAKE_INSTALL_LIBDIR=lib" |
| 2159 | "-DCMAKE_INSTALL_RPATH_USE_LINK_PATH=TRUE" |
| 2160 | (string-append "-DCMAKE_INSTALL_RPATH=" out |
| 2161 | "/lib") |
| 2162 | "-DCMAKE_VERBOSE_MAKEFILE=ON" |
| 2163 | (string-append "-DONNX_BUILD_TESTS=" |
| 2164 | (if tests? "ON" "OFF")) |
| 2165 | "-DBUILD_SHARED_LIBS=ON" |
| 2166 | "-DONNX_USE_PROTOBUF_SHARED_LIBS=ON" |
| 2167 | (string-append |
| 2168 | "-DONNX_ROOT=" #$(package-source this-package)))) |
| 2169 | |
| 2170 | ;; This environment variable is honored by 'setup.py', |
| 2171 | ;; which passes it down to 'cmake'. |
| 2172 | (setenv "CMAKE_ARGS" (string-join args)) |
| 2173 | |
| 2174 | ;; This one is honored by 'setup.py' and passed to 'make -j'. |
| 2175 | (setenv "MAX_JOBS" |
| 2176 | (number->string (parallel-job-count))))) |
| 2177 | (add-before 'check 'make-test-directory-writable |
| 2178 | (lambda _ |
| 2179 | ;; Make things writable for tests. |
| 2180 | (setenv "HOME" (getcwd)) |
| 2181 | (for-each make-file-writable |
| 2182 | (find-files "onnx/examples" "." |
| 2183 | #:directories? #t)))) |
| 2184 | (add-after 'install 'install-from-cmake |
| 2185 | (lambda _ |
| 2186 | ;; Run "make install" in the build tree 'setup.py' |
| 2187 | ;; created for CMake so that libonnx.so, |
| 2188 | ;; libonnx_proto.so, etc. are installed. |
| 2189 | (invoke "make" "install" |
| 2190 | "-C" ".setuptools-cmake-build")))))) |
| 2191 | (native-inputs |
| 2192 | (append |
| 2193 | (list cmake-minimal |
| 2194 | googletest |
| 2195 | pybind11-2 |
| 2196 | python-fb-re2 |
| 2197 | python-parameterized |
| 2198 | python-pytest |
| 2199 | python-setuptools |
| 2200 | python-wheel) |
| 2201 | (filter |
| 2202 | (lambda (pkg) |
| 2203 | (member (or (%current-target-system) |
| 2204 | (%current-system)) |
| 2205 | (package-transitive-supported-systems pkg))) |
| 2206 | (list python-nbval)))) |
| 2207 | (inputs |
| 2208 | (list protobuf)) |
| 2209 | (propagated-inputs |
| 2210 | (list python-numpy python-protobuf python-six python-tabulate |
| 2211 | python-typing-extensions)) |
| 2212 | (home-page "https://onnx.ai/") |
| 2213 | (synopsis "Open Neural Network Exchange") |
| 2214 | (description |
| 2215 | "@acronym{ONNX, Open Neural Network Exchange} is a format for AI models, |
| 2216 | both deep learning and traditional @acronym{ML, machine learning}. It defines |
| 2217 | an extensible computation graph model, as well as definitions of built-in |
| 2218 | operators and standard data types.") |
| 2219 | (license license:expat))) |
| 2220 | |
| 2221 | (define-public onnx-for-onnxruntime |
| 2222 | (package |
| 2223 | (inherit onnx) |
| 2224 | (name "onnx-for-onnxruntime") |
| 2225 | (source |
| 2226 | (origin |
| 2227 | (inherit (package-source onnx)) |
| 2228 | (patches (append (search-patches "onnx-1.17.0-for-onnxruntime.patch") |
| 2229 | (origin-patches (package-source onnx)))))) |
| 2230 | ;; Use cmake-build-system instead of pyproject-build-system so that |
| 2231 | ;; only the C++ static library is built and installed. The Python |
| 2232 | ;; onnx package (with schemas registered) is provided by regular |
| 2233 | ;; onnx; this variant has ONNX_DISABLE_STATIC_REGISTRATION=ON which |
| 2234 | ;; would break onnx.shape_inference if its Python files ended up on |
| 2235 | ;; PYTHONPATH. |
| 2236 | (build-system cmake-build-system) |
| 2237 | (arguments |
| 2238 | (list |
| 2239 | #:tests? #f ; tests fail without schema registration |
| 2240 | #:configure-flags |
| 2241 | #~(list "-DBUILD_SHARED_LIBS=OFF" |
| 2242 | "-DONNX_USE_PROTOBUF_SHARED_LIBS=OFF" |
| 2243 | ;; Disable ONNX's schema registration so onnxruntime |
| 2244 | ;; handles it. See |
| 2245 | ;; <https://onnxruntime.ai/docs/build/dependencies.html>. |
| 2246 | "-DONNX_DISABLE_STATIC_REGISTRATION=ON" |
| 2247 | "-DONNX_BUILD_TESTS=OFF" |
| 2248 | "-DONNX_BUILD_BENCHMARKS=OFF" |
| 2249 | "-DONNX_ML=ON" |
| 2250 | "-DONNX_USE_LITE_PROTO=OFF" |
| 2251 | "-DBUILD_ONNX_PYTHON=OFF"))) |
| 2252 | (inputs (list protobuf-static-for-onnxruntime)) |
| 2253 | (native-inputs (list python-minimal-wrapper pybind11)) |
| 2254 | (propagated-inputs '()) |
| 2255 | (description |
| 2256 | "This variant of the ONNX package is built as a static C++ library |
| 2257 | with schema registration disabled, for linking into onnxruntime. It |
| 2258 | should not be used standalone."))) |
| 2259 | |
| 2260 | (define-public onnx-optimizer |
| 2261 | (package |
| 2262 | (name "onnx-optimizer") |
| 2263 | ;; Note: 0.3.x is *more* recent than 1.5.0. |
| 2264 | (version "0.3.19") |
| 2265 | (home-page "https://github.com/onnx/optimizer") |
| 2266 | (source (origin |
| 2267 | (method git-fetch) |
| 2268 | (uri (git-reference |
| 2269 | (url home-page) |
| 2270 | (commit (string-append "v" version)))) |
| 2271 | (sha256 |
| 2272 | (base32 |
| 2273 | "1mx3hsl42na6fr05nh2x3j9kxm56cpfmwk6lwl2cfq9zs3gv929w")) |
| 2274 | (file-name (git-file-name name version)) |
| 2275 | (patches (search-patches "onnx-optimizer-system-library.patch")) |
| 2276 | (modules '((guix build utils))) |
| 2277 | (snippet '(delete-file-recursively "third_party")))) |
| 2278 | (build-system pyproject-build-system) |
| 2279 | (arguments |
| 2280 | ;; reuse build system tweaks |
| 2281 | (substitute-keyword-arguments (package-arguments onnx) |
| 2282 | ((#:phases phases) |
| 2283 | #~(modify-phases #$phases |
| 2284 | (add-after 'pass-cmake-arguments |
| 2285 | 'pass-onnx-optimizer-cmake-arguments |
| 2286 | (lambda _ |
| 2287 | (setenv |
| 2288 | "CMAKE_ARGS" |
| 2289 | (string-append |
| 2290 | (getenv "CMAKE_ARGS") |
| 2291 | " -DONNX_OPT_USE_SYSTEM_PROTOBUF=ON" |
| 2292 | " -DCMAKE_CXX_FLAGS=\"-DONNX_ML=1 -DONNX_NAMESPACE=onnx\"")))) |
| 2293 | (replace 'check |
| 2294 | (lambda* (#:key tests? #:allow-other-keys) |
| 2295 | (if tests? |
| 2296 | (invoke "pytest" "-vv" "-k" |
| 2297 | ;; These tests fail with upstream ONNX: |
| 2298 | ;; https://github.com/onnx/optimizer/issues/138 |
| 2299 | (string-append |
| 2300 | "not test_fuse_matmul" |
| 2301 | " and not test_fuse_consecutive" |
| 2302 | " and not test_fuse_transpose"))))))))) |
| 2303 | (native-inputs |
| 2304 | (append |
| 2305 | (list cmake-minimal |
| 2306 | python-pytest |
| 2307 | python-setuptools) |
| 2308 | (filter |
| 2309 | (lambda (pkg) |
| 2310 | (member (or (%current-target-system) |
| 2311 | (%current-system)) |
| 2312 | (package-transitive-supported-systems pkg))) |
| 2313 | (list python-nbval)))) |
| 2314 | (inputs |
| 2315 | ;; Must use the same protobuf as onnx to avoid loading two protobuf |
| 2316 | ;; libraries causing duplicate schema registration. |
| 2317 | (list onnx protobuf pybind11-2)) |
| 2318 | (propagated-inputs |
| 2319 | (list python-numpy)) |
| 2320 | (synopsis "Library to optimize ONNX models") |
| 2321 | (description |
| 2322 | "This package provides a C++ and Python library for performing arbitrary |
| 2323 | optimizations on ONNX models, as well as a growing list of prepackaged |
| 2324 | optimization passes. |
| 2325 | |
| 2326 | Not all possible optimizations can be directly implemented on ONNX graphs--- |
| 2327 | some will need additional backend-specific information---but many can, and the |
| 2328 | aim is to provide all such passes along with ONNX so that they can be re-used |
| 2329 | with a single function call.") |
| 2330 | (license license:expat))) |
| 2331 | |
| 2332 | (define-public onnxruntime |
| 2333 | (package |
| 2334 | (name "onnxruntime") |
| 2335 | (version "1.22.0") |
| 2336 | (source |
| 2337 | (origin |
| 2338 | (method git-fetch) |
| 2339 | (uri (git-reference |
| 2340 | (url "https://github.com/microsoft/onnxruntime") |
| 2341 | (commit (string-append "v" version)))) |
| 2342 | (file-name (git-file-name name version)) |
| 2343 | (sha256 |
| 2344 | (base32 "0z2s79l4wdilssw9lmj319ypyyqi2y0dx0fpwr2yhq8bax3ci50n")) |
| 2345 | (patches |
| 2346 | (search-patches |
| 2347 | ;; SplitToSequence operator was missing bool type support, despite |
| 2348 | ;; the ONNX spec allowing it in models. Fixed upstream in 1.23+, |
| 2349 | ;; but we cannot upgrade because ROCm support was removed in 1.23. |
| 2350 | ;; <https://github.com/microsoft/onnxruntime/pull/24929> |
| 2351 | "onnxruntime-1.22.0-splittosequence-bool.patch")))) |
| 2352 | (build-system cmake-build-system) |
| 2353 | (arguments |
| 2354 | (list |
| 2355 | #:modules '((guix build cmake-build-system) |
| 2356 | (guix build utils) |
| 2357 | ((guix build pyproject-build-system) #:prefix py:)) |
| 2358 | #:imported-modules (append %cmake-build-system-modules |
| 2359 | %pyproject-build-system-modules) |
| 2360 | #:configure-flags |
| 2361 | #~(list "-Donnxruntime_BUILD_UNIT_TESTS=OFF" |
| 2362 | "-Donnxruntime_BUILD_SHARED_LIB=ON" |
| 2363 | "-Donnxruntime_ENABLE_LTO=ON" |
| 2364 | "-Donnxruntime_ENABLE_PYTHON=ON" |
| 2365 | "-Donnxruntime_USE_FULL_PROTOBUF=ON" |
| 2366 | ;; Link static protobuf, matching upstream's vcpkg approach. |
| 2367 | ;; This keeps the protobuf descriptor pool inside |
| 2368 | ;; onnxruntime_pybind11_state.so separate from the shared |
| 2369 | ;; libprotobuf.so used by the regular onnx Python package. |
| 2370 | "-DProtobuf_USE_STATIC_LIBS=ON" |
| 2371 | ;; XXX: Fixes build with gcc@14. |
| 2372 | "-DCMAKE_CXX_FLAGS=-Wl,-z,noexecstack") |
| 2373 | #:phases |
| 2374 | #~(modify-phases %standard-phases |
| 2375 | (add-after 'unpack 'chdir |
| 2376 | (lambda _ |
| 2377 | (chdir "cmake"))) |
| 2378 | (add-after 'unpack 'relax-dependencies |
| 2379 | (lambda _ |
| 2380 | (with-output-to-file "cmake/external/eigen.cmake" |
| 2381 | (lambda _ |
| 2382 | (display "find_package(Eigen3 REQUIRED)\n"))) |
| 2383 | (substitute* "cmake/external/abseil-cpp.cmake" |
| 2384 | (("20240722") |
| 2385 | (car (string-split #$(package-version |
| 2386 | (this-package-input "abseil-cpp")) |
| 2387 | #\.)))))) |
| 2388 | (add-after 'unpack 'fix-numpy |
| 2389 | ;; Prevent TypeError: only 0-dimensional arrays can be converted |
| 2390 | ;; to Python scalars. |
| 2391 | ;; Fixed upstream in 71d2cd0173c863ee67b3a6803aea248c563d5427. |
| 2392 | (lambda _ |
| 2393 | (substitute* |
| 2394 | "onnxruntime/python/tools/transformers/fusion_attention.py" |
| 2395 | (("float\\(mul_val)") "mul_val.item()")))) |
| 2396 | (add-after 'install 'build-python |
| 2397 | (lambda _ |
| 2398 | (invoke "python3" "../setup.py" "bdist_wheel"))) |
| 2399 | (add-after 'build-python 'install-python |
| 2400 | (lambda* (#:key inputs #:allow-other-keys) |
| 2401 | ((assoc-ref py:%standard-phases 'install) |
| 2402 | #:inputs inputs |
| 2403 | #:outputs `(("out" . ,#$output:python))))) |
| 2404 | (add-after 'install-python 'add-install-to-pythonpath |
| 2405 | (lambda* (#:key inputs #:allow-other-keys) |
| 2406 | ((assoc-ref py:%standard-phases 'add-install-to-pythonpath) |
| 2407 | #:inputs inputs |
| 2408 | #:outputs `(("out" . ,#$output:python))))) |
| 2409 | (delete 'check) |
| 2410 | (add-after 'add-install-to-pythonpath 'check |
| 2411 | (lambda* (#:key tests? #:allow-other-keys) |
| 2412 | (with-directory-excursion "../onnxruntime/test/python" |
| 2413 | ((assoc-ref py:%standard-phases 'check) |
| 2414 | #:tests? tests? |
| 2415 | #:test-flags |
| 2416 | `(;; XXX: NotImplementedError |
| 2417 | "--ignore-glob=quantization/*" |
| 2418 | ;; XXX: These tests require transformer models or have |
| 2419 | ;; import issues. |
| 2420 | "--ignore=transformers/test_generation.py" |
| 2421 | "--ignore=transformers/test_gpt2_benchmark.py" |
| 2422 | "--ignore=transformers/test_gpt2_to_onnx.py" |
| 2423 | "--ignore=transformers/test_optimizer_huggingface_bert.py" |
| 2424 | "--ignore=transformers/test_parity_huggingface_gpt_attention.py" |
| 2425 | "--ignore=transformers/test_shape_infer_helper.py" |
| 2426 | ;; XXX: onnxscript ModuleNotFound |
| 2427 | "--ignore=transformers/test_gelu_fusions.py" |
| 2428 | "--ignore=transformers/test_gemma3_vision.py" |
| 2429 | ;; XXX: PyTorch 2.9 ONNX exporter requires python-onnxscript. |
| 2430 | ;; Skip exporter-dependent tests until packaged/enabled. |
| 2431 | "--ignore=test_pytorch_export_contrib_ops.py" |
| 2432 | "--ignore=transformers/test_parity_gelu.py" |
| 2433 | "--ignore=transformers/test_parity_layernorm.py" |
| 2434 | "--ignore=transformers/test_phi_vision.py" |
| 2435 | ;; XXX: Other failing tests. |
| 2436 | "-k" ,(string-append |
| 2437 | "not test_gelu_is_fused_by_default" |
| 2438 | " and not test_inverse")))))) |
| 2439 | (add-after 'check 'python-sanity-check |
| 2440 | (lambda* (#:key tests? inputs #:allow-other-keys) |
| 2441 | ((assoc-ref py:%standard-phases 'sanity-check) |
| 2442 | #:inputs `(("sanity-check.py" . ,#$(default-sanity-check.py)) |
| 2443 | ,@inputs) |
| 2444 | #:outputs `(("out" . ,#$output:python)))))))) |
| 2445 | (outputs (list "out" "python")) |
| 2446 | (inputs |
| 2447 | (list abseil-cpp-20250127 |
| 2448 | boost |
| 2449 | cpuinfo |
| 2450 | dlpack |
| 2451 | c++-gsl |
| 2452 | date |
| 2453 | eigen-for-onnxruntime |
| 2454 | flatbuffers-23.5 |
| 2455 | googletest |
| 2456 | nlohmann-json |
| 2457 | onnx-for-onnxruntime |
| 2458 | protobuf-static-for-onnxruntime |
| 2459 | pybind11-2 |
| 2460 | re2-next |
| 2461 | safeint |
| 2462 | zlib)) |
| 2463 | (native-inputs |
| 2464 | (list pkg-config |
| 2465 | python-einops |
| 2466 | python-wrapper |
| 2467 | python-numpy |
| 2468 | python-parameterized |
| 2469 | python-psutil |
| 2470 | python-pytest |
| 2471 | python-pytorch |
| 2472 | python-sentencepiece |
| 2473 | python-setuptools)) |
| 2474 | (propagated-inputs |
| 2475 | (list python-coloredlogs |
| 2476 | python-flatbuffers |
| 2477 | python-protobuf |
| 2478 | python-sympy)) |
| 2479 | (home-page "https://github.com/microsoft/onnxruntime") |
| 2480 | (synopsis "Cross-platform, high performance scoring engine for ML models") |
| 2481 | (description |
| 2482 | "ONNX Runtime is a performance-focused complete scoring engine |
| 2483 | for Open Neural Network Exchange (ONNX) models, with an open |
| 2484 | extensible architecture to continually address the latest developments |
| 2485 | in AI and Deep Learning. ONNX Runtime stays up to date with the ONNX |
| 2486 | standard with complete implementation of all ONNX operators, and |
| 2487 | supports all ONNX releases (1.2+) with both future and backwards |
| 2488 | compatibility.") |
| 2489 | (license license:expat))) |
| 2490 | |
| 2491 | (define-public gemmlowp |
| 2492 | (let ((commit "16e8662c34917be0065110bfcd9cc27d30f52fdf") |
| 2493 | (version "0.1") |
| 2494 | (revision "2")) |
| 2495 | (package |
| 2496 | (name "gemmlowp") |
| 2497 | (version (git-version version revision commit)) |
| 2498 | (source |
| 2499 | (origin |
| 2500 | (method git-fetch) |
| 2501 | (uri (git-reference |
| 2502 | (url "https://github.com/google/gemmlowp") |
| 2503 | (commit commit))) |
| 2504 | (file-name (git-file-name name version)) |
| 2505 | (sha256 |
| 2506 | (base32 "02xmrv921al94nihiqrvi8inlq6qc07j0zll3f9qi8322r31x83v")))) |
| 2507 | (arguments |
| 2508 | (list |
| 2509 | #:configure-flags |
| 2510 | #~(list #$@(match (%current-system) |
| 2511 | ((or "x86_64-linux" "i686-linux") |
| 2512 | '("-DCMAKE_CXX_FLAGS=-msse2")) |
| 2513 | (_ '())) |
| 2514 | "-DBUILD_SHARED_LIBS=ON") |
| 2515 | #:phases |
| 2516 | #~(modify-phases %standard-phases |
| 2517 | ;; This directory contains the CMakeLists.txt. |
| 2518 | (add-after 'unpack 'chdir |
| 2519 | (lambda _ |
| 2520 | (chdir "contrib"))) |
| 2521 | ;; There is no install target |
| 2522 | (replace 'install |
| 2523 | (lambda _ |
| 2524 | (let ((lib (string-append #$output "/lib/"))) |
| 2525 | (install-file "../build/libeight_bit_int_gemm.so" lib) |
| 2526 | (for-each |
| 2527 | (lambda (dir) |
| 2528 | (let ((target (string-append #$output |
| 2529 | "/include/gemmlowp/" dir))) |
| 2530 | (for-each (lambda (h) |
| 2531 | (install-file h target)) |
| 2532 | (find-files (string-append "../" dir) |
| 2533 | "\\.h$")))) |
| 2534 | '("meta" "profiling" "public" "fixedpoint" |
| 2535 | "eight_bit_int_gemm" "internal")))))))) |
| 2536 | (build-system cmake-build-system) |
| 2537 | (home-page "https://github.com/google/gemmlowp") |
| 2538 | (synopsis "Small self-contained low-precision GEMM library") |
| 2539 | (description |
| 2540 | "This is a small self-contained low-precision @dfn{general matrix |
| 2541 | multiplication} (GEMM) library. It is not a full linear algebra library. |
| 2542 | Low-precision means that the input and output matrix entries are integers on |
| 2543 | at most 8 bits. To avoid overflow, results are internally accumulated on more |
| 2544 | than 8 bits, and at the end only some significant 8 bits are kept.") |
| 2545 | (license license:asl2.0)))) |
| 2546 | |
| 2547 | (define-public dlib |
| 2548 | (package |
| 2549 | (name "dlib") |
| 2550 | (version "20.0") |
| 2551 | (source |
| 2552 | (origin |
| 2553 | (method git-fetch) |
| 2554 | (uri (git-reference |
| 2555 | (url "https://github.com/davisking/dlib.git") |
| 2556 | (commit (string-append "v" version)))) |
| 2557 | (file-name (git-file-name name version)) |
| 2558 | (sha256 |
| 2559 | (base32 "1k6jwbb7n842snd2znv0ix8zwa4a35y324jb7mpkj0vn9arzndam")) |
| 2560 | (modules '((guix build utils))) |
| 2561 | (snippet |
| 2562 | '(begin |
| 2563 | (delete-file-recursively "dlib/external"))))) |
| 2564 | (build-system cmake-build-system) |
| 2565 | (arguments |
| 2566 | (list |
| 2567 | #:configure-flags #~(list "-DBUILD_SHARED_LIBS=ON") |
| 2568 | #:phases |
| 2569 | #~(modify-phases %standard-phases |
| 2570 | (add-after 'unpack 'disable-asserts |
| 2571 | (lambda _ |
| 2572 | ;; config.h recommends explicitly enabling or disabling asserts |
| 2573 | ;; when building as a shared library. By default neither is set. |
| 2574 | (substitute* "dlib/config.h" |
| 2575 | (("^//#define DLIB_DISABLE_ASSERTS") "#define DLIB_DISABLE_ASSERTS")))) |
| 2576 | (add-after 'disable-asserts 'disable-failing-tests |
| 2577 | (lambda _ |
| 2578 | ;; One test times out on MIPS, so we need to disable it. |
| 2579 | ;; Others are flaky on some platforms. |
| 2580 | (let* ((system #$(or (%current-target-system) |
| 2581 | (%current-system))) |
| 2582 | (disabled-tests (cond |
| 2583 | ((string-prefix? "mips64" system) |
| 2584 | '("object_detector" ; timeout |
| 2585 | "data_io")) |
| 2586 | ((string-prefix? "armhf" system) |
| 2587 | '("learning_to_track")) |
| 2588 | ((string-prefix? "i686" system) |
| 2589 | '("optimization")) |
| 2590 | (else '())))) |
| 2591 | (for-each |
| 2592 | (lambda (test) |
| 2593 | (substitute* "dlib/test/makefile" |
| 2594 | (((string-append "SRC \\+= " test "\\.cpp")) ""))) |
| 2595 | disabled-tests)))) |
| 2596 | (add-before 'configure 'fix-test-build-error-with-gcc-14.3 |
| 2597 | (lambda _ |
| 2598 | (substitute* "dlib/test/CMakeLists.txt" |
| 2599 | ;; Some flags are activated specifically for GCC 14.0 but they |
| 2600 | ;; are needed for current GCC version (14.3.0). Let's assume |
| 2601 | ;; they are needed for the 14.x branch. |
| 2602 | (("14\\.0\\.1") |
| 2603 | "14.99") |
| 2604 | ;; Tests use deprecated functions from ffmpeg. |
| 2605 | (("-Werror\\)") |
| 2606 | "-Werror -Wno-error=deprecated-declarations)")))) |
| 2607 | (replace 'check |
| 2608 | (lambda _ |
| 2609 | ;; XXX: This causes a rebuild--however, trying to run the tests |
| 2610 | ;; without rebuilding causes a lot of errors. |
| 2611 | ;; Also, the official way is to rebuild. |
| 2612 | (mkdir "../source/dlib/test/build") |
| 2613 | (with-directory-excursion "../source/dlib/test/build" |
| 2614 | (invoke "cmake" "-DBUILD_SHARED_LIBS=ON" "..") |
| 2615 | (invoke "cmake" "--build" "." "--config" "Release") |
| 2616 | (invoke "./dtest" "--runall"))))))) |
| 2617 | (native-inputs |
| 2618 | (list pkg-config)) |
| 2619 | (inputs |
| 2620 | (list ffmpeg |
| 2621 | giflib |
| 2622 | libjpeg-turbo |
| 2623 | libjxl |
| 2624 | libpng |
| 2625 | libwebp |
| 2626 | libx11 |
| 2627 | openblas |
| 2628 | pybind11-2 |
| 2629 | zlib)) |
| 2630 | (synopsis |
| 2631 | "Toolkit for making machine learning and data analysis applications in |
| 2632 | C++") |
| 2633 | (description |
| 2634 | "Dlib is a modern C++ toolkit containing machine learning algorithms and |
| 2635 | tools. It is used in both industry and academia in a wide range of domains |
| 2636 | including robotics, embedded devices, mobile phones, and large high |
| 2637 | performance computing environments.") |
| 2638 | (home-page "http://dlib.net") |
| 2639 | (license license:boost1.0))) |
| 2640 | |
| 2641 | (define-public python-scikit-learn |
| 2642 | (package |
| 2643 | (name "python-scikit-learn") |
| 2644 | (version "1.9.0") |
| 2645 | (source |
| 2646 | (origin |
| 2647 | (method git-fetch) |
| 2648 | (uri (git-reference |
| 2649 | (url "https://github.com/scikit-learn/scikit-learn") |
| 2650 | (commit version))) |
| 2651 | (file-name (git-file-name name version)) |
| 2652 | (sha256 |
| 2653 | (base32 "0ay9zqlv1gimi6smxxjf29rvq3d923hdp34x731gd57rvbf22h3p")))) |
| 2654 | (build-system pyproject-build-system) |
| 2655 | (arguments |
| 2656 | (list |
| 2657 | ;; tests: 35838 passed, 11167 skipped, 226 xfailed, 119 xpassed, 5155 warnings |
| 2658 | #:test-flags |
| 2659 | #~(list "--numprocesses" (number->string (min 8 (parallel-job-count))) |
| 2660 | "-m" "not network" |
| 2661 | "--pyargs" "sklearn") |
| 2662 | #:phases |
| 2663 | #~(modify-phases %standard-phases |
| 2664 | (add-after 'unpack 'patch-pyproject |
| 2665 | (lambda _ |
| 2666 | (substitute* "pyproject.toml" |
| 2667 | ;; FIXME: This line contains regexps and breaks toml parser. |
| 2668 | (("notice-rgx.*") "")))) |
| 2669 | (add-before 'build 'configure |
| 2670 | (lambda _ |
| 2671 | (setenv "SKLEARN_BUILD_PARALLEL" |
| 2672 | (number->string (parallel-job-count))))) |
| 2673 | (add-before 'check 'pre-check |
| 2674 | (lambda _ |
| 2675 | ;; Restrict OpenBLAS threads to prevent segfaults while |
| 2676 | ;; testing! |
| 2677 | (setenv "OPENBLAS_NUM_THREADS" "1") |
| 2678 | ;; Some tests require write access to $HOME. |
| 2679 | (setenv "HOME" "/tmp") |
| 2680 | (delete-file-recursively "sklearn")))))) |
| 2681 | (inputs |
| 2682 | (list openblas)) |
| 2683 | (native-inputs |
| 2684 | (list gfortran |
| 2685 | python-meson |
| 2686 | python-pandas |
| 2687 | python-pytest |
| 2688 | python-pytest-xdist)) |
| 2689 | (propagated-inputs |
| 2690 | (list python-joblib |
| 2691 | python-narwhals-minimal ;to reduce closure size |
| 2692 | python-numpy |
| 2693 | python-scipy |
| 2694 | python-threadpoolctl)) |
| 2695 | (home-page "https://scikit-learn.org/") |
| 2696 | (synopsis "Machine Learning in Python") |
| 2697 | (description |
| 2698 | "Scikit-learn provides simple and efficient tools for data mining and |
| 2699 | data analysis.") |
| 2700 | (license license:bsd-3))) |
| 2701 | |
| 2702 | ;; 1.7 intorduced breaking changes in API. |
| 2703 | (define-public python-scikit-learn-1.6 |
| 2704 | (package |
| 2705 | (inherit python-scikit-learn) |
| 2706 | (name "python-scikit-learn") |
| 2707 | (version "1.6.1") |
| 2708 | (source |
| 2709 | (origin |
| 2710 | (method git-fetch) |
| 2711 | (uri (git-reference |
| 2712 | (url "https://github.com/scikit-learn/scikit-learn") |
| 2713 | (commit version))) |
| 2714 | (file-name (git-file-name name version)) |
| 2715 | (sha256 |
| 2716 | (base32 "08z1b58n31grfvl42wi6rdwrfhrdhnzkkxhg19iag3zkvkcvxqjl")))) |
| 2717 | (arguments |
| 2718 | (substitute-keyword-arguments arguments |
| 2719 | ((#:test-flags flags) |
| 2720 | #~(list "--numprocesses" (number->string (min 8 (parallel-job-count))) |
| 2721 | "-m" "not network" |
| 2722 | "-k" (string-join |
| 2723 | (list "not test_ard_accuracy_on_easy_problem" |
| 2724 | "test_check_inplace_ensure_writeable" |
| 2725 | "test_check_is_fitted_with_attributes" |
| 2726 | "test_covariance" |
| 2727 | "test_estimators" |
| 2728 | "test_newton_cholesky_fallback_to_lbfgs[42]" |
| 2729 | "test_logistic_regression_path_convergence_fail" |
| 2730 | "test_linalg_warning_with_newton_solver[42]" |
| 2731 | "test_ledoit_wolf" |
| 2732 | "test_mcd" |
| 2733 | "test_mcd_issue1127" |
| 2734 | "test_mcd_support_covariance_is_zero" |
| 2735 | "test_oas" |
| 2736 | "test_shrunk_covariance" |
| 2737 | "test_toy_ard_object") |
| 2738 | " and not "))) |
| 2739 | ((#:phases phases) |
| 2740 | #~(modify-phases #$phases |
| 2741 | (add-after 'unpack 'remove-broken-tests |
| 2742 | (lambda _ |
| 2743 | ;; ImportError: cannot import name 'ColMajor' from |
| 2744 | ;; 'sklearn.utils._cython_blas' |
| 2745 | ;; (<...>/_cython_blas.cpython-311-x86_64-linux-gnu.so) |
| 2746 | (delete-file-recursively "sklearn/utils/tests/test_cython_blas.py"))))))))) |
| 2747 | |
| 2748 | (define-public python-mord |
| 2749 | (package |
| 2750 | (name "python-mord") |
| 2751 | (version "0.7") |
| 2752 | (source |
| 2753 | (origin |
| 2754 | (method url-fetch) |
| 2755 | (uri (pypi-uri "mord" version)) |
| 2756 | (sha256 |
| 2757 | (base32 "1cvv9b9w69v0inq0zgcw0vmkiq3zn9q9r6clkynpzjik9rrh405n")))) |
| 2758 | (build-system pyproject-build-system) |
| 2759 | ;; v0.7 does not provide any test cases |
| 2760 | ;; v0.6 relies on deprecated scikit-learn functionality |
| 2761 | (arguments `(#:tests? #f)) |
| 2762 | (inputs (list python-numpy python-scipy python-scikit-learn)) |
| 2763 | (native-inputs (list python-setuptools python-wheel)) |
| 2764 | (home-page "https://pypi.org/project/mord/") |
| 2765 | (synopsis "Ordinal regression models for scikit-learn") |
| 2766 | (description |
| 2767 | "This package provides a collection of ordinal regression models for |
| 2768 | machine learning in Python. They are intended to be used with scikit-learn |
| 2769 | and are compatible with its API.") |
| 2770 | (license license:bsd-3))) |
| 2771 | |
| 2772 | (define-public python-thinc |
| 2773 | (package |
| 2774 | (name "python-thinc") |
| 2775 | (version "8.3.13") |
| 2776 | (source |
| 2777 | (origin |
| 2778 | (method git-fetch) |
| 2779 | (uri (git-reference |
| 2780 | (url "https://github.com/explosion/thinc") |
| 2781 | ;; XXX: It may be "release-v" as well. |
| 2782 | (commit (string-append "v" version)))) |
| 2783 | (file-name (git-file-name name version)) |
| 2784 | (sha256 |
| 2785 | (base32 "04h9lbcaf4xizmcyq10lgr1gsxlgb8w4zcbf9vnzp37bn344cvig")))) |
| 2786 | (build-system pyproject-build-system) |
| 2787 | (arguments |
| 2788 | (list |
| 2789 | ;; tests: 1215 passed, 113 skipped, 1657 warnings |
| 2790 | #:phases |
| 2791 | #~(modify-phases %standard-phases |
| 2792 | (add-before 'check 'remove-local-thinc |
| 2793 | (lambda _ |
| 2794 | (copy-recursively "thinc/tests" "tests") |
| 2795 | ;; This would otherwise interfere with finding the installed |
| 2796 | ;; thinc when running tests. |
| 2797 | (delete-file-recursively "thinc")))))) |
| 2798 | (propagated-inputs |
| 2799 | (list python-blis |
| 2800 | python-catalogue |
| 2801 | python-confection |
| 2802 | python-cymem |
| 2803 | python-murmurhash |
| 2804 | python-numpy |
| 2805 | python-packaging |
| 2806 | python-preshed |
| 2807 | python-pydantic |
| 2808 | python-srsly |
| 2809 | python-wasabi)) |
| 2810 | (native-inputs |
| 2811 | (list python-coverage |
| 2812 | python-cython |
| 2813 | python-mock |
| 2814 | python-pytest |
| 2815 | python-setuptools)) |
| 2816 | (home-page "https://github.com/explosion/thinc") |
| 2817 | (synopsis "Functional take on deep learning") |
| 2818 | (description |
| 2819 | "This package provides a functional take on deep learning, compatible |
| 2820 | with your favorite libraries.") |
| 2821 | (license license:expat))) |
| 2822 | |
| 2823 | (define-public python-threadpoolctl |
| 2824 | (package |
| 2825 | (name "python-threadpoolctl") |
| 2826 | (version "3.6.0") |
| 2827 | (source |
| 2828 | (origin |
| 2829 | (method git-fetch) |
| 2830 | (uri (git-reference |
| 2831 | (url "https://github.com/joblib/threadpoolctl") |
| 2832 | (commit version))) |
| 2833 | (file-name (git-file-name name version)) |
| 2834 | (sha256 |
| 2835 | (base32 |
| 2836 | "1n5gkcw619d9qgnnk46v2v77safj0z8hngb8i75dawq7lan89068")))) |
| 2837 | (build-system pyproject-build-system) |
| 2838 | (native-inputs |
| 2839 | (list python-flit-core python-pytest)) |
| 2840 | (home-page "https://github.com/joblib/threadpoolctl") |
| 2841 | (synopsis "Python helpers for common threading libraries") |
| 2842 | (description "Thread-pool Controls provides Python helpers to limit the |
| 2843 | number of threads used in the threadpool-backed of common native libraries used |
| 2844 | for scientific computing and data science (e.g. BLAS and OpenMP).") |
| 2845 | (license license:bsd-3))) |
| 2846 | |
| 2847 | (define-public python-hdbscan |
| 2848 | (package |
| 2849 | (name "python-hdbscan") |
| 2850 | (version "0.8.41") |
| 2851 | (source |
| 2852 | (origin |
| 2853 | (method url-fetch) |
| 2854 | (uri (pypi-uri "hdbscan" version)) |
| 2855 | (sha256 |
| 2856 | (base32 "0ggzkm153cmgdq88jpdsvcmsipdicri24b957wbz47xjbcz847p4")))) |
| 2857 | (build-system pyproject-build-system) |
| 2858 | (arguments |
| 2859 | (list |
| 2860 | #:test-flags |
| 2861 | ;; XXX: Test files which fail during test collection. |
| 2862 | #~(list "--ignore=hdbscan/tests/test_branches.py" |
| 2863 | "--ignore=hdbscan/tests/test_hdbscan.py" |
| 2864 | "--ignore=hdbscan/tests/test_rsl.py") |
| 2865 | #:phases |
| 2866 | #~(modify-phases %standard-phases |
| 2867 | (add-before 'check 'pre-check |
| 2868 | (lambda _ |
| 2869 | (delete-file "hdbscan/tests/__init__.py"))) |
| 2870 | (add-before 'check 'build-extensions |
| 2871 | (lambda _ |
| 2872 | (invoke "python" "setup.py" "build_ext" "--inplace")))))) |
| 2873 | (native-inputs |
| 2874 | (list python-cython |
| 2875 | python-numpy |
| 2876 | python-pytest |
| 2877 | python-setuptools)) |
| 2878 | (propagated-inputs |
| 2879 | (list python-joblib |
| 2880 | python-numpy |
| 2881 | python-scikit-learn |
| 2882 | python-scipy)) |
| 2883 | (home-page "https://github.com/scikit-learn-contrib/hdbscan") |
| 2884 | (synopsis "High performance implementation of HDBSCAN clustering") |
| 2885 | (description "HDBSCAN - Hierarchical Density-Based Spatial Clustering of |
| 2886 | Applications with Noise. Performs DBSCAN over varying epsilon values and |
| 2887 | integrates the result to find a clustering that gives the best stability over |
| 2888 | epsilon. This allows HDBSCAN to find clusters of varying densities (unlike |
| 2889 | DBSCAN), and be more robust to parameter selection. HDBSCAN is ideal for |
| 2890 | exploratory data analysis; it's a fast and robust algorithm that you can trust |
| 2891 | to return meaningful clusters (if there are any).") |
| 2892 | (license license:bsd-3))) |
| 2893 | |
| 2894 | (define-public python-pynndescent |
| 2895 | (package |
| 2896 | (name "python-pynndescent") |
| 2897 | (version "0.6.0") |
| 2898 | (source |
| 2899 | (origin |
| 2900 | (method url-fetch) |
| 2901 | (uri (pypi-uri "pynndescent" version)) |
| 2902 | (sha256 |
| 2903 | (base32 "1x94mjdf62f2yqql50qn0qjh4bkhwdvx7xx9aph421s0bgxy1zbz")))) |
| 2904 | (build-system pyproject-build-system) |
| 2905 | (arguments |
| 2906 | (list |
| 2907 | #:test-flags |
| 2908 | ;; Error from ScikitLearn: The 'metric' parameter of pairwise_distances |
| 2909 | ;; must be a str among <...>. Fixed in upcomming release of pynndescent. |
| 2910 | #~(list (string-append "--deselect=pynndescent/tests/test_distances.py" |
| 2911 | "::test_binary_check[sokalmichener]") |
| 2912 | (string-append "--deselect=ynndescent/tests/test_distances.py" |
| 2913 | "::test_sparse_binary_check[sokalmichener]") |
| 2914 | (string-append "--deselect=pynndescent/tests/test_distances.py" |
| 2915 | "::test_sparse_binary_check[sokalmichener]")))) |
| 2916 | (native-inputs (list python-pytest python-setuptools)) |
| 2917 | (propagated-inputs (list python-joblib |
| 2918 | python-llvmlite |
| 2919 | python-numba |
| 2920 | python-scikit-learn |
| 2921 | python-scipy)) |
| 2922 | (home-page "https://github.com/lmcinnes/pynndescent") |
| 2923 | (synopsis "Nearest neighbor descent for approximate nearest neighbors") |
| 2924 | (description |
| 2925 | "PyNNDescent provides a Python implementation of Nearest Neighbor Descent |
| 2926 | for k-neighbor-graph construction and approximate nearest neighbor search.") |
| 2927 | (license license:bsd-2))) |
| 2928 | |
| 2929 | (define-public python-opentsne |
| 2930 | (package |
| 2931 | (name "python-opentsne") |
| 2932 | (version "1.0.2") |
| 2933 | (source |
| 2934 | (origin |
| 2935 | (method git-fetch) ;no tests in PyPI release |
| 2936 | (uri (git-reference |
| 2937 | (url "https://github.com/pavlin-policar/openTSNE") |
| 2938 | (commit (string-append "v" version)))) |
| 2939 | (file-name (git-file-name name version)) |
| 2940 | (sha256 |
| 2941 | (base32 "1jn15h8b53z6ah47y51y8y4m0mk20xr6l2kd27ffr0hxswbifmkv")))) |
| 2942 | (build-system pyproject-build-system) |
| 2943 | (arguments |
| 2944 | (list |
| 2945 | #:phases |
| 2946 | #~(modify-phases %standard-phases |
| 2947 | ;; Benchmarks require the 'macosko2015' data files. |
| 2948 | (add-after 'unpack 'delete-benchmark |
| 2949 | (lambda _ |
| 2950 | (delete-file-recursively "benchmarks"))) |
| 2951 | (add-before 'check 'preparations |
| 2952 | (lambda _ |
| 2953 | ;; Numba needs a writable dir to cache functions. |
| 2954 | (setenv "NUMBA_CACHE_DIR" "/tmp") |
| 2955 | ;; Rebuild extensions to run tests. |
| 2956 | (invoke "python" "setup.py" "build_ext" "--inplace")))))) |
| 2957 | (native-inputs (list python-cython python-pytest python-setuptools |
| 2958 | python-wheel)) |
| 2959 | (inputs (list fftw)) |
| 2960 | (propagated-inputs (list python-numpy python-pynndescent |
| 2961 | python-scikit-learn python-scipy)) |
| 2962 | (home-page "https://github.com/pavlin-policar/openTSNE") |
| 2963 | (synopsis "Extensible, parallel implementations of t-SNE") |
| 2964 | (description |
| 2965 | "This is a modular Python implementation of t-Distributed Stochastic |
| 2966 | Neighbor Embedding (t-SNE), a popular dimensionality-reduction algorithm for |
| 2967 | visualizing high-dimensional data sets.") |
| 2968 | (license license:bsd-3))) |
| 2969 | |
| 2970 | ;; XXX: Potentially it is an abandonware, no releases or any updates since |
| 2971 | ;; 2021, consider to remove when become incompatible with propagated inputs. |
| 2972 | (define-public python-scikit-rebate |
| 2973 | (package |
| 2974 | (name "python-scikit-rebate") |
| 2975 | (version "0.62") |
| 2976 | (source |
| 2977 | (origin |
| 2978 | (method url-fetch) |
| 2979 | (uri (pypi-uri "skrebate" version)) |
| 2980 | (sha256 |
| 2981 | (base32 "0n55ghvnv7rxqa5agq6a4892ad0ghha165b0g4ghwr9gqm6ss3dj")))) |
| 2982 | (build-system pyproject-build-system) |
| 2983 | ;; There are no tests on PyPI, tests in git repo require an old release of |
| 2984 | ;; python-scikit-learn |
| 2985 | (arguments '(#:tests? #f)) |
| 2986 | (native-inputs |
| 2987 | (list python-setuptools)) |
| 2988 | (propagated-inputs |
| 2989 | (list python-numpy python-scipy python-scikit-learn python-joblib)) |
| 2990 | (home-page "https://epistasislab.github.io/scikit-rebate/") |
| 2991 | (synopsis "Relief-based feature selection algorithms for Python") |
| 2992 | (description "Scikit-rebate is a scikit-learn-compatible Python |
| 2993 | implementation of ReBATE, a suite of Relief-based feature selection algorithms |
| 2994 | for Machine Learning. These algorithms excel at identifying features that are |
| 2995 | predictive of the outcome in supervised learning problems, and are especially |
| 2996 | good at identifying feature interactions that are normally overlooked by |
| 2997 | standard feature selection algorithms.") |
| 2998 | (license license:expat))) |
| 2999 | |
| 3000 | (define-public python-cleanlab |
| 3001 | (package |
| 3002 | (name "python-cleanlab") |
| 3003 | (version "2.9.0") |
| 3004 | (source (origin |
| 3005 | (method git-fetch) |
| 3006 | (uri (git-reference |
| 3007 | (url "https://github.com/cleanlab/cleanlab") |
| 3008 | (commit (string-append "v" version)))) |
| 3009 | (file-name (git-file-name name version)) |
| 3010 | (sha256 |
| 3011 | (base32 |
| 3012 | "1fks57i3a297fcm1k4bs48x9cc7rs0jp4ij92r42pd1n0x60jznh")))) |
| 3013 | (build-system pyproject-build-system) |
| 3014 | (arguments |
| 3015 | (list |
| 3016 | ;; tests: 616 passed, 3 deselected, 99 warnings |
| 3017 | #:test-flags |
| 3018 | ;; This test fails because the newer version of scikit learn returns one |
| 3019 | ;; more classification result than expected. This should be harmless. |
| 3020 | #~(list "-k" "not test_aux_inputs" |
| 3021 | ;; Requires Tensorflow |
| 3022 | "--ignore=tests/test_frameworks.py" |
| 3023 | ;; These need datasets, which needs jax, so it could only live in |
| 3024 | ;; the guix-science channel. |
| 3025 | "--ignore-glob=tests/datalab/**" |
| 3026 | ;; Tries to download datasets from the internet at runtime. |
| 3027 | "--ignore=tests/test_dataset.py" |
| 3028 | ;; Test requiring not packaged dataset. |
| 3029 | "--ignore=tests/spurious_correlation/test_correlation_visualizer.py" |
| 3030 | "--ignore=tests/spurious_correlation/test_spurious_correlation.py" |
| 3031 | ;; AssertionError: assert 'Annotators [1] did not label any |
| 3032 | ;; examples.' in 'labels_multiannotator cannot have columns with |
| 3033 | ;; all NaN, each annotator must annotator at least one example. |
| 3034 | "--deselect=tests/test_multiannotator.py::test_label_quality_scores_multiannotator" |
| 3035 | ;; These tests were not updated to the last numpy update. |
| 3036 | ;; Try enabling them on the next update. |
| 3037 | (string-append "--deselect=tests/test_object_detection.py::" |
| 3038 | "test_return_issues_ranked_by_scores") |
| 3039 | (string-append "--deselect=tests/test_object_detection.py::" |
| 3040 | "test_bad_input_find_label_issues_internal")) |
| 3041 | #:phases |
| 3042 | #~(modify-phases %standard-phases |
| 3043 | (add-after 'unpack 'remove-datasets |
| 3044 | (lambda _ |
| 3045 | (delete-file "tests/datalab/conftest.py")))))) |
| 3046 | (propagated-inputs |
| 3047 | (list python-numpy |
| 3048 | python-pandas |
| 3049 | python-scikit-learn |
| 3050 | python-termcolor |
| 3051 | python-tqdm)) |
| 3052 | (native-inputs |
| 3053 | (list ;; python-dataset ; https://github.com/huggingface/datasets |
| 3054 | python-matplotlib |
| 3055 | python-pytest |
| 3056 | python-pytorch |
| 3057 | python-setuptools |
| 3058 | python-torchvision |
| 3059 | python-typing-extensions)) |
| 3060 | (home-page "https://cleanlab.ai") |
| 3061 | (synopsis "Automatically find and fix dataset issues") |
| 3062 | (description |
| 3063 | "cleanlab automatically finds and fixes errors in any ML dataset. This |
| 3064 | data-centric AI package facilitates machine learning with messy, real-world |
| 3065 | data by providing clean labels during training.") |
| 3066 | (license license:agpl3+))) |
| 3067 | |
| 3068 | (define-public python-cma |
| 3069 | (package |
| 3070 | (name "python-cma") |
| 3071 | (version "4.4.4") |
| 3072 | (source (origin |
| 3073 | (method url-fetch) |
| 3074 | (uri (pypi-uri "cma" version)) |
| 3075 | (sha256 |
| 3076 | (base32 |
| 3077 | "0w41k8dk8g1ngx797dzag5xcwws77sfnfriim874rq6wnmadcav3")))) |
| 3078 | (build-system pyproject-build-system) |
| 3079 | (arguments |
| 3080 | (list #:test-backend #~'custom |
| 3081 | #:test-flags #~'("src/cma/test.py"))) |
| 3082 | (native-inputs (list python-pytest python-setuptools)) |
| 3083 | (propagated-inputs (list python-numpy)) |
| 3084 | (home-page "https://github.com/CMA-ES/pycma") |
| 3085 | (synopsis "Python implementation of CMA-ES") |
| 3086 | (description "This package provides a Python implementation of the |
| 3087 | @acronym{CMA-ES, Covariance Matrix Adaptation Evolution Strategy} algorithm |
| 3088 | and a few related numerical optimization tools.") |
| 3089 | (license license:bsd-3))) |
| 3090 | |
| 3091 | (define-public python-cmaes |
| 3092 | (package |
| 3093 | (name "python-cmaes") |
| 3094 | (version "0.13.0") |
| 3095 | (source |
| 3096 | (origin |
| 3097 | (method git-fetch) ;no tests in PyPI |
| 3098 | (uri (git-reference |
| 3099 | (url "https://github.com/CyberAgentAILab/cmaes") |
| 3100 | (commit (string-append "v" version)))) |
| 3101 | (sha256 |
| 3102 | (base32 "1fqy7829isg3j00849ifya4ncf9qpm0b3jhw47ibjn4jv0kn4zmh")) |
| 3103 | (file-name (git-file-name name version)))) |
| 3104 | (build-system pyproject-build-system) |
| 3105 | (native-inputs (list python-pytest python-setuptools)) |
| 3106 | (propagated-inputs (list python-numpy)) |
| 3107 | (home-page "https://github.com/CyberAgentAILab/cmaes") |
| 3108 | (synopsis "CMA-ES implementation for Python") |
| 3109 | (description "This package provides provides an implementation of the |
| 3110 | Covariance Matrix Adaptation Evolution Strategy (CMA-ES) for Python.") |
| 3111 | (license license:expat))) |
| 3112 | |
| 3113 | (define-public python-autograd |
| 3114 | (package |
| 3115 | (name "python-autograd") |
| 3116 | (version "1.8.0") |
| 3117 | (source |
| 3118 | (origin |
| 3119 | (method git-fetch) |
| 3120 | (uri (git-reference |
| 3121 | (url "https://github.com/HIPS/autograd") |
| 3122 | (commit (string-append "v" version)))) |
| 3123 | (sha256 |
| 3124 | (base32 "054pkhzz0h9p1jzva8774wb9dj7rvax4rcpr8ava971kbimdr2lk")) |
| 3125 | (file-name (git-file-name name version)))) |
| 3126 | (build-system pyproject-build-system) |
| 3127 | (native-inputs |
| 3128 | (list python-hatchling |
| 3129 | python-pytest |
| 3130 | python-pytest-cov ; This package has a custom way of running tests. |
| 3131 | python-pytest-xdist)) |
| 3132 | (propagated-inputs |
| 3133 | (list python-future |
| 3134 | python-numpy)) |
| 3135 | (home-page "https://github.com/HIPS/autograd") |
| 3136 | (synopsis "Efficiently computes derivatives of NumPy code") |
| 3137 | (description |
| 3138 | "Autograd can automatically differentiate native Python and NumPy code. |
| 3139 | It can handle a large subset of Python's features, including loops, ifs, |
| 3140 | recursion and closures, and it can even take derivatives of derivatives of |
| 3141 | derivatives. It supports reverse-mode differentiation |
| 3142 | (a.k.a. backpropagation), which means it can efficiently take gradients of |
| 3143 | scalar-valued functions with respect to array-valued arguments, as well as |
| 3144 | forward-mode differentiation, and the two can be composed arbitrarily. The |
| 3145 | main intended application of Autograd is gradient-based optimization.") |
| 3146 | (license license:expat))) |
| 3147 | |
| 3148 | (define-public python-torchdiffeq |
| 3149 | ;; There are neither releases nor tags. |
| 3150 | (let ((commit "657943acefa826ef04c025ebeb1ff5e9d60dc268") |
| 3151 | (revision "1")) |
| 3152 | (package |
| 3153 | (name "python-torchdiffeq") |
| 3154 | (version (git-version "0.2.5" revision commit)) |
| 3155 | (source |
| 3156 | (origin |
| 3157 | (method git-fetch) |
| 3158 | (uri (git-reference |
| 3159 | (url "https://github.com/rtqichen/torchdiffeq") |
| 3160 | (commit commit))) |
| 3161 | (file-name (git-file-name name version)) |
| 3162 | (sha256 |
| 3163 | (base32 "15ccm7m0k3jjvx0sxz4s453pb0sxlbg5569dqlvj6ql74jl356x0")))) |
| 3164 | (build-system pyproject-build-system) |
| 3165 | (arguments |
| 3166 | (list |
| 3167 | #:test-flags |
| 3168 | #~(list |
| 3169 | ;; This starts failing with [email protected], it looks like a regression |
| 3170 | ;; in the fortran to C rewrite of LSODA. |
| 3171 | "-k" "not test_min_max_step" |
| 3172 | "tests/run_all.py"))) |
| 3173 | (propagated-inputs (list python-numpy python-scipy python-pytorch)) |
| 3174 | (native-inputs (list python-pytest python-setuptools)) |
| 3175 | (home-page "https://github.com/rtqichen/torchdiffeq") |
| 3176 | (synopsis "ODE solvers and adjoint sensitivity analysis in PyTorch") |
| 3177 | (description |
| 3178 | "This tool provides ordinary differential equation solvers implemented |
| 3179 | in PyTorch. Backpropagation through ODE solutions is supported using the |
| 3180 | adjoint method for constant memory cost.") |
| 3181 | (license license:expat)))) |
| 3182 | |
| 3183 | (define-public lightgbm |
| 3184 | (package |
| 3185 | (name "lightgbm") |
| 3186 | (version "2.0.12") |
| 3187 | (source |
| 3188 | (origin |
| 3189 | (method git-fetch) |
| 3190 | (uri (git-reference |
| 3191 | (url "https://github.com/Microsoft/LightGBM") |
| 3192 | (commit (string-append "v" version)))) |
| 3193 | (sha256 |
| 3194 | (base32 "0jlvyn7k81dzrh9ij3zw576wbgiwmmr26rzpdxjn1dbpc3njpvzi")) |
| 3195 | (file-name (git-file-name name version)))) |
| 3196 | (build-system cmake-build-system) |
| 3197 | (arguments |
| 3198 | (list |
| 3199 | #:configure-flags |
| 3200 | #~(list "-DUSE_MPI=ON") |
| 3201 | #:phases |
| 3202 | #~(modify-phases %standard-phases |
| 3203 | (replace 'check |
| 3204 | (lambda* (#:key tests? #:allow-other-keys) |
| 3205 | (when tests? |
| 3206 | (with-directory-excursion "../source" |
| 3207 | (invoke "pytest" "tests/c_api_test/test_.py")))))))) |
| 3208 | (native-inputs (list python python-pytest)) |
| 3209 | (inputs (list openmpi)) |
| 3210 | (propagated-inputs (list python-numpy python-scipy)) |
| 3211 | (home-page "https://github.com/Microsoft/LightGBM") |
| 3212 | (synopsis "Gradient boosting framework based on decision tree algorithms") |
| 3213 | (description |
| 3214 | "LightGBM is a gradient boosting framework that uses tree based learning |
| 3215 | algorithms. It is designed to be distributed and efficient with the following |
| 3216 | advantages: |
| 3217 | |
| 3218 | @itemize |
| 3219 | @item Faster training speed and higher efficiency |
| 3220 | @item Lower memory usage |
| 3221 | @item Better accuracy |
| 3222 | @item Parallel and GPU learning supported (not enabled in this package) |
| 3223 | @item Capable of handling large-scale data |
| 3224 | @end itemize |
| 3225 | ") |
| 3226 | (license license:expat))) |
| 3227 | |
| 3228 | (define-public python-lightgbm |
| 3229 | (package/inherit lightgbm |
| 3230 | (name "python-lightgbm") |
| 3231 | (build-system pyproject-build-system) |
| 3232 | (arguments |
| 3233 | (list |
| 3234 | ;; XXX: Enable when lightgbm is updated. |
| 3235 | #:tests? #f |
| 3236 | #:phases |
| 3237 | #~(modify-phases %standard-phases |
| 3238 | (add-after 'unpack 'chdir |
| 3239 | (lambda _ |
| 3240 | (chdir "python-package")))))) |
| 3241 | (native-inputs |
| 3242 | (list cmake-minimal |
| 3243 | python-pytest |
| 3244 | python-setuptools)) |
| 3245 | (propagated-inputs |
| 3246 | (list python-numpy |
| 3247 | python-scipy |
| 3248 | python-scikit-learn)) |
| 3249 | (synopsis "LightGBM Python Package"))) |
| 3250 | |
| 3251 | (define-public vowpal-wabbit |
| 3252 | ;; Language bindings not included. |
| 3253 | (package |
| 3254 | (name "vowpal-wabbit") |
| 3255 | (version "8.5.0") |
| 3256 | (source (origin |
| 3257 | (method git-fetch) |
| 3258 | (uri (git-reference |
| 3259 | (url "https://github.com/JohnLangford/vowpal_wabbit") |
| 3260 | (commit version))) |
| 3261 | (sha256 |
| 3262 | (base32 |
| 3263 | "04bwzk6ifgnz3fmzid8b7avxf9n5pnx9xcjm61nkjng1vv0bpj8x")) |
| 3264 | (file-name (git-file-name name version)))) |
| 3265 | (build-system gnu-build-system) |
| 3266 | (arguments |
| 3267 | (list |
| 3268 | #:configure-flags |
| 3269 | #~(list (string-append "--with-boost=" #$(this-package-input "boost"))) |
| 3270 | #:phases |
| 3271 | #~(modify-phases %standard-phases |
| 3272 | (add-after 'unpack 'make-files-writable |
| 3273 | (lambda _ |
| 3274 | (for-each make-file-writable (find-files "." ".*")))) |
| 3275 | (add-after 'install 'install-more-headers |
| 3276 | (lambda _ |
| 3277 | (for-each |
| 3278 | (lambda (file) |
| 3279 | (install-file |
| 3280 | file (string-append #$output "/include/vowpalwabbit"))) |
| 3281 | (find-files "vowpalwabbit" "\\.h$"))))))) |
| 3282 | (inputs |
| 3283 | (list boost zlib)) |
| 3284 | (home-page "https://vowpalwabbit.org") |
| 3285 | (synopsis "Fast machine learning library for online learning") |
| 3286 | (description "Vowpal Wabbit is a machine learning system with techniques |
| 3287 | such as online, hashing, allreduce, reductions, learning2search, active, and |
| 3288 | interactive learning.") |
| 3289 | (license license:bsd-3))) |
| 3290 | |
| 3291 | (define-public python-hyperopt |
| 3292 | (package |
| 3293 | (name "python-hyperopt") |
| 3294 | (version "0.3.0") |
| 3295 | (source |
| 3296 | (origin |
| 3297 | (method git-fetch) |
| 3298 | (uri (git-reference |
| 3299 | (url "https://github.com/hyperopt/hyperopt") |
| 3300 | (commit (string-append "v" version)))) |
| 3301 | (file-name (git-file-name name version)) |
| 3302 | (sha256 |
| 3303 | (base32 "09qy7v01dy2kk9wpd50ras5m9qzysg6h23lm3yl5ylr8hk7hi8wj")))) |
| 3304 | (build-system pyproject-build-system) |
| 3305 | (arguments |
| 3306 | (list |
| 3307 | ;; tests: 225 passed, 2 skipped, 1 deselected, 2 warnings |
| 3308 | #:test-flags |
| 3309 | #~(list "-m" "not slow" |
| 3310 | "--ignore=hyperopt/tests/integration/" |
| 3311 | ;; XXX: Segfaulted test. |
| 3312 | (string-append "--deselect=hyperopt/tests/unit/" |
| 3313 | "test_atpe_basic.py::test_run_basic_search")))) |
| 3314 | (native-inputs |
| 3315 | (list python-hatchling |
| 3316 | python-hatch-vcs |
| 3317 | python-pytest)) |
| 3318 | (propagated-inputs |
| 3319 | (list python-cloudpickle |
| 3320 | python-networkx |
| 3321 | python-numpy |
| 3322 | python-scipy |
| 3323 | python-tqdm |
| 3324 | ;; [optional] |
| 3325 | python-lightgbm |
| 3326 | python-py4j |
| 3327 | python-pymongo |
| 3328 | ;; python-pyspark ;not packaged yet in Guix |
| 3329 | python-scikit-learn)) |
| 3330 | (home-page "https://hyperopt.github.io/hyperopt/") |
| 3331 | (synopsis "Library for hyperparameter optimization") |
| 3332 | (description "Hyperopt is a Python library for serial and parallel |
| 3333 | optimization over awkward search spaces, which may include real-valued, |
| 3334 | discrete, and conditional dimensions.") |
| 3335 | (license license:bsd-3))) |
| 3336 | |
| 3337 | (define-public python-deepxde |
| 3338 | (package |
| 3339 | (name "python-deepxde") |
| 3340 | (version "1.15.0") |
| 3341 | (source (origin |
| 3342 | (method url-fetch) |
| 3343 | (uri (pypi-uri "deepxde" version)) |
| 3344 | (sha256 |
| 3345 | (base32 |
| 3346 | "1m21hkcj386p0bwf19cr6r2kkv0c054yv1ipsl1rqav13g88l3ic")))) |
| 3347 | (build-system pyproject-build-system) |
| 3348 | (arguments |
| 3349 | (list #:tests? #f ; there are no tests |
| 3350 | #:phases #~(modify-phases %standard-phases |
| 3351 | (add-before 'sanity-check 'writable-home |
| 3352 | ;; sanity-check writes ~/.deepxde/config.json to set |
| 3353 | ;; the default backend. |
| 3354 | (lambda _ |
| 3355 | (setenv "HOME" "/tmp")))))) |
| 3356 | ;; DeepXDE supported backends are TensorFlow (v1 and v2), PyTorch, JAX and |
| 3357 | ;; PaddlePaddle. We test with PyTorch because we have it up to date. |
| 3358 | (native-inputs (list python-pytorch python-setuptools |
| 3359 | python-setuptools-scm)) |
| 3360 | (propagated-inputs (list python-matplotlib python-numpy |
| 3361 | python-scikit-learn python-scikit-optimize |
| 3362 | python-scipy)) |
| 3363 | (home-page "https://deepxde.readthedocs.io/en/latest/") |
| 3364 | (synopsis "Library for scientific machine learning") |
| 3365 | (description "DeepXDE is a library for scientific machine learning and |
| 3366 | physics-informed learning. It includes implementations for the PINN |
| 3367 | (physics-informed neural networks), DeepONet (deep operator network) and |
| 3368 | MFNN (multifidelity neural network) algorithms.") |
| 3369 | (license license:lgpl2.1+))) |
| 3370 | |
| 3371 | (define-public python-jaxtyping |
| 3372 | (package |
| 3373 | (name "python-jaxtyping") |
| 3374 | (version "0.3.3") |
| 3375 | (source (origin |
| 3376 | (method url-fetch) |
| 3377 | (uri (pypi-uri "jaxtyping" version)) |
| 3378 | (sha256 |
| 3379 | (base32 |
| 3380 | "11pajw47mj0px787m6f7vwfsi6ajawm9h7fszm3rpbd2dg8wy0w0")))) |
| 3381 | (build-system pyproject-build-system) |
| 3382 | ;; Tests require JAX, but JAX can't be packaged because it uses the Bazel |
| 3383 | ;; build system. |
| 3384 | (arguments (list #:tests? #f)) |
| 3385 | (native-inputs (list python-hatchling)) |
| 3386 | (propagated-inputs (list python-numpy python-typeguard |
| 3387 | python-typing-extensions python-wadler-lindig)) |
| 3388 | (home-page "https://github.com/google/jaxtyping") |
| 3389 | (synopsis |
| 3390 | "Type annotations and runtime checking for JAX arrays and others") |
| 3391 | (description "@code{jaxtyping} provides type annotations and runtime |
| 3392 | checking for shape and dtype of JAX arrays, PyTorch, NumPy, TensorFlow, and |
| 3393 | PyTrees.") |
| 3394 | (license license:expat))) |
| 3395 | |
| 3396 | ;; There have been no proper releases yet. |
| 3397 | (define-public kaldi |
| 3398 | (let ((commit "01aadd7c19372e3eacadec88caabd86162f33d69") |
| 3399 | (revision "2")) |
| 3400 | (package |
| 3401 | (name "kaldi") |
| 3402 | (version (git-version "0" revision commit)) |
| 3403 | (source (origin |
| 3404 | (method git-fetch) |
| 3405 | (uri (git-reference |
| 3406 | (url "https://github.com/kaldi-asr/kaldi") |
| 3407 | (commit commit))) |
| 3408 | (file-name (git-file-name name version)) |
| 3409 | (sha256 |
| 3410 | (base32 |
| 3411 | "08l31g256wl81fsrm1dvi0d2rq2vk5zq7ihbbiw7hp51iwg2miif")) |
| 3412 | (patches |
| 3413 | (search-patches "kaldi-openblas-0.3.29-compatibility.patch" |
| 3414 | "kaldi-ignore-failing-test.patch")))) |
| 3415 | (build-system gnu-build-system) |
| 3416 | (arguments |
| 3417 | (list |
| 3418 | #:test-target "test" |
| 3419 | #:configure-flags |
| 3420 | (let ((fst (this-package-input "openfst"))) |
| 3421 | #~(list "--use-cuda=no" |
| 3422 | "--shared" |
| 3423 | (string-append "--openblas-root=" |
| 3424 | #$(this-package-input "openblas")) |
| 3425 | (string-append "--fst-root=" #$fst) |
| 3426 | (string-append "--fst-version=" #$(package-version fst)))) |
| 3427 | #:phases |
| 3428 | #~(modify-phases %standard-phases |
| 3429 | (add-after 'unpack 'chdir |
| 3430 | (lambda _ (chdir "src"))) |
| 3431 | (replace 'configure |
| 3432 | (lambda* (#:key build system inputs configure-flags |
| 3433 | #:allow-other-keys) |
| 3434 | (when (not (or (string-prefix? "x86_64" system) |
| 3435 | (string-prefix? "i686" system))) |
| 3436 | (substitute* "makefiles/linux_openblas.mk" |
| 3437 | (("-msse -msse2") ""))) |
| 3438 | (substitute* "makefiles/default_rules.mk" |
| 3439 | (("/bin/bash") (which "bash"))) |
| 3440 | (substitute* "Makefile" |
| 3441 | (("ext_depend: check_portaudio") |
| 3442 | "ext_depend:")) |
| 3443 | (substitute* '("online/Makefile" |
| 3444 | "onlinebin/Makefile" |
| 3445 | "gst-plugin/Makefile") |
| 3446 | (("../../tools/portaudio/install") |
| 3447 | (assoc-ref inputs "portaudio"))) |
| 3448 | (substitute* "matrix/Makefile" ;temporary test bypass |
| 3449 | (("matrix-lib-test sparse-matrix-test") "")) |
| 3450 | |
| 3451 | ;; This `configure' script doesn't support variables passed as |
| 3452 | ;; arguments, nor does it support "prefix". |
| 3453 | (substitute* "configure" |
| 3454 | (("check_for_slow_expf;") "") |
| 3455 | ;; This affects the RPATH and also serves as the installation |
| 3456 | ;; directory. |
| 3457 | (("KALDILIBDIR=`pwd`/lib") |
| 3458 | (string-append "KALDILIBDIR=" #$output "/lib"))) |
| 3459 | (mkdir-p #$output) ; must exist |
| 3460 | (setenv "CONFIG_SHELL" (which "bash")) |
| 3461 | (apply invoke "./configure" configure-flags))) |
| 3462 | (add-after 'build 'build-ext-and-gstreamer-plugin |
| 3463 | (lambda _ |
| 3464 | (invoke "make" "-C" "online" "depend") |
| 3465 | (invoke "make" "-C" "online") |
| 3466 | (invoke "make" "-C" "onlinebin" "depend") |
| 3467 | (invoke "make" "-C" "onlinebin") |
| 3468 | (invoke "make" "-C" "gst-plugin" "depend") |
| 3469 | (invoke "make" "-C" "gst-plugin"))) |
| 3470 | ;; TODO: also install the executables. |
| 3471 | (replace 'install |
| 3472 | (lambda _ |
| 3473 | (let* ((inc (string-append #$output "/include")) |
| 3474 | (lib (string-append #$output "/lib"))) |
| 3475 | (mkdir-p lib) |
| 3476 | ;; The build phase installed symlinks to the actual |
| 3477 | ;; libraries. Install the actual targets. |
| 3478 | (for-each (lambda (file) |
| 3479 | (let ((target (readlink file))) |
| 3480 | (delete-file file) |
| 3481 | (install-file target lib))) |
| 3482 | (find-files lib "\\.so")) |
| 3483 | ;; Install headers |
| 3484 | (for-each (lambda (file) |
| 3485 | (let ((target-dir (string-append inc "/" (dirname file)))) |
| 3486 | (install-file file target-dir))) |
| 3487 | (find-files "." "\\.h")) |
| 3488 | (install-file "gst-plugin/libgstonlinegmmdecodefaster.so" |
| 3489 | (string-append lib "/gstreamer-1.0")))))))) |
| 3490 | (inputs |
| 3491 | (list alsa-lib |
| 3492 | `(,gfortran "lib") |
| 3493 | glib |
| 3494 | gstreamer |
| 3495 | jack-1 |
| 3496 | openblas |
| 3497 | openfst |
| 3498 | portaudio |
| 3499 | python)) |
| 3500 | (native-inputs |
| 3501 | (list `(,glib "bin") ; glib-genmarshal |
| 3502 | grep |
| 3503 | sed |
| 3504 | pkg-config |
| 3505 | which)) |
| 3506 | (home-page "https://kaldi-asr.org/") |
| 3507 | (synopsis "Speech recognition toolkit") |
| 3508 | (description "Kaldi is an extensible toolkit for speech recognition |
| 3509 | written in C++.") |
| 3510 | (license license:asl2.0)))) |
| 3511 | |
| 3512 | (define kaldi-for-vosk |
| 3513 | ;; Commit of branch "vosk" |
| 3514 | (let* ((commit "bc5baf14231660bd50b7d05788865b4ac6c34481") |
| 3515 | (revision "1") |
| 3516 | (openfst openfst-for-vosk)) |
| 3517 | (package |
| 3518 | (inherit kaldi) |
| 3519 | (name "kaldi") |
| 3520 | (version (git-version "0" revision commit)) |
| 3521 | (source |
| 3522 | (origin |
| 3523 | (method git-fetch) |
| 3524 | (uri (git-reference |
| 3525 | (url "https://github.com/alphacep/kaldi") |
| 3526 | (commit commit))) |
| 3527 | (file-name (git-file-name name version)) |
| 3528 | (sha256 |
| 3529 | (base32 "1y3d6918srr7cn5r72v5wvbdwz9p9j2bjw1x78sfis2r2k60lllw")) |
| 3530 | (patches |
| 3531 | (search-patches "kaldi-openblas-0.3.29-compatibility.patch")))) |
| 3532 | (inputs |
| 3533 | (list alsa-lib |
| 3534 | glib |
| 3535 | gstreamer |
| 3536 | jack-2 |
| 3537 | openblas |
| 3538 | openfst |
| 3539 | portaudio |
| 3540 | python)) |
| 3541 | (arguments |
| 3542 | (list |
| 3543 | #:test-target "test" |
| 3544 | #:make-flags '(list "online2" "lm" "rnnlm") |
| 3545 | #:phases |
| 3546 | #~(modify-phases %standard-phases |
| 3547 | (add-after 'unpack 'chdir |
| 3548 | (lambda _ (chdir "src"))) |
| 3549 | (replace 'configure |
| 3550 | (lambda _ |
| 3551 | (let ((portaudio #$(this-package-input "portaudio")) |
| 3552 | (openfst #$(this-package-input "openfst")) |
| 3553 | (openblas #$(this-package-input "openblas"))) |
| 3554 | #$@(if (target-x86?) |
| 3555 | '() |
| 3556 | #~((substitute* "makefiles/linux_openblas.mk" |
| 3557 | (("-msse -msse2") "")))) |
| 3558 | (substitute* "makefiles/default_rules.mk" |
| 3559 | (("/bin/bash") (which "bash"))) |
| 3560 | (substitute* "Makefile" |
| 3561 | (("ext_depend: check_portaudio") |
| 3562 | "ext_depend:")) |
| 3563 | (substitute* '("online/Makefile" |
| 3564 | "onlinebin/Makefile" |
| 3565 | "gst-plugin/Makefile") |
| 3566 | (("../../tools/portaudio/install") |
| 3567 | portaudio)) |
| 3568 | (substitute* "matrix/Makefile" ;temporary test bypass |
| 3569 | (("matrix-lib-test sparse-matrix-test") "")) |
| 3570 | (substitute* "cudamatrix/Makefile" |
| 3571 | ((" cu-array-test") "")) |
| 3572 | |
| 3573 | ;; This `configure' script doesn't support variables passed as |
| 3574 | ;; arguments, nor does it support "prefix". |
| 3575 | (substitute* "configure" |
| 3576 | (("check_for_slow_expf;") "") |
| 3577 | ;; This affects the RPATH and also serves as the installation |
| 3578 | ;; directory. |
| 3579 | (("KALDILIBDIR=`pwd`/lib") |
| 3580 | (string-append "KALDILIBDIR=" #$output "/lib")) |
| 3581 | (("OPENBLASROOT=\\\"\\$\\(rel2abs ..\\/tools\\/OpenBLAS\\/install\\)\\\"") |
| 3582 | (string-append "OPENBLASROOT=\"" openblas "\"")) |
| 3583 | (("-L\\$OPENBLASLIBDIR -l:libopenblas.a -l:libblas.a -l:liblapack.a -l:libf2c.a") |
| 3584 | (string-append "-L$OPENBLASLIBDIR -lopenblas " |
| 3585 | "-L" openblas "/lib -lopenblas"))) |
| 3586 | (mkdir-p #$output) ; must exist |
| 3587 | (setenv "CONFIG_SHELL" (which "bash")) |
| 3588 | (setenv "OPENFST_VER" #$(package-version openfst)) |
| 3589 | (invoke "./configure" |
| 3590 | "--use-cuda=no" |
| 3591 | "--mathlib=OPENBLAS_CLAPACK" |
| 3592 | "--shared" |
| 3593 | (string-append "--fst-root=" openfst))))) |
| 3594 | (add-after 'configure 'optimize-build |
| 3595 | (lambda _ (substitute* "kaldi.mk" ((" -O1") " -O3")))) |
| 3596 | (replace 'install |
| 3597 | (lambda _ |
| 3598 | (let ((inc (string-append #$output "/include")) |
| 3599 | (lib (string-append #$output "/lib"))) |
| 3600 | ;; The build phase installed symlinks to the actual |
| 3601 | ;; libraries. Install the actual targets. |
| 3602 | (for-each (lambda (file) |
| 3603 | (let ((target (readlink file))) |
| 3604 | (delete-file file) |
| 3605 | (install-file target lib))) |
| 3606 | (find-files lib "\\.so")) |
| 3607 | ;; Install headers |
| 3608 | (for-each (lambda (file) |
| 3609 | (let ((target-dir (string-append inc "/" (dirname file)))) |
| 3610 | (install-file file target-dir))) |
| 3611 | (find-files "." "\\.h"))))))))))) |
| 3612 | |
| 3613 | (define-public gst-kaldi-nnet2-online |
| 3614 | (let ((commit "7888ae562a65bd7e406783ce2c33535bc66a30ef") |
| 3615 | (revision "3")) |
| 3616 | (package |
| 3617 | (name "gst-kaldi-nnet2-online") |
| 3618 | (version (git-version "0" revision commit)) |
| 3619 | (source (origin |
| 3620 | (method git-fetch) |
| 3621 | (uri (git-reference |
| 3622 | (url "https://github.com/alumae/gst-kaldi-nnet2-online") |
| 3623 | (commit commit))) |
| 3624 | (file-name (git-file-name name version)) |
| 3625 | (sha256 |
| 3626 | (base32 |
| 3627 | "0xp59a6lmx1y24i8bkmxcm27lhm5x5m6y41670yjzhamcbnx8jcr")))) |
| 3628 | (build-system gnu-build-system) |
| 3629 | (arguments |
| 3630 | (list |
| 3631 | #:tests? #f ; there are none |
| 3632 | #:make-flags |
| 3633 | (let ((kaldi (this-package-input "kaldi")) |
| 3634 | (bash (this-package-native-input "bash"))) |
| 3635 | #~(list (string-append "SHELL=" #$bash "/bin/bash") |
| 3636 | (string-append "KALDI_ROOT=" #$(package-source kaldi)) |
| 3637 | (string-append "KALDILIBDIR=" #$kaldi "/lib") |
| 3638 | "KALDI_FLAVOR=dynamic")) |
| 3639 | #:phases |
| 3640 | #~(modify-phases %standard-phases |
| 3641 | (add-after 'unpack 'chdir |
| 3642 | (lambda _ (chdir "src"))) |
| 3643 | (replace 'configure |
| 3644 | (lambda* (#:key inputs #:allow-other-keys) |
| 3645 | (let ((fst-version #$(package-version |
| 3646 | (this-package-input "openfst")))) |
| 3647 | (setenv "CXXFLAGS" |
| 3648 | (string-append |
| 3649 | "-fPIC -DOPENFST_VER=" |
| 3650 | (string-join (string-split fst-version #\.) "0"))) |
| 3651 | (setenv "CPLUS_INCLUDE_PATH" |
| 3652 | (string-join |
| 3653 | (append (map (lambda (dir) |
| 3654 | (search-input-directory inputs dir)) |
| 3655 | '("/include/glib-2.0" |
| 3656 | "/lib/glib-2.0/include" |
| 3657 | "/include/gstreamer-1.0")) |
| 3658 | (list (getenv "CPLUS_INCLUDE_PATH"))) |
| 3659 | ":"))) |
| 3660 | (substitute* "Makefile" |
| 3661 | (("include \\$\\(KALDI_ROOT\\)/src/kaldi.mk") "") |
| 3662 | (("\\$\\(error Cannot find") "#")))) |
| 3663 | (add-before 'build 'build-depend |
| 3664 | (lambda* (#:key make-flags #:allow-other-keys) |
| 3665 | (apply invoke "make" "depend" make-flags))) |
| 3666 | (replace 'install |
| 3667 | (lambda _ |
| 3668 | (install-file |
| 3669 | "libgstkaldinnet2onlinedecoder.so" |
| 3670 | (string-append #$output "/lib/gstreamer-1.0"))))))) |
| 3671 | (inputs |
| 3672 | (list glib gstreamer jansson openfst kaldi)) |
| 3673 | (native-inputs |
| 3674 | (list bash `(,glib "bin") pkg-config)) |
| 3675 | (home-page "https://kaldi-asr.org/") |
| 3676 | (synopsis "Gstreamer plugin for decoding speech") |
| 3677 | (description "This package provides a GStreamer plugin that wraps |
| 3678 | Kaldi's @code{SingleUtteranceNnet2Decoder}. It requires iVector-adapted DNN |
| 3679 | acoustic models. The iVectors are adapted to the current audio stream |
| 3680 | automatically.") |
| 3681 | (license license:asl2.0)))) |
| 3682 | |
| 3683 | (define-public kaldi-gstreamer-server |
| 3684 | ;; This is the tip of the py3 branch |
| 3685 | (let ((commit "f79e204d751a5964918001822e4520fa2acfd246") |
| 3686 | (revision "3")) |
| 3687 | (package |
| 3688 | (name "kaldi-gstreamer-server") |
| 3689 | (version (git-version "0" revision commit)) |
| 3690 | (source (origin |
| 3691 | (method git-fetch) |
| 3692 | (uri (git-reference |
| 3693 | (url "https://github.com/alumae/kaldi-gstreamer-server") |
| 3694 | (commit commit))) |
| 3695 | (file-name (git-file-name name version)) |
| 3696 | (sha256 |
| 3697 | (base32 |
| 3698 | "1iijq8jmgdxr7961inal1ggs496ymxradm51m4sqx8vl983x14y8")) |
| 3699 | ;; XXX: Tests are broken beyond repair and are mingled in the |
| 3700 | ;; source directory. Remove them to avoid installing them. |
| 3701 | ;; See https://github.com/nicolas-graves/kaldi-gstreamer-server |
| 3702 | ;; for a fork that tries to repair them, to no avail. |
| 3703 | (snippet #~(for-each delete-file |
| 3704 | '("kaldigstserver/test-buffer.py" |
| 3705 | "kaldigstserver/decoder_test.py" |
| 3706 | "kaldigstserver/decoder2_test.py"))))) |
| 3707 | (build-system gnu-build-system) |
| 3708 | (arguments |
| 3709 | `(#:tests? #f |
| 3710 | #:modules ((guix build utils) |
| 3711 | (guix build gnu-build-system) |
| 3712 | (srfi srfi-26)) |
| 3713 | #:phases |
| 3714 | (modify-phases %standard-phases |
| 3715 | (delete 'configure) |
| 3716 | (replace 'build |
| 3717 | (lambda _ |
| 3718 | ;; Migrate to Glib.MainLoop. |
| 3719 | (substitute* (find-files "kaldigstserver" "\\.py") |
| 3720 | (("GObject\\.threads_init\\(\\)") "") |
| 3721 | (("GObject") "GLib")) |
| 3722 | ;; Disable hash randomization to ensure the generated .pycs |
| 3723 | ;; are reproducible. |
| 3724 | (setenv "PYTHONHASHSEED" "0") |
| 3725 | (with-directory-excursion "kaldigstserver" |
| 3726 | ;; See https://github.com/alumae/kaldi-gstreamer-server/issues/232 |
| 3727 | (substitute* "master_server.py" |
| 3728 | (("\\.replace\\('\\\\.*") ")")) |
| 3729 | |
| 3730 | (for-each (lambda (file) |
| 3731 | (apply invoke |
| 3732 | `("python" |
| 3733 | "-m" "compileall" |
| 3734 | "-f" ; force rebuild |
| 3735 | ,file))) |
| 3736 | (find-files "." "\\.py$"))))) |
| 3737 | (replace 'install |
| 3738 | (lambda* (#:key inputs outputs #:allow-other-keys) |
| 3739 | (let* ((out (assoc-ref outputs "out")) |
| 3740 | (bin (string-append out "/bin")) |
| 3741 | (share (string-append out "/share/kaldi-gstreamer-server/"))) |
| 3742 | ;; Install Python files |
| 3743 | (with-directory-excursion "kaldigstserver" |
| 3744 | (for-each (cut install-file <> share) |
| 3745 | (find-files "." ".*"))) |
| 3746 | |
| 3747 | ;; Install sample configuration files |
| 3748 | (for-each (cut install-file <> share) |
| 3749 | (find-files "." "\\.yaml")) |
| 3750 | |
| 3751 | ;; Install executables |
| 3752 | (mkdir-p bin) |
| 3753 | (let* ((server (string-append bin "/kaldi-gst-server")) |
| 3754 | (client (string-append bin "/kaldi-gst-client")) |
| 3755 | (worker (string-append bin "/kaldi-gst-worker")) |
| 3756 | (wrap (lambda (wrapper what) |
| 3757 | (with-output-to-file wrapper |
| 3758 | (lambda _ |
| 3759 | (format #t "#!~a |
| 3760 | export GUIX_PYTHONPATH=~a |
| 3761 | export GI_TYPELIB_PATH=~a:${GI_TYPELIB_PATH} |
| 3762 | export GST_PLUGIN_SYSTEM_PATH=~a:${GST_PLUGIN_SYSTEM_PATH} |
| 3763 | exec ~a ~a~a \"$@\"~%" |
| 3764 | (which "bash") |
| 3765 | (getenv "GUIX_PYTHONPATH") |
| 3766 | (getenv "GI_TYPELIB_PATH") |
| 3767 | (getenv "GST_PLUGIN_SYSTEM_PATH") |
| 3768 | (which "python") share what))) |
| 3769 | (chmod wrapper #o555)))) |
| 3770 | (for-each wrap |
| 3771 | (list server client worker) |
| 3772 | (list "master_server.py" |
| 3773 | "client.py" |
| 3774 | "worker.py"))))))))) |
| 3775 | (inputs |
| 3776 | (list gstreamer |
| 3777 | gst-kaldi-nnet2-online |
| 3778 | gst-plugins-base |
| 3779 | gst-plugins-good |
| 3780 | kaldi |
| 3781 | python-wrapper |
| 3782 | python-pygobject |
| 3783 | python-pyyaml |
| 3784 | python-tornado)) |
| 3785 | (home-page "https://github.com/alumae/kaldi-gstreamer-server") |
| 3786 | (synopsis "Real-time full-duplex speech recognition server") |
| 3787 | (description "This is a real-time full-duplex speech recognition server, |
| 3788 | based on the Kaldi toolkit and the GStreamer framework and implemented in |
| 3789 | Python.") |
| 3790 | (license license:bsd-2)))) |
| 3791 | |
| 3792 | (define-public tensorflow-lite |
| 3793 | (package |
| 3794 | (name "tensorflow-lite") |
| 3795 | (version "2.16.2") |
| 3796 | (source |
| 3797 | (origin |
| 3798 | (method git-fetch) |
| 3799 | (uri (git-reference |
| 3800 | (url "https://github.com/tensorflow/tensorflow") |
| 3801 | (commit (string-append "v" version)))) |
| 3802 | (file-name (git-file-name name version)) |
| 3803 | (sha256 |
| 3804 | (base32 |
| 3805 | "082qqv7987qb559skkh8199v8knigylbxgxvz49v4m3467sd4468")))) |
| 3806 | (build-system cmake-build-system) |
| 3807 | (outputs (list "out" "python")) |
| 3808 | (arguments |
| 3809 | (list |
| 3810 | #:build-type "Release" |
| 3811 | #:imported-modules (append %cmake-build-system-modules |
| 3812 | %pyproject-build-system-modules) |
| 3813 | #:modules '((ice-9 match) |
| 3814 | (srfi srfi-26) |
| 3815 | (guix build utils) |
| 3816 | (guix build cmake-build-system) |
| 3817 | ((guix build pyproject-build-system) #:prefix py:)) |
| 3818 | #:configure-flags |
| 3819 | #~(list |
| 3820 | ;; Access some OpenCL functions in mesa-headers |
| 3821 | "-DCMAKE_C_FLAGS=-DCL_ENABLE_BETA_EXTENSIONS=1" |
| 3822 | "-DCMAKE_CXX_FLAGS=-DCL_ENABLE_BETA_EXTENSIONS=1" |
| 3823 | ;; "-DTFLITE_KERNEL_TEST=ON" ; TODO: build tests |
| 3824 | ;; so cmake can be used to find this from other packages |
| 3825 | "-DTFLITE_ENABLE_INSTALL=ON" |
| 3826 | |
| 3827 | ;; Use Guix's own packages as dependencies. |
| 3828 | "-DCMAKE_FIND_PACKAGE_PREFER_CONFIG=ON" |
| 3829 | |
| 3830 | "-DTFLITE_ENABLE_GPU=ON" |
| 3831 | "-DTFLITE_ENABLE_RUY=ON" |
| 3832 | "-DTFLITE_ENABLE_XNNPACK=ON" |
| 3833 | |
| 3834 | "-DSYSTEM_PTHREADPOOL=ON" |
| 3835 | ;; TODO: turn on Farmhash |
| 3836 | ;;"-DSYSTEM_FARMHASH=ON" |
| 3837 | (string-append "-Dabsl_DIR=" #$(this-package-input "abseil-cpp") |
| 3838 | "/lib/cmake/absl") |
| 3839 | (string-append "-DEigen3_DIR=" #$(this-package-input "eigen") |
| 3840 | "/share/eigen3/cmake") |
| 3841 | (string-append "-DFlatBuffers_DIR=" |
| 3842 | #$(this-package-input "flatbuffers") |
| 3843 | "/lib/cmake/flatbuffers") |
| 3844 | (string-append "-DNEON_2_SSE_DIR=" #$(this-package-input "neon2sse") |
| 3845 | "/lib/cmake/NEON_2_SSE") |
| 3846 | (string-append "-Dcpuinfo_DIR=" #$(this-package-input "cpuinfo") |
| 3847 | "/share/cpuinfo") |
| 3848 | (string-append "-Druy_DIR=" #$(this-package-input "ruy") |
| 3849 | "/lib/cmake/ruy") |
| 3850 | (string-append "-DML_DTYPES_LIBRARY_DIRS=" |
| 3851 | #$(this-package-input "python-ml-dtypes") "/lib") |
| 3852 | |
| 3853 | ;; Don't fetch the sources. We have these already |
| 3854 | "-Dml_dtypes_POPULATED=ON" |
| 3855 | "-Dgemmlowp_POPULATED=TRUE" |
| 3856 | "-Degl_headers_POPULATED=TRUE" |
| 3857 | "-Dfp16_headers_POPULATED=TRUE" |
| 3858 | "-Dopencl_headers_POPULATED=TRUE" |
| 3859 | "-Dopengl_headers_POPULATED=TRUE" |
| 3860 | "-Dvulkan_headers_POPULATED=TRUE" |
| 3861 | "-Dgoogletest_POPULATED=TRUE" |
| 3862 | "-Dgoogle_benchmark_POPULATED=TRUE" |
| 3863 | "-Dnsync_POPULATED=TRUE" |
| 3864 | "-Dre2_POPULATED=TRUE" |
| 3865 | "-Dxnnpack_POPULATED=TRUE" |
| 3866 | |
| 3867 | "-DFFT2D_SOURCE_DIR=/tmp/fft2d" |
| 3868 | "-DFARMHASH_SOURCE_DIR=/tmp/farmhash" |
| 3869 | (string-append "-Dgemmlowp_ROOT=" #$(this-package-input "gemmlowp"))) |
| 3870 | #:phases |
| 3871 | #~(modify-phases %standard-phases |
| 3872 | (add-after 'unpack 'chdir |
| 3873 | (lambda _ |
| 3874 | (chdir "tensorflow/lite"))) |
| 3875 | (add-after 'chdir 'unbundle-gemmlowp |
| 3876 | (lambda _ |
| 3877 | (call-with-output-file "tools/cmake/modules/gemmlowp.cmake" |
| 3878 | (lambda (port) |
| 3879 | (display "\ |
| 3880 | add_library(gemmlowp INTERFACE IMPORTED) |
| 3881 | include_directories(\"${gemmlowp_ROOT}/include/gemmlowp\")" port))) |
| 3882 | (call-with-output-file "tools/cmake/modules/ml_dtypes.cmake" |
| 3883 | (lambda (port) |
| 3884 | (display "\ |
| 3885 | add_library(ml_dtypes INTERFACE IMPORTED) |
| 3886 | find_library(ML_DTYPES_LIBRARIES |
| 3887 | NAMES ml_dtypes.so |
| 3888 | PATHS \"${ML_DTYPES_LIBRARY_DIRS}\" |
| 3889 | NO_DEFAULT_PATH)" port))))) |
| 3890 | (add-after 'chdir 'copy-sources |
| 3891 | (lambda _ |
| 3892 | ;; Don't fetch source code; we already have everything we need. |
| 3893 | (substitute* '("tools/cmake/modules/fft2d.cmake" |
| 3894 | "tools/cmake/modules/farmhash.cmake" |
| 3895 | "tools/cmake/modules/gemmlowp.cmake") |
| 3896 | (("^ *OverridableFetchContent_Populate.*") "")) |
| 3897 | |
| 3898 | (mkdir-p "/tmp/farmhash") |
| 3899 | (with-directory-excursion "/tmp/farmhash" |
| 3900 | (copy-recursively #$(this-package-native-input "farmhash-src") ".") |
| 3901 | (for-each make-file-writable (find-files "."))) |
| 3902 | |
| 3903 | (mkdir-p "/tmp/fft2d") |
| 3904 | (with-directory-excursion "/tmp/fft2d" |
| 3905 | (copy-recursively #$(this-package-native-input "fft2d-src") ".") |
| 3906 | (for-each make-file-writable (find-files "."))))) |
| 3907 | (add-after 'copy-sources 'opencl-fix |
| 3908 | (lambda _ |
| 3909 | (substitute* "delegates/gpu/cl/opencl_wrapper.h" |
| 3910 | (("cl_ndrange_kernel_command_properties_khr") |
| 3911 | "cl_command_properties_khr")))) |
| 3912 | (add-after 'opencl-fix 'absl-fix |
| 3913 | (lambda _ |
| 3914 | (substitute* '("delegates/gpu/cl/cl_operation.h" |
| 3915 | "delegates/gpu/common/task/qcom_thin_filter_desc.cc" |
| 3916 | "delegates/gpu/common/tasks/special/thin_pointwise_fuser.cc") |
| 3917 | (("#include <vector>") |
| 3918 | "#include <vector>\n\n#include \"absl/strings/str_cat.h\"\n")))) |
| 3919 | (add-after 'opencl-fix 'stdint-fix |
| 3920 | (lambda _ |
| 3921 | (substitute* "kernels/internal/spectrogram.cc" |
| 3922 | (("#include <math.h>") |
| 3923 | "#include <math.h>\n#include <cstdint>\n")))) |
| 3924 | (add-after 'stdint-fix 'gemmlowp-fix |
| 3925 | (lambda _ |
| 3926 | (substitute* "kernels/internal/common.h" |
| 3927 | (("#include \"fixedpoint/fixedpoint\\.h\"") |
| 3928 | "#include <fixedpoint/fixedpoint.h>")))) |
| 3929 | (add-after 'gemmlowp-fix 'xnn-delegate-fix |
| 3930 | ;; undefined function, remove call to it |
| 3931 | (lambda _ |
| 3932 | (substitute* "delegates/xnnpack/xnnpack_delegate.cc" |
| 3933 | (("xnn_experiment_enable_adaptive_avx_optimization\\(\\);") |
| 3934 | "")))) |
| 3935 | (add-after 'build 'build-shared-library |
| 3936 | (lambda* (#:key configure-flags #:allow-other-keys) |
| 3937 | (mkdir-p "c") |
| 3938 | (with-directory-excursion "c" |
| 3939 | (apply invoke "cmake" (append configure-flags (list "../../lite/c"))) |
| 3940 | (invoke "cmake" "--build" "." "-j" (number->string |
| 3941 | (parallel-job-count)))))) |
| 3942 | (add-after 'build-shared-library 'build-benchmark-model |
| 3943 | (lambda _ |
| 3944 | (invoke "cmake" "--build" "." "--target" "benchmark_model" |
| 3945 | "-j" (number->string (parallel-job-count))))) |
| 3946 | (add-after 'build-benchmark-model 'build-python |
| 3947 | (lambda* (#:key configure-flags #:allow-other-keys) |
| 3948 | (let ((script (string-append "../lite/tools/pip_package/" |
| 3949 | "build_pip_package_with_cmake.sh"))) |
| 3950 | (substitute* script |
| 3951 | (("\"\\$\\{TENSORFLOW_LITE_DIR\\}\"" all) |
| 3952 | (string-append "${ADDITIONAL_CMAKE_FLAGS} " all)) |
| 3953 | (("-D(CMAKE_C[X]*_FLAGS)=\"\\$\\{BUILD_FLAGS\\}" all flag) |
| 3954 | (string-append all " ${ADDITIONAL_" flag "}"))) |
| 3955 | (setenv "BUILD_NUM_JOBS" (number->string (parallel-job-count))) |
| 3956 | (let* ((config (map (cut string-split <> #\=) configure-flags)) |
| 3957 | (c-flags (assoc-ref config "-DCMAKE_C_FLAGS")) |
| 3958 | (config (alist-delete "-DCMAKE_C_FLAGS" config)) |
| 3959 | (cxx-flags (assoc-ref config "-DCMAKE_CXX_FLAGS")) |
| 3960 | (config (alist-delete "-DCMAKE_CXX_FLAGS" config))) |
| 3961 | (setenv "ADDITIONAL_CMAKE_C_FLAGS" |
| 3962 | (string-join c-flags "=")) |
| 3963 | (setenv "ADDITIONAL_CMAKE_CXX_FLAGS" |
| 3964 | (string-join cxx-flags "=")) |
| 3965 | (setenv "ADDITIONAL_CMAKE_FLAGS" |
| 3966 | (string-join (map (cut string-join <> "=") config) |
| 3967 | " "))) |
| 3968 | (invoke "sh" script)))) |
| 3969 | (add-after 'install 'install-extra |
| 3970 | (lambda _ |
| 3971 | (install-file "../build/c/libtensorflowlite_c.so" |
| 3972 | (string-append #$output "/lib")) |
| 3973 | (install-file "../build/tools/benchmark/benchmark_model" |
| 3974 | (string-append #$output "/bin")))) |
| 3975 | (add-after 'install-extra 'install-python |
| 3976 | (lambda* (#:key inputs outputs #:allow-other-keys) |
| 3977 | (with-directory-excursion |
| 3978 | "../lite/tools/pip_package/gen/tflite_pip/python3" |
| 3979 | ((assoc-ref py:%standard-phases 'install) |
| 3980 | #:inputs inputs |
| 3981 | #:outputs `(("out" . ,#$output:python)))))) |
| 3982 | (replace 'check |
| 3983 | (lambda* (#:key tests? #:allow-other-keys) |
| 3984 | (when tests? |
| 3985 | (invoke "ctest" "-L" "plain")))) |
| 3986 | (add-after 'install-python 'add-install-to-pythonpath |
| 3987 | (lambda* (#:key inputs #:allow-other-keys) |
| 3988 | ((assoc-ref py:%standard-phases 'add-install-to-pythonpath) |
| 3989 | #:inputs inputs |
| 3990 | #:outputs `(("out" . ,#$output:python))))) |
| 3991 | (add-after 'add-install-to-pythonpath 'python-sanity-check |
| 3992 | (lambda* (#:key tests? inputs #:allow-other-keys) |
| 3993 | ((assoc-ref py:%standard-phases 'sanity-check) |
| 3994 | #:inputs `(("sanity-check.py" . ,#$(default-sanity-check.py)) |
| 3995 | ,@inputs) |
| 3996 | #:outputs `(("out" . ,#$output:python)))))))) |
| 3997 | (inputs |
| 3998 | (list abseil-cpp |
| 3999 | cpuinfo |
| 4000 | eigen |
| 4001 | fp16 |
| 4002 | flatbuffers-23.5 |
| 4003 | gemmlowp |
| 4004 | mesa-headers |
| 4005 | neon2sse |
| 4006 | nsync |
| 4007 | opencl-icd-loader |
| 4008 | pthreadpool |
| 4009 | python ;for its /lib |
| 4010 | python-ml-dtypes |
| 4011 | ruy |
| 4012 | re2 |
| 4013 | xnnpack |
| 4014 | vulkan-headers |
| 4015 | zlib)) |
| 4016 | (propagated-inputs |
| 4017 | (list python-numpy)) |
| 4018 | (native-inputs |
| 4019 | (list pkg-config |
| 4020 | googletest |
| 4021 | pybind11-2 |
| 4022 | python-setuptools |
| 4023 | python-wrapper ;for its /bin |
| 4024 | swig-4.0 |
| 4025 | (origin |
| 4026 | (method git-fetch) |
| 4027 | (uri (git-reference |
| 4028 | (url "https://github.com/google/farmhash") |
| 4029 | (commit "816a4ae622e964763ca0862d9dbd19324a1eaf45"))) |
| 4030 | (file-name "farmhash-src") |
| 4031 | (sha256 |
| 4032 | (base32 |
| 4033 | "1mqxsljq476n1hb8ilkrpb39yz3ip2hnc7rhzszz4sri8ma7qzp6"))) |
| 4034 | (origin |
| 4035 | (method git-fetch) |
| 4036 | (uri (git-reference |
| 4037 | (url "https://github.com/petewarden/OouraFFT") |
| 4038 | (commit "v1.0"))) |
| 4039 | (file-name "fft2d-src") |
| 4040 | (sha256 |
| 4041 | (base32 |
| 4042 | "1gla8m477din9k7jnbkzzbvc0wzw2rn97skxwf0k0mwcdf6vlhcs"))))) |
| 4043 | (home-page "https://www.tensorflow.org") |
| 4044 | (synopsis "Machine learning framework") |
| 4045 | (description |
| 4046 | "TensorFlow is a flexible platform for building and training machine |
| 4047 | learning models. This package provides the \"lite\" variant for mobile |
| 4048 | devices.") |
| 4049 | (license license:asl2.0))) |
| 4050 | |
| 4051 | (define-public python-tflite-runtime |
| 4052 | (package |
| 4053 | (inherit tensorflow-lite) |
| 4054 | (name "python-tflite-runtime") |
| 4055 | (source #f) |
| 4056 | (build-system trivial-build-system) |
| 4057 | (outputs (list "out")) |
| 4058 | (arguments |
| 4059 | (list |
| 4060 | #:builder |
| 4061 | #~(begin |
| 4062 | (mkdir #$output) |
| 4063 | (symlink (string-append #$tensorflow-lite:python "/lib") |
| 4064 | (string-append #$output "/lib"))))) |
| 4065 | (native-inputs '()))) |
| 4066 | |
| 4067 | (define-public tflite-micro |
| 4068 | (let ((commit "a94423c6b6ee1f30c82d60952d70f749296a5581") |
| 4069 | (version "0") |
| 4070 | (revision "0")) |
| 4071 | (package |
| 4072 | (name "tflite-micro") |
| 4073 | (version (git-version version revision commit)) |
| 4074 | (source |
| 4075 | (origin |
| 4076 | (method git-fetch) |
| 4077 | (uri (git-reference |
| 4078 | (url "https://github.com/tensorflow/tflite-micro") |
| 4079 | (commit commit))) |
| 4080 | (file-name (git-file-name name version)) |
| 4081 | (sha256 |
| 4082 | (base32 "0q1l4chj4p7vwclcysjmzl017l08kpqh498wpghq1cph641mdp5n")) |
| 4083 | (patches (search-patches "tflite-micro_disable_download.patch")))) |
| 4084 | (build-system gnu-build-system) |
| 4085 | (arguments |
| 4086 | (list |
| 4087 | #:tests? #f ;there are no tests |
| 4088 | #:phases |
| 4089 | #~(modify-phases %standard-phases |
| 4090 | (add-after 'unpack 'patch-makefile |
| 4091 | (lambda _ |
| 4092 | (let ((makefile "tensorflow/lite/micro/tools/make/Makefile")) |
| 4093 | (substitute* makefile |
| 4094 | (("BASE_GENDIR := gen") |
| 4095 | "BASE_GENDIR := $(BUILD_DIR)"))))) |
| 4096 | (delete 'configure) |
| 4097 | |
| 4098 | (replace 'build |
| 4099 | (lambda _ |
| 4100 | (let ((makefile "tensorflow/lite/micro/tools/make/Makefile")) |
| 4101 | (mkdir-p "../build") |
| 4102 | (invoke "make" |
| 4103 | "-j" |
| 4104 | (number->string (parallel-job-count)) |
| 4105 | "-f" |
| 4106 | makefile |
| 4107 | (string-append "BASE_GENDIR=" "../build") |
| 4108 | (string-append "PRE_INCLUDES=-I" |
| 4109 | #$(this-package-input "gemmlowp") |
| 4110 | "/include/gemmlowp" |
| 4111 | " -I" |
| 4112 | #$(this-package-input |
| 4113 | "kissfft-for-tflite-micro") |
| 4114 | "/include/kissfft" |
| 4115 | " -I" |
| 4116 | #$(package-source |
| 4117 | kissfft-for-tflite-micro)))))) |
| 4118 | (replace 'install |
| 4119 | (lambda _ |
| 4120 | (for-each (lambda (f) |
| 4121 | (install-file f |
| 4122 | (string-append #$output "/lib"))) |
| 4123 | (find-files "../build" "\\.a$")) |
| 4124 | (for-each (lambda (f) |
| 4125 | (install-file f |
| 4126 | (string-append #$output |
| 4127 | "/include/lite/micro"))) |
| 4128 | (find-files "tensorflow/lite/micro" "\\.h$")) |
| 4129 | (for-each (lambda (f) |
| 4130 | (install-file f |
| 4131 | (string-append #$output |
| 4132 | "/include/lite/micro"))) |
| 4133 | (find-files "signal/micro/kernels" "\\.h$")) |
| 4134 | (for-each (lambda (f) |
| 4135 | (install-file f |
| 4136 | (string-append #$output |
| 4137 | "/include/lite/micro"))) |
| 4138 | (find-files "signal/src" "\\.h$"))))))) |
| 4139 | (inputs (list flatbuffers-23.5 gemmlowp kissfft-for-tflite-micro ruy)) |
| 4140 | (native-inputs (list python python-pillow python-numpy)) |
| 4141 | (home-page "https://ai.google.dev/edge/litert/microcontrollers/overview") |
| 4142 | (synopsis |
| 4143 | "Infrastructure to enable deployment of ML models to embedded targets ") |
| 4144 | (description |
| 4145 | "TensorFlow Lite for Microcontrollers is a port of TensorFlow Lite designed |
| 4146 | to run machine learning models on DSPs, microcontrollers and other devices with |
| 4147 | limited memory.") |
| 4148 | (license license:asl2.0)))) |
| 4149 | |
| 4150 | |
| 4151 | (define-public dmlc-core |
| 4152 | (package |
| 4153 | (name "dmlc-core") |
| 4154 | (version "0.5") |
| 4155 | (source |
| 4156 | (origin |
| 4157 | (method git-fetch) |
| 4158 | (uri (git-reference |
| 4159 | (url "https://github.com/dmlc/dmlc-core") |
| 4160 | (commit (string-append "v" version)))) |
| 4161 | (file-name (git-file-name name version)) |
| 4162 | (sha256 |
| 4163 | (base32 "1x4ad1jhn84fywlk031fmv1kxyiscclmrqn9hhj8gz0mh7z9vcrh")))) |
| 4164 | (build-system cmake-build-system) |
| 4165 | (arguments |
| 4166 | (list |
| 4167 | #:configure-flags #~(list "-DGOOGLE_TEST=ON"))) |
| 4168 | (native-inputs |
| 4169 | (list googletest python-wrapper)) |
| 4170 | (home-page "https://github.com/dmlc/dmlc-core") |
| 4171 | (synopsis "Common bricks library for machine learning") |
| 4172 | (description |
| 4173 | "DMLC-Core is the backbone library to support all DMLC projects, |
| 4174 | offers the bricks to build efficient and scalable distributed machine |
| 4175 | learning libraries.") |
| 4176 | (license license:asl2.0))) |
| 4177 | |
| 4178 | (define-public dmlc-core-next |
| 4179 | ;; Apache Tvm needs the latest code from git commits in May 2024 or later. |
| 4180 | (let ((commit "13341857549852a9a86b1894b5ba84c6276ab381") |
| 4181 | (revision "1")) |
| 4182 | (package |
| 4183 | (inherit dmlc-core) |
| 4184 | (name "dmlc-core") |
| 4185 | (version (git-version "0.5" revision commit)) |
| 4186 | (source |
| 4187 | (origin |
| 4188 | (method git-fetch) |
| 4189 | (uri (git-reference |
| 4190 | (url "https://github.com/dmlc/dmlc-core") |
| 4191 | (commit commit))) |
| 4192 | (file-name (git-file-name name version)) |
| 4193 | (sha256 |
| 4194 | (base32 "1hj6h103lal3cm5mnry9lrm3d7aij89rxv46yv6y49vzd5hbnfbd"))))))) |
| 4195 | |
| 4196 | (define-public xgboost |
| 4197 | (package |
| 4198 | (name "xgboost") |
| 4199 | (version "1.7.6") |
| 4200 | (source |
| 4201 | (origin |
| 4202 | (method git-fetch) |
| 4203 | (uri (git-reference |
| 4204 | (url "https://github.com/dmlc/xgboost") |
| 4205 | (commit (string-append "v" version)))) |
| 4206 | (file-name (git-file-name name version)) |
| 4207 | (patches (search-patches "xgboost-use-system-dmlc-core.patch")) |
| 4208 | (sha256 |
| 4209 | (base32 "16fbm5y3hn6ccflmbdlmn7krrdq7c0az3mxd8j1d23s9ky8niw05")))) |
| 4210 | (build-system cmake-build-system) |
| 4211 | (arguments |
| 4212 | (list #:configure-flags #~(list "-DGOOGLE_TEST=ON"))) |
| 4213 | (native-inputs |
| 4214 | (list googletest |
| 4215 | python-wrapper)) |
| 4216 | (inputs |
| 4217 | (list dmlc-core)) |
| 4218 | (home-page "https://xgboost.ai/") |
| 4219 | (synopsis "Gradient boosting (GBDT, GBRT or GBM) library") |
| 4220 | (description |
| 4221 | "XGBoost is an optimized distributed gradient boosting library designed |
| 4222 | to be highly efficient, flexible and portable. It implements machine learning |
| 4223 | algorithms under the Gradient Boosting framework. XGBoost provides a parallel |
| 4224 | tree boosting (also known as GBDT, GBM) that solve many data science problems |
| 4225 | in a fast and accurate way.") |
| 4226 | (license license:asl2.0))) |
| 4227 | |
| 4228 | (define-public python-xgboost |
| 4229 | (package |
| 4230 | (inherit xgboost) |
| 4231 | (name "python-xgboost") |
| 4232 | (source (package-source xgboost)) |
| 4233 | (build-system pyproject-build-system) |
| 4234 | (arguments |
| 4235 | (list |
| 4236 | #:tests? #f ; all tests require network access |
| 4237 | #:phases |
| 4238 | #~(modify-phases %standard-phases |
| 4239 | (add-after 'unpack 'preparations |
| 4240 | (lambda _ |
| 4241 | ;; Move python-package content to parent directory to silence |
| 4242 | ;; some warnings about files not being found if we chdir. |
| 4243 | (rename-file "python-package/xgboost" "xgboost") |
| 4244 | (rename-file "python-package/README.rst" "README.rst") |
| 4245 | (rename-file "python-package/setup.cfg" "setup.cfg") |
| 4246 | (rename-file "python-package/setup.py" "setup.py") |
| 4247 | ;; Skip rebuilding libxgboost.so. |
| 4248 | (substitute* "setup.py" |
| 4249 | (("ext_modules=\\[CMakeExtension\\('libxgboost'\\)\\],") "") |
| 4250 | (("'install_lib': InstallLib,") "")) |
| 4251 | ;; Remove bad dataset. This has been removed in scipy. |
| 4252 | (substitute* "tests/python/testing.py" |
| 4253 | (("TestDataset\\('boston', get_boston, 'reg:squarederror', 'rmse'\\),") |
| 4254 | "") |
| 4255 | (("datasets.load_boston") |
| 4256 | "datasets.load_digits")))) |
| 4257 | (add-after 'install 'install-version-and-libxgboost |
| 4258 | (lambda* (#:key inputs #:allow-other-keys) |
| 4259 | (let* ((pylib (string-append #$output "/lib/python" |
| 4260 | #$(version-major+minor |
| 4261 | (package-version python)) |
| 4262 | "/site-packages")) |
| 4263 | (xgbdir (string-append pylib "/xgboost")) |
| 4264 | (version-file (string-append xgbdir "/VERSION")) |
| 4265 | (libxgboost (string-append (assoc-ref inputs "xgboost") |
| 4266 | "/lib/libxgboost.so"))) |
| 4267 | (with-output-to-file version-file |
| 4268 | (lambda () |
| 4269 | (display #$(package-version xgboost)))) |
| 4270 | (mkdir-p (string-append xgbdir "/lib")) |
| 4271 | (symlink libxgboost (string-append xgbdir "/lib" |
| 4272 | "/libxgboost.so")))))))) |
| 4273 | (native-inputs |
| 4274 | (list python-pandas python-pytest python-scikit-learn python-setuptools |
| 4275 | python-wheel)) |
| 4276 | (inputs |
| 4277 | (list xgboost)) |
| 4278 | (propagated-inputs |
| 4279 | (list python-numpy python-scipy)) |
| 4280 | (synopsis "Python interface for the XGBoost library"))) |
| 4281 | |
| 4282 | (define-public gloo |
| 4283 | (let ((version "0.0.0") ; no proper version tag |
| 4284 | (commit "54cbae0d3a67fa890b4c3d9ee162b7860315e341") |
| 4285 | (revision "4")) |
| 4286 | (package |
| 4287 | (name "gloo") |
| 4288 | (version (git-version version revision commit)) |
| 4289 | (source |
| 4290 | (origin |
| 4291 | (method git-fetch) |
| 4292 | (uri (git-reference |
| 4293 | (url "https://github.com/pytorch/gloo") |
| 4294 | (commit commit))) |
| 4295 | (file-name (git-file-name name version)) |
| 4296 | (sha256 |
| 4297 | (base32 |
| 4298 | "1zixybyma7zpwdawy1qi2d48g9r65wcjjd4j3j2bhlvjymyw03z2")))) |
| 4299 | (build-system cmake-build-system) |
| 4300 | (native-inputs |
| 4301 | (list googletest)) |
| 4302 | (inputs |
| 4303 | (append (list openssl) |
| 4304 | (if (supported-package? rdma-core) |
| 4305 | (list rdma-core) |
| 4306 | '()))) |
| 4307 | (arguments |
| 4308 | (list #:configure-flags #~'("-DBUILD_SHARED_LIBS=ON" |
| 4309 | "-DBUILD_TEST=1" |
| 4310 | "-DCMAKE_CXX_STANDARD=17" |
| 4311 | #$@(if (this-package-input "rdma-core") |
| 4312 | #~("-DUSE_IBVERBS=ON") |
| 4313 | #~())) |
| 4314 | #:phases |
| 4315 | #~(modify-phases %standard-phases |
| 4316 | (replace 'check |
| 4317 | (lambda* (#:key tests? #:allow-other-keys) |
| 4318 | (when tests? |
| 4319 | (invoke "make" "gloo_test"))))))) |
| 4320 | (synopsis "Collective communications library") |
| 4321 | (description |
| 4322 | "Gloo is a collective communications library. It comes with a |
| 4323 | number of collective algorithms useful for machine learning applications. |
| 4324 | These include a barrier, broadcast, and allreduce.") |
| 4325 | (home-page "https://github.com/facebookincubator/gloo") |
| 4326 | (supported-systems %64bit-supported-systems) |
| 4327 | (license license:bsd-3)))) |
| 4328 | |
| 4329 | (define-public gloo-for-r-torch |
| 4330 | (let ((version "0.0.0") ; no proper version tag |
| 4331 | (commit "81925d1c674c34f0dc34dd9a0f2151c1b6f701eb") |
| 4332 | (revision "2")) |
| 4333 | (package |
| 4334 | (name "gloo") |
| 4335 | (version (git-version version revision commit)) |
| 4336 | (source |
| 4337 | (origin |
| 4338 | (method git-fetch) |
| 4339 | (uri (git-reference |
| 4340 | (url "https://github.com/facebookincubator/gloo") |
| 4341 | (commit commit))) |
| 4342 | (file-name (git-file-name name version)) |
| 4343 | (sha256 |
| 4344 | (base32 |
| 4345 | "16zs8ndbiv9nppn8bv6lfanzyyssz7g5pawxiqcnafwq3nvxpj9m")))) |
| 4346 | (build-system cmake-build-system) |
| 4347 | (native-inputs |
| 4348 | (list googletest)) |
| 4349 | (inputs |
| 4350 | (append (list openssl-1.1) |
| 4351 | (if (supported-package? rdma-core) |
| 4352 | (list rdma-core) |
| 4353 | '()))) |
| 4354 | (arguments |
| 4355 | (list #:configure-flags #~'("-DBUILD_SHARED_LIBS=ON" |
| 4356 | "-DBUILD_TEST=1" |
| 4357 | "-DCMAKE_CXX_STANDARD=17" |
| 4358 | #$@(if (this-package-input "rdma-core") |
| 4359 | #~("-DUSE_IBVERBS=ON") |
| 4360 | #~())) |
| 4361 | #:phases |
| 4362 | #~(modify-phases %standard-phases |
| 4363 | (replace 'check |
| 4364 | (lambda* (#:key tests? #:allow-other-keys) |
| 4365 | (when tests? |
| 4366 | (invoke "make" "gloo_test"))))))) |
| 4367 | (synopsis "Collective communications library") |
| 4368 | (description |
| 4369 | "Gloo is a collective communications library. It comes with a |
| 4370 | number of collective algorithms useful for machine learning applications. |
| 4371 | These include a barrier, broadcast, and allreduce.") |
| 4372 | (home-page "https://github.com/facebookincubator/gloo") |
| 4373 | (supported-systems %64bit-supported-systems) |
| 4374 | (license license:bsd-3)))) |
| 4375 | |
| 4376 | (define-public python-tensorly |
| 4377 | (package |
| 4378 | (name "python-tensorly") |
| 4379 | (version "0.9.0") |
| 4380 | (source (origin |
| 4381 | (method git-fetch) |
| 4382 | (uri (git-reference |
| 4383 | (url "https://github.com/tensorly/tensorly") |
| 4384 | (commit version))) |
| 4385 | (file-name (git-file-name name version)) |
| 4386 | (sha256 |
| 4387 | (base32 |
| 4388 | "01xdkhzwq7s18pp6433d4dhyblmlhjs87acagxh73vfsqyknb9h3")))) |
| 4389 | (build-system pyproject-build-system) |
| 4390 | (arguments |
| 4391 | (list |
| 4392 | #:test-flags |
| 4393 | '(list |
| 4394 | ;; These tests fail due to missing example, documentation, or tutorial files. |
| 4395 | "--ignore=doc/sphinx_ext/sphinx_gallery/tests/test_gen_rst.py" |
| 4396 | ;; XXX There is no "get_marker" method. |
| 4397 | "--ignore=doc/sphinx_ext/sphinx_gallery/tests/test_gen_gallery.py" |
| 4398 | "-k" |
| 4399 | (string-append |
| 4400 | ;; tutorials/plot_parse.py is not included |
| 4401 | "not test_jupyter_notebook" |
| 4402 | ;; nor is examples/plot_quantum.py |
| 4403 | " and not test_file_is_generated")))) |
| 4404 | (propagated-inputs (list python-jsmin python-numpy python-scipy)) |
| 4405 | (native-inputs (list python-pytest python-sphinx |
| 4406 | python-setuptools python-wheel)) |
| 4407 | (home-page "https://github.com/tensorly/tensorly") |
| 4408 | (synopsis "Tensor learning in Python") |
| 4409 | (description |
| 4410 | "This is a Python library that aims at making tensor learning simple and |
| 4411 | accessible. It allows performing tensor decomposition, tensor learning and |
| 4412 | tensor algebra easily. Its backend system allows seamlessly perform |
| 4413 | computation with NumPy, PyTorch, JAX, MXNet, TensorFlow or CuPy and run |
| 4414 | methodxs at scale on CPU or GPU.") |
| 4415 | (license license:bsd-3))) |
| 4416 | |
| 4417 | (define-public python-umap-learn |
| 4418 | (package |
| 4419 | (name "python-umap-learn") |
| 4420 | (version "0.5.12") |
| 4421 | (source |
| 4422 | (origin |
| 4423 | (method git-fetch) |
| 4424 | (uri (git-reference |
| 4425 | (url "https://github.com/lmcinnes/umap") |
| 4426 | (commit (string-append "release-" version)))) |
| 4427 | (file-name (git-file-name name version)) |
| 4428 | (sha256 |
| 4429 | (base32 "18ahwd94z6ravxvgpw3wyclcyzfpig93197znzgsg2350bgnzr1l")))) |
| 4430 | (build-system pyproject-build-system) |
| 4431 | (arguments |
| 4432 | (list |
| 4433 | ;; tests: 253 passed, 65 skipped, 56 warnings |
| 4434 | #:phases |
| 4435 | #~(modify-phases %standard-phases |
| 4436 | ;; Numba needs a writable dir to cache functions. |
| 4437 | (add-before 'build 'set-numba-cache-dir |
| 4438 | (lambda _ |
| 4439 | (setenv "NUMBA_CACHE_DIR" "/tmp")))))) |
| 4440 | (native-inputs |
| 4441 | (list python-pytest |
| 4442 | python-setuptools)) |
| 4443 | (propagated-inputs |
| 4444 | (list python-numba |
| 4445 | python-numpy |
| 4446 | python-pynndescent |
| 4447 | python-scikit-learn |
| 4448 | python-scipy |
| 4449 | python-tqdm)) |
| 4450 | (home-page "https://github.com/lmcinnes/umap") |
| 4451 | (synopsis "Uniform Manifold Approximation and Projection") |
| 4452 | (description |
| 4453 | "Uniform Manifold Approximation and Projection is a dimension reduction |
| 4454 | technique that can be used for visualization similarly to t-SNE, but also for |
| 4455 | general non-linear dimension reduction.") |
| 4456 | (license license:bsd-3))) |
| 4457 | |
| 4458 | ;; Requires AVX2 on x86_64-linux. |
| 4459 | (define-public nnpack |
| 4460 | (let ((version "0.0") |
| 4461 | (commit "70a77f485e8b934224f3a79efd8edcd84cd377b8") |
| 4462 | (revision "2")) |
| 4463 | (package |
| 4464 | (name "nnpack") |
| 4465 | (version (git-version version revision commit)) |
| 4466 | (home-page "https://github.com/Maratyszcza/NNPACK") |
| 4467 | (source (origin |
| 4468 | (method git-fetch) |
| 4469 | (uri (git-reference (url home-page) (commit commit))) |
| 4470 | (file-name (git-file-name name version)) |
| 4471 | (sha256 |
| 4472 | (base32 |
| 4473 | "0c4pw926279s3rlx7mg4l4vhnfy6dh374n6w7zqhcn0bxpym1hv1")) |
| 4474 | (patches (search-patches "nnpack-system-libraries.patch")))) |
| 4475 | (build-system cmake-build-system) |
| 4476 | ;; XXX: The test suite runs but it's very expensive. On x86_64-linux, it |
| 4477 | ;; requires AVX2 instructions. |
| 4478 | (arguments '(#:tests? #f)) |
| 4479 | (synopsis "Acceleration package for neural network computations") |
| 4480 | (description |
| 4481 | "NNPACK is an acceleration package for neural network computations. |
| 4482 | NNPACK aims to provide high-performance implementations of convnet layers for |
| 4483 | multi-core CPUs. |
| 4484 | |
| 4485 | NNPACK is not intended to be directly used by machine learning researchers; |
| 4486 | instead it provides low-level performance primitives leveraged in leading deep |
| 4487 | learning frameworks, such as PyTorch, Caffe2, MXNet, tiny-dnn, Caffe, Torch, |
| 4488 | and Darknet.") |
| 4489 | (inputs |
| 4490 | (list cpuinfo |
| 4491 | fp16 |
| 4492 | fxdiv |
| 4493 | psimd |
| 4494 | pthreadpool |
| 4495 | googletest-1.12)) |
| 4496 | (native-inputs |
| 4497 | `(,python |
| 4498 | ,@(if (target-x86-64?) (list python-peachpy) '()) |
| 4499 | ,python-six)) |
| 4500 | ;; Supported for Linux. |
| 4501 | (supported-systems '("x86_64-linux" "armhf-linux" "aarch64-linux")) |
| 4502 | (license license:bsd-2)))) |
| 4503 | |
| 4504 | (define-public qnnpack |
| 4505 | (let ((commit "7d2a4e9931a82adc3814275b6219a03e24e36b4c") |
| 4506 | (revision "0")) |
| 4507 | (package |
| 4508 | (name "qnnpack") |
| 4509 | (version (git-version "0" revision commit)) |
| 4510 | (source |
| 4511 | (origin |
| 4512 | (method git-fetch) |
| 4513 | (uri (git-reference |
| 4514 | (url "https://github.com/pytorch/qnnpack") |
| 4515 | (commit commit))) |
| 4516 | (file-name (git-file-name name version)) |
| 4517 | (sha256 |
| 4518 | (base32 |
| 4519 | "1dgzriiaz9arsrfwhx42y4l74wbzn6xvdmllfb66v4pmvi5gpxc5")) |
| 4520 | (modules '((guix build utils))) |
| 4521 | (snippet |
| 4522 | '(delete-file-recursively "deps")))) |
| 4523 | (build-system cmake-build-system) |
| 4524 | (arguments |
| 4525 | (list |
| 4526 | ;; The shared library build fails with linker errors, so we build the |
| 4527 | ;; static library with -fPIC as in the bundled PyTorch version. |
| 4528 | #:configure-flags |
| 4529 | ''("-DQNNPACK_LIBRARY_TYPE=static" |
| 4530 | "-DCMAKE_POSITION_INDEPENDENT_CODE=ON") |
| 4531 | #:phases |
| 4532 | #~(modify-phases %standard-phases |
| 4533 | (add-after 'unpack 'patch-cmake |
| 4534 | (lambda _ |
| 4535 | (substitute* "CMakeLists.txt" |
| 4536 | (("IF.*SOURCE_DIR.*") |
| 4537 | "IF(FALSE)\n") |
| 4538 | (("IF\\(NOT TARGET.*") |
| 4539 | "IF(FALSE)\n") |
| 4540 | (("TARGET_LINK_LIBRARIES.*(fxdiv|psimd|fp16)\\).*") |
| 4541 | "") |
| 4542 | (("(TARGET_LINK_LIBRARIES.*) fp16 (.*)" _ before after) |
| 4543 | (string-append before " " after)))))))) |
| 4544 | (inputs (list clog cpuinfo fp16 fxdiv psimd pthreadpool)) |
| 4545 | (native-inputs (list googletest-1.12 googlebenchmark)) |
| 4546 | (home-page "https://github.com/pytorch/qnnpack") |
| 4547 | (synopsis "Quantized Neural Network PACKage") |
| 4548 | (description "QNNPACK is a library for low-precision neural network |
| 4549 | inference. It contains the implementation of common neural network operators |
| 4550 | on quantized 8-bit tensors.") |
| 4551 | (supported-systems |
| 4552 | '("armhf-linux" "aarch64-linux" "i686-linux" "x86_64-linux")) |
| 4553 | (license license:bsd-3)))) |
| 4554 | |
| 4555 | (define-public xnnpack |
| 4556 | ;; There's currently no tag on this repo. |
| 4557 | (let ((version "0.0") |
| 4558 | (commit "51a0103656eff6fc9bfd39a4597923c4b542c883") |
| 4559 | (revision "4")) |
| 4560 | (package |
| 4561 | (name "xnnpack") |
| 4562 | (version (git-version version revision commit)) |
| 4563 | (home-page "https://github.com/google/XNNPACK") ;fork of QNNPACK |
| 4564 | (source |
| 4565 | (origin |
| 4566 | (method git-fetch) |
| 4567 | (uri (git-reference (url home-page) (commit commit))) |
| 4568 | (file-name (git-file-name name version)) |
| 4569 | (sha256 |
| 4570 | (base32 |
| 4571 | "0j4smj8yaxw6r14caqn0dgb8j8mnsq34zp8ixxn11zmgbab306ly")) |
| 4572 | (modules '((guix build utils) |
| 4573 | (ice-9 ftw) |
| 4574 | (ice-9 textual-ports) |
| 4575 | (srfi srfi-26))) |
| 4576 | (snippet |
| 4577 | '(begin |
| 4578 | (when (directory-exists? "bench/models") |
| 4579 | (delete-file-recursively "bench/models")) |
| 4580 | ;; Remove autogenerated files, which contain the string |
| 4581 | ;; "Auto-generated file" |
| 4582 | (for-each |
| 4583 | (lambda (dir) |
| 4584 | (for-each |
| 4585 | (lambda (name) |
| 4586 | (let ((path (string-append dir "/" name))) |
| 4587 | (when (call-with-input-file path |
| 4588 | (lambda (port) |
| 4589 | (string-contains |
| 4590 | (get-string-all port) |
| 4591 | "Auto-generated file"))) |
| 4592 | (delete-file path)))) |
| 4593 | (or (scandir dir (negate (cut member <> '("." ".." "simd")))) |
| 4594 | '()))) |
| 4595 | (cons* |
| 4596 | "test" "bench" "src/enums" "src/xnnpack" "gen" "cmake/gen" |
| 4597 | (filter file-exists? |
| 4598 | (map (cut string-append "src/" <> "/gen") |
| 4599 | (scandir "src"))))))))) |
| 4600 | (build-system cmake-build-system) |
| 4601 | (arguments |
| 4602 | (list |
| 4603 | #:build-type "Release" ;; Debugging symbols require a lot of disk space |
| 4604 | #:configure-flags |
| 4605 | #~(list "-DXNNPACK_USE_SYSTEM_LIBS=YES" |
| 4606 | "-DBUILD_SHARED_LIBS=ON" |
| 4607 | "-DCMAKE_POSITION_INDEPENDENT_CODE=ON" |
| 4608 | "-DXNNPACK_LIBRARY_TYPE=shared" |
| 4609 | "-DXNNPACK_BUILD_BENCHMARKS=FALSE" |
| 4610 | ;; Tests fail to build with -DXNNPACK_LIBRARY_TYPE=shared: |
| 4611 | ;; https://github.com/google/XNNPACK/issues/6285 |
| 4612 | "-DXNNPACK_BUILD_TESTS=OFF") |
| 4613 | #:tests? #f |
| 4614 | #:modules '((ice-9 ftw) |
| 4615 | (guix build cmake-build-system) |
| 4616 | (guix build utils)) |
| 4617 | #:phases |
| 4618 | #~(modify-phases %standard-phases |
| 4619 | (add-after 'unpack 'fix-cmake |
| 4620 | (lambda _ |
| 4621 | (substitute* "CMakeLists.txt" |
| 4622 | (("TARGET_INCLUDE_DIRECTORIES\\((pthreadpool|cpuinfo).*") "") |
| 4623 | (("AMD64") "x86_64")))) |
| 4624 | (add-after 'patch-source-shebangs 'generate-files |
| 4625 | (lambda _ |
| 4626 | ;; This script just calls two other scripts. |
| 4627 | (delete-file "scripts/generate-tests-and-benchmarks.sh") |
| 4628 | ;; The bash scripts run all jobs at once and then wait, so we |
| 4629 | ;; convert them to Makefiles. |
| 4630 | (for-each |
| 4631 | (lambda (name) |
| 4632 | (define counter 0) ; For the targets |
| 4633 | (define target-deps "") |
| 4634 | (when (and (string-prefix? "generate" name) |
| 4635 | (string-suffix? ".sh" name)) |
| 4636 | (let ((file (string-append "scripts/" name))) |
| 4637 | (substitute* file |
| 4638 | ;; Turn the commands into targets. Avoid comments and |
| 4639 | ;; lines starting with - (rest of multilines). |
| 4640 | (("\ |
| 4641 | ^[[:space:]]*([^ #-].*/.*(\\.sh|\\.py|-o |--output)[^&]*).*$" |
| 4642 | _ command) |
| 4643 | (begin |
| 4644 | (set! counter (+ counter 1)) |
| 4645 | (string-append "\ntarget" (number->string counter) |
| 4646 | ":" target-deps |
| 4647 | "\n\t" command))) |
| 4648 | ;; Remove trailing '&' characters. |
| 4649 | (("&?[[:space:]]*$") "\n") |
| 4650 | (("[[:space:]]*wait[[:space:]]*") "") |
| 4651 | ;; The commands after this line depend on the |
| 4652 | ;; previous commands in the file. |
| 4653 | (("JIT requires assembly files to be generated first.*" all) |
| 4654 | (begin |
| 4655 | (set! target-deps |
| 4656 | (string-append |
| 4657 | target-deps " target" |
| 4658 | (string-join |
| 4659 | (map number->string (iota counter 1)) " target"))) |
| 4660 | all))) |
| 4661 | (display (string-append "Running " name "\n")) |
| 4662 | (apply invoke "make" "-s" "-f" file "-j" |
| 4663 | (number->string (parallel-job-count)) |
| 4664 | (map |
| 4665 | (lambda (i) |
| 4666 | (string-append "target" (number->string i))) |
| 4667 | (iota counter 1)))))) |
| 4668 | (scandir "scripts")) |
| 4669 | ;; These need to run after the above scripts |
| 4670 | (display "Remaining files\n") |
| 4671 | (invoke "python3" "tools/update-microkernels.py") |
| 4672 | (invoke "python3" "tools/generate-lut-norm-test.py" |
| 4673 | "--spec" "test/u8-lut32norm.yaml" |
| 4674 | "--output" "test/u8-lut32norm.cc") |
| 4675 | ;; Check existence to avoid doubling the phase for r-torch. |
| 4676 | (when (file-exists? "test/qd8-f16-qb4w-gemm-minmax.yaml") |
| 4677 | (invoke "python3" "tools/generate-gemm-test.py" |
| 4678 | "--spec" "test/qd8-f16-qb4w-gemm-minmax.yaml" |
| 4679 | "--output-test" "test/qd8-f16-qb4w-gemm-minmax.cc")) |
| 4680 | (when (file-exists? "test/qd8-f32-qb4w-gemm-minmax.yaml") |
| 4681 | (invoke "python3" "tools/generate-gemm-test.py" |
| 4682 | "--spec" "test/qd8-f32-qb4w-gemm-minmax.yaml" |
| 4683 | "--output-test" "test/qd8-f32-qb4w-gemm-minmax.cc"))))))) |
| 4684 | (inputs |
| 4685 | (list clog |
| 4686 | cpuinfo |
| 4687 | pthreadpool |
| 4688 | googletest |
| 4689 | googlebenchmark |
| 4690 | fxdiv |
| 4691 | fp16 |
| 4692 | psimd)) |
| 4693 | (native-inputs |
| 4694 | (list python-pyyaml python-wrapper)) |
| 4695 | (synopsis "Optimized floating-point neural network inference operators") |
| 4696 | (description |
| 4697 | "XNNPACK is a highly optimized library of floating-point neural network |
| 4698 | inference operators for ARM, WebAssembly, and x86 platforms. XNNPACK is not |
| 4699 | intended for direct use by deep learning practitioners and researchers; |
| 4700 | instead it provides low-level performance primitives for accelerating |
| 4701 | high-level machine learning frameworks, such as TensorFlow Lite, |
| 4702 | TensorFlow.js, PyTorch, and MediaPipe.") |
| 4703 | (supported-systems |
| 4704 | '("armhf-linux" "aarch64-linux" "riscv64-linux" |
| 4705 | "i686-linux" "x86_64-linux")) |
| 4706 | (license license:bsd-3)))) |
| 4707 | |
| 4708 | ;; Warning: This package requires AVX2 or AVX-512 instructions. |
| 4709 | (define-public fbgemm |
| 4710 | (package |
| 4711 | (name "fbgemm") |
| 4712 | (version "1.5.0") |
| 4713 | (source (origin |
| 4714 | (method git-fetch) |
| 4715 | (uri (git-reference |
| 4716 | (url "https://github.com/pytorch/fbgemm") |
| 4717 | (commit (string-append "v" version)))) |
| 4718 | (file-name (git-file-name name version)) |
| 4719 | (sha256 |
| 4720 | (base32 |
| 4721 | "0dygcx42iijyhgpj5a8lams1avvc4d0cnvrbgzgxvpf3zx50n01a")) |
| 4722 | (patches (search-patches "fbgemm-use-system-libraries.patch")))) |
| 4723 | (build-system cmake-build-system) |
| 4724 | (arguments |
| 4725 | (list |
| 4726 | #:configure-flags |
| 4727 | ''("-DFBGEMM_LIBRARY_TYPE=shared") |
| 4728 | ;; Tests require AVX2 or AVX-512 instructions |
| 4729 | #:tests? #f)) |
| 4730 | (inputs (list cpuinfo)) |
| 4731 | (propagated-inputs (list asmjit-1.17)) |
| 4732 | (native-inputs (list pkg-config python googletest)) |
| 4733 | (home-page "https://github.com/pytorch/fbgemm") |
| 4734 | (synopsis "Facebook GEneral Matrix Multiplication") |
| 4735 | (description "Low-precision, high-performance matrix-matrix |
| 4736 | multiplications and convolution library for server-side inference.") |
| 4737 | (supported-systems '("x86_64-linux")) |
| 4738 | (license license:bsd-3))) |
| 4739 | |
| 4740 | (define-public tensorpipe |
| 4741 | (let ((commit "bb1473a4b38b18268e8693044afdb8635bc8351b") |
| 4742 | (revision "0")) |
| 4743 | (package |
| 4744 | (name "tensorpipe") |
| 4745 | (version (git-version "0" revision commit)) |
| 4746 | (source (origin |
| 4747 | (method git-fetch) |
| 4748 | (uri (git-reference |
| 4749 | (url "https://github.com/pytorch/tensorpipe") |
| 4750 | (commit commit))) |
| 4751 | (file-name (git-file-name name version)) |
| 4752 | (sha256 |
| 4753 | (base32 |
| 4754 | "0sbpkd69rzybw2j89sjkf4s0j8vkk96d51bsps28894989a75j6v")) |
| 4755 | (modules '((guix build utils))) |
| 4756 | (snippet |
| 4757 | '(delete-file-recursively "third_party")))) |
| 4758 | (build-system cmake-build-system) |
| 4759 | (arguments |
| 4760 | (list |
| 4761 | #:configure-flags |
| 4762 | ''("-DBUILD_SHARED_LIBS=ON") |
| 4763 | ;; There are no tests |
| 4764 | #:tests? #f)) |
| 4765 | (inputs (list libuv)) |
| 4766 | (native-inputs (list googletest pkg-config pybind11-2 libnop)) |
| 4767 | (home-page "https://github.com/pytorch/tensorpipe") |
| 4768 | (synopsis "Tensor-aware point-to-point communication primitive for |
| 4769 | machine learning") |
| 4770 | (description "TensorPipe provides a tensor-aware channel to transfer |
| 4771 | rich objects from one process to another while using the fastest transport for |
| 4772 | the tensors contained therein.") |
| 4773 | (license license:bsd-3)))) |
| 4774 | |
| 4775 | (define-public tvm |
| 4776 | ;; Include a bug fix post 0.19 release. |
| 4777 | (let ((commit "d3a2ed68a42f8b51d8ab1533b62e837b9ea74b8e") |
| 4778 | (revision "1")) |
| 4779 | (package |
| 4780 | (name "tvm") |
| 4781 | (version (git-version "0.20.dev0" revision commit)) |
| 4782 | (source |
| 4783 | (origin |
| 4784 | (method git-fetch) |
| 4785 | (uri (git-reference |
| 4786 | (url "https://github.com/apache/tvm") |
| 4787 | (commit commit))) |
| 4788 | (patches (search-patches "tvm_fix_cpptest_build.patch")) |
| 4789 | (file-name (git-file-name name version)) |
| 4790 | (sha256 |
| 4791 | (base32 "0ng52i0aydljnlrkyg069838c8b311fynhy5976w74smma2m7v72")))) |
| 4792 | (build-system cmake-build-system) |
| 4793 | (arguments |
| 4794 | (list |
| 4795 | #:configure-flags |
| 4796 | #~(list "-DUSE_OPENCL=ON" |
| 4797 | "-DUSE_VULKAN=ON" |
| 4798 | "-DUSE_OPENCL_ENABLE_HOST_PTR=ON" |
| 4799 | "-DINSTALL_DEV=ON" |
| 4800 | "-DUSE_GTEST=ON" |
| 4801 | "-DHIDE_PRIVATE_SYMBOLS=ON" ;per upstream build instructions |
| 4802 | "-DUSE_LLVM=llvm-config\\ --ignore-libllvm\\ --link-static" |
| 4803 | ;; per upstream build instructions |
| 4804 | (string-append "-DDLPACK_PATH=" |
| 4805 | (assoc-ref %build-inputs "dlpack") "/include") |
| 4806 | (string-append "-DDMLC_PATH=" |
| 4807 | (assoc-ref %build-inputs "dmlc-core") |
| 4808 | "/include") |
| 4809 | (string-append "-DRANG_PATH=" |
| 4810 | (assoc-ref %build-inputs "rang") "/include")) |
| 4811 | #:phases |
| 4812 | #~(modify-phases %standard-phases |
| 4813 | (replace 'check |
| 4814 | (lambda* (#:key source tests? #:allow-other-keys) |
| 4815 | (when tests? |
| 4816 | (begin |
| 4817 | (invoke "make" "-j" |
| 4818 | (number->string (parallel-job-count)) "cpptest") |
| 4819 | ;; Disable below the actual run of the tests because |
| 4820 | ;; several fail due to platform variations (for example, |
| 4821 | ;; fp16 tests fail because not supported on CPUs). |
| 4822 | ;; |
| 4823 | ;; (chdir "..") |
| 4824 | ;; (invoke (which "bash") |
| 4825 | ;; (string-append source |
| 4826 | ;; "/tests/scripts/task_cpp_unittest.sh")) |
| 4827 | ))))))) |
| 4828 | (inputs (list dmlc-core-next |
| 4829 | dlpack |
| 4830 | libedit |
| 4831 | libxml2 |
| 4832 | llvm-19 |
| 4833 | opencl-clhpp |
| 4834 | opencl-headers-2022 |
| 4835 | rang |
| 4836 | mesa |
| 4837 | mesa-opencl |
| 4838 | spirv-headers |
| 4839 | spirv-tools |
| 4840 | vulkan-headers ;TODO; now not building due to missing vta-hw |
| 4841 | vulkan-loader |
| 4842 | zlib |
| 4843 | zstd)) |
| 4844 | (native-inputs (list bash-minimal |
| 4845 | googletest |
| 4846 | (module-ref (resolve-interface '(gnu packages debug)) |
| 4847 | 'libbacktrace) |
| 4848 | pkg-config |
| 4849 | python |
| 4850 | python-cython)) |
| 4851 | (home-page "https://tvm.apache.org/") |
| 4852 | (synopsis |
| 4853 | "Machine learning compiler framework for CPUs, GPUs and accelerators") |
| 4854 | (description |
| 4855 | "Apache TVM is a compiler stack for deep learning systems. It is |
| 4856 | designed to close the gap between the productivity-focused deep learning |
| 4857 | frameworks, and the performance- and efficiency-focused hardware backends. |
| 4858 | TVM works with deep learning frameworks to provide end to end compilation to |
| 4859 | different backends") |
| 4860 | (license license:asl2.0)))) |
| 4861 | |
| 4862 | (define-public usearch |
| 4863 | (package |
| 4864 | (name "usearch") |
| 4865 | (version "2.26.2") |
| 4866 | (source |
| 4867 | (origin |
| 4868 | (method git-fetch) |
| 4869 | (uri (git-reference |
| 4870 | (url "https://github.com/unum-cloud/usearch") |
| 4871 | (commit (string-append "v" version)))) |
| 4872 | (file-name (git-file-name name version)) |
| 4873 | (sha256 |
| 4874 | (base32 |
| 4875 | "0j8nmqh9q7n15xwlpcs6zd8wr93yl4hn0kbf4v2fcwkkbj6khhzk")))) |
| 4876 | (build-system cmake-build-system) |
| 4877 | (arguments |
| 4878 | (list |
| 4879 | #:configure-flags |
| 4880 | #~(list "-DUSEARCH_INSTALL=ON" |
| 4881 | "-DUSEARCH_BUILD_BENCH_CPP=OFF"))) |
| 4882 | (inputs (list stringzilla)) |
| 4883 | (home-page "https://unum.cloud/usearch") |
| 4884 | (synopsis "Similarity search engine") |
| 4885 | (description "USearch is a free software search and clustering |
| 4886 | engine. Features include: |
| 4887 | @itemize |
| 4888 | @item 10x faster HNSW implementation than FAISS. |
| 4889 | @item Simple and extensible single C++11 header library. |
| 4890 | @item SIMD-optimized and user-defined metrics with JIT compilation. |
| 4891 | @item Hardware-agnostic bf16, e5m2, & i8 - half-precision & quarter-precision support. |
| 4892 | @item View large indexes from disk without loading into RAM. |
| 4893 | @item Heterogeneous lookups, renaming/relabeling, and on-the-fly deletions. |
| 4894 | @item Binary Tanimoto and Sorensen coefficients for Genomics and Chemistry applications. |
| 4895 | @item Space-efficient point-clouds with uint40_t, accommodating 4B+ size. |
| 4896 | @item Compatible with OpenMP and custom \"executors\" for fine-grained parallelism. |
| 4897 | @item Semantic Search and Joins. |
| 4898 | @item Near-real-time clustering and sub-clustering for Tens or Millions of clusters. |
| 4899 | @end itemize") |
| 4900 | (license license:asl2.0))) |
| 4901 | |
| 4902 | (define-public koboldcpp |
| 4903 | (package |
| 4904 | (name "koboldcpp") |
| 4905 | (version "1.122") |
| 4906 | (source |
| 4907 | (origin |
| 4908 | (method git-fetch) |
| 4909 | (uri (git-reference |
| 4910 | (url "https://github.com/LostRuins/koboldcpp") |
| 4911 | (commit (string-append "v" version)))) |
| 4912 | (file-name (git-file-name name version)) |
| 4913 | (snippet |
| 4914 | #~(begin |
| 4915 | (use-modules (guix build utils) |
| 4916 | (ice-9 ftw) |
| 4917 | (srfi srfi-26)) |
| 4918 | (delete-file-recursively "lib") |
| 4919 | (for-each delete-file (find-files "." "\\.(dll|exe|bat)$")))) |
| 4920 | (sha256 |
| 4921 | (base32 "184336v6n3a3l7gvirm230rkknn93i91710wpr6190sdg5rbmmgl")))) |
| 4922 | (build-system gnu-build-system) |
| 4923 | (arguments |
| 4924 | (list |
| 4925 | #:tests? #f ;no tests |
| 4926 | #:imported-modules %pyproject-build-system-modules |
| 4927 | #:modules '((guix build gnu-build-system) |
| 4928 | ((guix build pyproject-build-system) #:prefix python:) |
| 4929 | (guix build utils)) |
| 4930 | #:make-flags |
| 4931 | #~(list (string-append "CC=" #$(cc-for-target)) |
| 4932 | "LLAMA_VULKAN=1" |
| 4933 | "LLAMA_PORTABLE=1" |
| 4934 | "LLAMA_USE_BUNDLED_GLSLC=0") ;; TODO: add CLBLAS |
| 4935 | #:phases |
| 4936 | #~(modify-phases %standard-phases |
| 4937 | (delete 'configure) |
| 4938 | (replace 'install |
| 4939 | (lambda _ |
| 4940 | (for-each |
| 4941 | (lambda (f) |
| 4942 | (install-file f (string-append #$output "/bin"))) |
| 4943 | (append (list "koboldcpp.py" |
| 4944 | "vulkan-shaders-gen" |
| 4945 | "vulkan-shaders-gen-noext") |
| 4946 | ;; koboldcpp.py expects .so's |
| 4947 | ;; to be in the same directory |
| 4948 | (find-files "." "\\.so$"))))) |
| 4949 | (add-after 'install 'python-wrap |
| 4950 | (lambda* (#:key inputs outputs #:allow-other-keys) |
| 4951 | ;; Upstream removed the executable flag. |
| 4952 | (chmod (search-input-file outputs "bin/koboldcpp.py") #o744) |
| 4953 | (wrap-program (search-input-file outputs "bin/koboldcpp.py") |
| 4954 | `("GUIX_PYTHONPATH" = |
| 4955 | (,(getenv "GUIX_PYTHONPATH") |
| 4956 | ,(python:site-packages inputs outputs))))))))) |
| 4957 | (inputs |
| 4958 | (list bash-minimal |
| 4959 | python |
| 4960 | python-customtkinter |
| 4961 | spirv-headers |
| 4962 | spirv-tools |
| 4963 | vulkan-headers |
| 4964 | vulkan-loader)) ;TODO: add OpenCL |
| 4965 | (native-inputs |
| 4966 | (list shaderc)) |
| 4967 | (properties '((tunable? . #t))) |
| 4968 | (home-page "https://github.com/lostruins/koboldcpp") |
| 4969 | (synopsis "AI text-generation software for GGML and GGUF models") |
| 4970 | (description "KoboldCpp is an easy-to-use AI text-generation software for |
| 4971 | GGML and GGUF models, builds off llama.cpp and adds many additional features: |
| 4972 | |
| 4973 | @itemize |
| 4974 | @item Runs on CPU or GPU, supports full or partial offloaded |
| 4975 | @item LLM text generation (Supports all GGML and GGUF models, backwards |
| 4976 | compatibility with ALL past models) |
| 4977 | @item Image Generation (Stable Diffusion 1.5, SDXL, SD3, Flux) |
| 4978 | @item Speech-To-Text (Voice Recognition) via Whisper |
| 4979 | @item Text-To-Speech (Voice Generation) via OuteTTS, Kokoro, Parler and Dia |
| 4980 | @item Provides many compatible APIs endpoints for many popular webservices |
| 4981 | (KoboldCppApi OpenAiApi OllamaApi A1111ForgeApi ComfyUiApi |
| 4982 | WhisperTranscribeApi XttsApi OpenAiSpeechApi) |
| 4983 | @item Bundled KoboldAI Lite UI with editing tools, save formats, memory, |
| 4984 | world info, author's note, characters, scenarios |
| 4985 | @item Includes multiple modes (chat, adventure, instruct, storywriter) |
| 4986 | and UI Themes (aesthetic roleplay, classic writer, corporate assistant |
| 4987 | messsenger) |
| 4988 | @item Supports loading Tavern Character Cards, importing many different |
| 4989 | data formats from various sites, reading or exporting JSON savefiles |
| 4990 | and persistent stories |
| 4991 | @item Many other features including new samplers, regex support |
| 4992 | websearch, RAG via TextDB, image recognition/vision and more |
| 4993 | @end itemize |
| 4994 | |
| 4995 | All up-to-date GGUF models are supported, and KoboldCpp also includes backward |
| 4996 | compatibility for older versions/legacy GGML models.") |
| 4997 | (license (list license:agpl3 license:expat)))) |
| 4998 | |
| 4999 | |
| 5000 | (define-public foxi |
| 5001 | (let |
| 5002 | ((commit "c278588e34e535f0bb8f00df3880d26928038cad") |
| 5003 | (revision "0")) |
| 5004 | (package |
| 5005 | (name "foxi") |
| 5006 | (version (git-version "1.4.1" revision commit)) |
| 5007 | (source (origin |
| 5008 | (method git-fetch) |
| 5009 | (uri (git-reference |
| 5010 | (url "https://github.com/houseroad/foxi") |
| 5011 | (commit commit))) |
| 5012 | (file-name (git-file-name name version)) |
| 5013 | (sha256 |
| 5014 | (base32 |
| 5015 | "0q3ssm5hmmvwfwx87mnnajbavzgpahybw6rpn8ysr9r095dwgq5a")) |
| 5016 | (patches (search-patches "foxi-fix-build.patch")))) |
| 5017 | (build-system cmake-build-system) |
| 5018 | (arguments |
| 5019 | (list |
| 5020 | ;; No tests |
| 5021 | #:tests? #f)) |
| 5022 | (home-page "https://github.com/houseroad/foxi") |
| 5023 | (synopsis "ONNXIFI with Facebook Extension") |
| 5024 | (description "ONNX Interface for Framework Integration is a cross-platform |
| 5025 | API for loading and executing ONNX graphs on optimized backends. This package |
| 5026 | contains facebook extensions and is used by PyTorch.") |
| 5027 | (license license:expat)))) |
| 5028 | |
| 5029 | (define-public ideep-pytorch |
| 5030 | (package |
| 5031 | (name "ideep-pytorch") |
| 5032 | (version "3.7.1") |
| 5033 | (source |
| 5034 | (origin |
| 5035 | (method git-fetch) |
| 5036 | (uri (git-reference |
| 5037 | (url "https://github.com/intel/ideep") |
| 5038 | (commit (string-append "pytorch-rls-v" version)))) |
| 5039 | (file-name (git-file-name name version)) |
| 5040 | (sha256 |
| 5041 | (base32 |
| 5042 | "06h3q86k9jlb1ia6rc8adzqypab7x9dmy2gc49v80ylqxab6azar")))) |
| 5043 | (build-system copy-build-system) |
| 5044 | (arguments |
| 5045 | (list |
| 5046 | #:install-plan |
| 5047 | ''(("include" "include")))) |
| 5048 | (home-page "https://github.com/intel/ideep") |
| 5049 | (synopsis "Ideep headers for internal use by PyTorch") |
| 5050 | (description "This library is used internally as header-only library by |
| 5051 | PyTorch.") |
| 5052 | (license license:expat))) |
| 5053 | |
| 5054 | (define %python-pytorch-version "2.10.0") |
| 5055 | |
| 5056 | (define %python-pytorch-src |
| 5057 | (origin |
| 5058 | (method git-fetch) |
| 5059 | (uri (git-reference |
| 5060 | (url "https://github.com/pytorch/pytorch") |
| 5061 | (commit (string-append "v" %python-pytorch-version)))) |
| 5062 | (file-name (git-file-name "python-pytorch" %python-pytorch-version)) |
| 5063 | (sha256 |
| 5064 | (base32 |
| 5065 | "05pw5i1fm1i3grj8ksyrprq3an4a5pzn4ii1jkym00d0nabgbkyx")) |
| 5066 | (patches (search-patches "python-pytorch-system-libraries.patch" |
| 5067 | "python-pytorch-runpath.patch" |
| 5068 | "python-pytorch-without-kineto.patch" |
| 5069 | ;; Some autogeneration scripts depend on the |
| 5070 | ;; compile PyTorch library. Therefore, we create |
| 5071 | ;; dummy versions which are regenerated later. |
| 5072 | "python-pytorch-fix-codegen.patch")) |
| 5073 | (modules '((guix build utils))) |
| 5074 | (snippet |
| 5075 | '(begin |
| 5076 | ;; Bundled or unused code |
| 5077 | (for-each |
| 5078 | (lambda (dir) |
| 5079 | (when (file-exists? dir) |
| 5080 | (delete-file-recursively dir))) |
| 5081 | '("android" |
| 5082 | "aten/src/ATen/native/cuda/cutlass_extensions" |
| 5083 | "aten/src/ATen/native/quantized/cpu/qnnpack" |
| 5084 | "caffe2/mobile/contrib/libopencl-stub" |
| 5085 | "caffe2/mobile/contrib/libvulkan-stub" |
| 5086 | "third_party")) |
| 5087 | |
| 5088 | ;; Autogenerated files |
| 5089 | (for-each |
| 5090 | delete-file |
| 5091 | '("aten/src/ATen/nnapi/nnapi_wrapper.cpp" |
| 5092 | "aten/src/ATen/nnapi/nnapi_wrapper.h" |
| 5093 | ;; These files contain just lists of floating point values and |
| 5094 | ;; might be as well hand-written. |
| 5095 | ;; "test/cpp/api/init_baseline.h" |
| 5096 | ;; "test/cpp/api/optim_baseline.h" |
| 5097 | "test/mobile/test_upgrader_bytecode_table_example.cpp" |
| 5098 | "torch/csrc/jit/mobile/upgrader_mobile.cpp" |
| 5099 | "torch/csrc/jit/runtime/decomposition_registry_util.cpp" |
| 5100 | "torch/csrc/jit/runtime/serialized_shape_function_registry.cpp" |
| 5101 | "torch/csrc/jit/tensorexpr/external_functions_codegen.cpp" |
| 5102 | "torch/csrc/jit/serialization/mobile_bytecode_generated.h")) |
| 5103 | (delete-file-recursively ".github") |
| 5104 | (for-each |
| 5105 | (lambda (dir) |
| 5106 | (for-each |
| 5107 | delete-file |
| 5108 | (find-files dir "\\.cu$"))) |
| 5109 | '("aten/src/ATen/native/transformers/cuda/flash_attn" |
| 5110 | "aten/src/ATen/native/transformers/cuda/mem_eff_attention" |
| 5111 | "aten/src/ATen/native/transformers/hip/flash_attn")))))) |
| 5112 | |
| 5113 | (define-public qnnpack-pytorch |
| 5114 | (package |
| 5115 | (inherit qnnpack) |
| 5116 | (name "qnnpack-pytorch") |
| 5117 | (version (string-append "pytorch-" %python-pytorch-version)) |
| 5118 | (source |
| 5119 | (origin |
| 5120 | (inherit %python-pytorch-src) |
| 5121 | (patches '()) |
| 5122 | (modules '((guix build utils) |
| 5123 | (srfi srfi-26) |
| 5124 | (ice-9 ftw))) |
| 5125 | (snippet |
| 5126 | '(begin |
| 5127 | (rename-file "aten/src/ATen/native/quantized/cpu/qnnpack" |
| 5128 | "../qnnpack") |
| 5129 | (let ((outdir (getcwd))) |
| 5130 | (chdir "..") |
| 5131 | (rename-file outdir "dummy") |
| 5132 | (rename-file "qnnpack" outdir) |
| 5133 | (chdir outdir) |
| 5134 | (delete-file-recursively "deps")))))) |
| 5135 | (arguments |
| 5136 | (substitute-keyword-arguments arguments |
| 5137 | ((#:phases phases #~%standard-phases) |
| 5138 | #~(modify-phases %standard-phases |
| 5139 | (add-after 'unpack 'patch-cmake |
| 5140 | (lambda _ |
| 5141 | (substitute* "CMakeLists.txt" |
| 5142 | (("project\\(.*" orig) |
| 5143 | (apply |
| 5144 | string-append |
| 5145 | orig "\n" |
| 5146 | (map (lambda (name) |
| 5147 | (string-append |
| 5148 | "option(" name " \"\" ON)\n")) |
| 5149 | '("USE_SYSTEM_CPUINFO" "USE_SYSTEM_FP16" "USE_SYSTEM_FXDIV" |
| 5150 | "USE_SYSTEM_PSIMD" "USE_SYSTEM_PTHREADPOOL")))) |
| 5151 | (("if.*SOURCE_DIR.*") |
| 5152 | "if(FALSE)\n") |
| 5153 | (("if\\(NOT TARGET (clog|gtest|benchmark).*") |
| 5154 | "if(FALSE)\n") |
| 5155 | (("target_link_libraries.*(fxdiv|psimd|fp16)\\).*") |
| 5156 | "") |
| 5157 | (("(target_link_libraries.*) fp16 (.*)" _ before after) |
| 5158 | (string-append before " " after))))) |
| 5159 | (add-after 'unpack 'fix-cstring-include |
| 5160 | (lambda _ |
| 5161 | (substitute* "include/pack_block_sparse.h" |
| 5162 | (("#include.*<vector>.*" orig) |
| 5163 | (string-append orig "\n#include <cstring>\n"))))) |
| 5164 | (add-after 'install 'install-missing-headers |
| 5165 | (lambda _ |
| 5166 | (for-each |
| 5167 | (lambda (name) |
| 5168 | (install-file (string-append "../source/include/" name) |
| 5169 | (string-append #$output "/include"))) |
| 5170 | '("pack_block_sparse.h" |
| 5171 | "pytorch_qnnpack.h" |
| 5172 | "qnnpack_func.h")) |
| 5173 | (copy-recursively |
| 5174 | "../source/src/qnnpack" |
| 5175 | (string-append #$output "/include/qnnpack")))))) |
| 5176 | ;; Some tests occasionally fail on i686 due to floating point rounding. |
| 5177 | ((#:tests? _ #t) |
| 5178 | (not (string-prefix? "i686" (or (%current-target-system) |
| 5179 | (%current-system))))))))) |
| 5180 | |
| 5181 | ;; Please also update python-torchvision when updating this package. |
| 5182 | (define-public python-pytorch |
| 5183 | (package |
| 5184 | (name "python-pytorch") |
| 5185 | (version %python-pytorch-version) |
| 5186 | (source %python-pytorch-src) |
| 5187 | (build-system pyproject-build-system) |
| 5188 | (arguments |
| 5189 | (list |
| 5190 | ;; Even only the core tests take a very long time to run. |
| 5191 | #:tests? #f |
| 5192 | #:phases |
| 5193 | #~(modify-phases %standard-phases |
| 5194 | (add-after 'unpack 'patch-build-system |
| 5195 | (lambda _ |
| 5196 | (substitute* "pyproject.toml" |
| 5197 | (("setuptools\\.build_meta:__legacy__") |
| 5198 | "setuptools.build_meta")))) |
| 5199 | (add-after 'unpack 'cmake-patches |
| 5200 | (lambda* (#:key inputs outputs #:allow-other-keys) |
| 5201 | (substitute* "cmake/Dependencies.cmake" |
| 5202 | (("#POCKETFFT_INCLUDE_DIR") |
| 5203 | (string-append |
| 5204 | #$(this-package-native-input "pocketfft-cpp") "/include")) |
| 5205 | (("#FP16_INCLUDE_DIR") |
| 5206 | (string-append |
| 5207 | #$(this-package-input "fp16") "/include")) |
| 5208 | (("#CONCURRENTQUEUE_INCLUDE_DIR") |
| 5209 | (search-input-directory inputs "include/concurrentqueue")) |
| 5210 | ;; Disable opentelemetry |
| 5211 | ((".*(add_library|target_include_directories).*opentelemetry.*") |
| 5212 | "")) |
| 5213 | (substitute* "torch/CMakeLists.txt" |
| 5214 | ((".*opentelemetry.*") "")) |
| 5215 | ;; Fix Python install directory |
| 5216 | (substitute* "caffe2/CMakeLists.txt" |
| 5217 | (("\\$\\{Python_SITELIB\\}") |
| 5218 | (site-packages inputs outputs))) |
| 5219 | ;; Ensure httplib::httplib is defined when used. |
| 5220 | (substitute* "cmake/Caffe2Config.cmake.in" |
| 5221 | (("include *\\(.*Caffe2Targets\\.cmake.*\\)" all) |
| 5222 | (string-append "find_package(httplib REQUIRED)\n" all))))) |
| 5223 | ;; This entry point is broken, because it refers to a module that is |
| 5224 | ;; (intentionally) not installed |
| 5225 | ;; (https://github.com/pytorch/pytorch/pull/134729), which causes |
| 5226 | ;; the 'sanity-check phase to fail. |
| 5227 | (add-after 'unpack 'remove-fr-trace-script |
| 5228 | (lambda _ |
| 5229 | (substitute* "setup.py" |
| 5230 | (("entry_points\\[\"console_scripts\"\\]\\.append\\(") "(")))) |
| 5231 | (add-after 'remove-fr-trace-script 'skip-pip-redirect |
| 5232 | (lambda _ |
| 5233 | ;; Keep using setup.py directly instead of invoking pip. |
| 5234 | (substitute* "setup.py" |
| 5235 | (("if arg == \"install\":") |
| 5236 | "if False and arg == \"install\":")))) |
| 5237 | (add-before 'build 'use-system-libraries |
| 5238 | (lambda _ |
| 5239 | (for-each |
| 5240 | (lambda (file) |
| 5241 | ;; Check whether the files exist for the |
| 5242 | ;; python-pytorch-for-r-torch package |
| 5243 | (when (file-exists? file) |
| 5244 | (substitute* file |
| 5245 | (("\"miniz\\.h\"") "<miniz/miniz.h>") |
| 5246 | (("<miniz\\.h>") "<miniz/miniz.h>")))) |
| 5247 | '("caffe2/serialize/crc.cc" |
| 5248 | "caffe2/serialize/inline_container.cc" |
| 5249 | "torch/csrc/inductor/aoti_package/model_package_loader.cpp")) |
| 5250 | |
| 5251 | ;; Fix missing include for getCvarString/getCvarInt. |
| 5252 | (substitute* "torch/csrc/distributed/c10d/GlooDeviceFactory.cpp" |
| 5253 | (("#include <c10/util/Exception.h>" all) |
| 5254 | (string-append |
| 5255 | "#include <torch/csrc/distributed/c10d/Utils.hpp>\n\n" |
| 5256 | all))) |
| 5257 | |
| 5258 | ;; Fix moodycamel/concurrentqueue includes for system package |
| 5259 | (substitute* '("c10/util/Semaphore.h" |
| 5260 | "c10/test/util/Semaphore_test.cpp" |
| 5261 | "torch/nativert/executor/ParallelGraphExecutor.cpp") |
| 5262 | (("<moodycamel/concurrentqueue\\.h>") "<concurrentqueue.h>") |
| 5263 | (("<moodycamel/lightweightsemaphore\\.h>") |
| 5264 | "<lightweightsemaphore.h>")) |
| 5265 | |
| 5266 | (substitute* "aten/src/ATen/native/vulkan/api/Allocator.h" |
| 5267 | (("<include/vk_mem_alloc.h>") |
| 5268 | "<vk_mem_alloc.h>")) |
| 5269 | ;; Fix missing <algorithm> header for std::for_each in Vulkan API |
| 5270 | (substitute* "aten/src/ATen/native/vulkan/api/QueryPool.cpp" |
| 5271 | (("#include <utility>" all) |
| 5272 | (string-append all "\n#include <algorithm>"))) |
| 5273 | ;; For Vulkan |
| 5274 | (substitute* "CMakeLists.txt" |
| 5275 | (("append_cxx_flag.*-Werror=(return-type|range-loop-construct).*") "")) |
| 5276 | (substitute* |
| 5277 | (cons* |
| 5278 | "torch/csrc/Module.cpp" |
| 5279 | (map |
| 5280 | (lambda (name) |
| 5281 | (string-append |
| 5282 | "torch/utils/benchmark/utils/valgrind_wrapper/" |
| 5283 | name)) |
| 5284 | '("compat_bindings.cpp" "timer_callgrind_template.cpp"))) |
| 5285 | (("<callgrind.h>") "<valgrind/callgrind.h>")) |
| 5286 | (setenv "USE_VULKAN" "1") |
| 5287 | ;; Tell 'setup.py' to let 'CMakeLists.txt' know that we |
| 5288 | ;; want to use "system libraries" instead of the bundled |
| 5289 | ;; ones. |
| 5290 | (setenv "USE_SYSTEM_LIBS" "1") |
| 5291 | ;; For oneDNN |
| 5292 | (setenv "USE_MKLDNN" "1") |
| 5293 | ;; Only works with CUPTI |
| 5294 | (setenv "USE_KINETO" "0") |
| 5295 | ;; Prevent CMake error by disabling explicitely |
| 5296 | (setenv "USE_ITT" "0") |
| 5297 | ;; Disable on unsupported systems |
| 5298 | (if #$(not (member |
| 5299 | (or (%current-target-system) |
| 5300 | (%current-system)) |
| 5301 | (package-transitive-supported-systems qnnpack))) |
| 5302 | (setenv "USE_QNNPACK" "0")) |
| 5303 | (substitute* '("requirements.txt" "setup.py") |
| 5304 | (("sympy>=1\\.13\\.3") |
| 5305 | "sympy>=1.13.1")) |
| 5306 | ;; Avoid ModuleNotFoundError while preserving setup.py flow. |
| 5307 | (substitute* "setup.py" |
| 5308 | (("from build_bundled import create_bundled") |
| 5309 | "create_bundled = lambda *args, **kwargs: None")))) |
| 5310 | (add-after 'use-system-libraries 'skip-nccl-call |
| 5311 | (lambda _ |
| 5312 | ;; Comment-out `checkout_nccl()` invokation in build_pytorch(). |
| 5313 | (substitute* "tools/build_pytorch_libs.py" |
| 5314 | (("^[[:blank:]]*checkout_nccl\\(\\)" all) |
| 5315 | (string-append "# " all "\n pass"))))) |
| 5316 | ;; PyTorch is still built with AVX2 and AVX-512 support selected at |
| 5317 | ;; runtime, but these dependencies require it (nnpack only for |
| 5318 | ;; x86_64). |
| 5319 | (add-before 'build 'disable-avx-dependencies |
| 5320 | (lambda _ |
| 5321 | (setenv "USE_FBGEMM" "0") |
| 5322 | (if #$(not |
| 5323 | (member (or (%current-target-system) |
| 5324 | (%current-system)) |
| 5325 | '("armhf-linux" "aarch64-linux"))) |
| 5326 | (setenv "USE_NNPACK" "0")))) |
| 5327 | (add-after 'use-system-libraries 'set-max-jobs |
| 5328 | (lambda _ |
| 5329 | (setenv "MAX_JOBS" (number->string (parallel-job-count))))) |
| 5330 | (add-after 'set-max-jobs 'codegen1 |
| 5331 | (lambda _ |
| 5332 | (with-directory-excursion "torch/csrc/jit/tensorexpr" |
| 5333 | (setenv "PYTHONPATH" "../../../..") |
| 5334 | (invoke "python3" "codegen_external.py") |
| 5335 | (setenv "PYTHONPATH" #f)) |
| 5336 | |
| 5337 | (invoke "python3" "aten/src/ATen/nnapi/codegen.py") |
| 5338 | |
| 5339 | (invoke "bash" "tools/gen_flatbuffers.sh") |
| 5340 | |
| 5341 | ;; Generate dummy files as the generation depends on the compiled |
| 5342 | ;; library. They are regenerated later. |
| 5343 | (setenv "PYTHONPATH" ".") |
| 5344 | (invoke "python3" |
| 5345 | "torchgen/operator_versions/gen_mobile_upgraders.py" |
| 5346 | "dummy") |
| 5347 | (setenv "PYTHONPATH" #f) |
| 5348 | |
| 5349 | (invoke "python3" |
| 5350 | "torchgen/shape_functions/gen_jit_shape_functions.py" |
| 5351 | "dummy") |
| 5352 | |
| 5353 | (invoke "python3" |
| 5354 | "torchgen/decompositions/gen_jit_decompositions.py" |
| 5355 | "dummy"))) |
| 5356 | ;; Properly generate autogenerated files ... |
| 5357 | (add-after 'install 'codegen2 |
| 5358 | (lambda* (#:key inputs outputs #:allow-other-keys) |
| 5359 | (add-installed-pythonpath inputs outputs) |
| 5360 | (invoke "python3" |
| 5361 | "torchgen/operator_versions/gen_mobile_upgraders.py") |
| 5362 | (invoke "python3" |
| 5363 | "torchgen/shape_functions/gen_jit_shape_functions.py") |
| 5364 | (invoke "python3" |
| 5365 | "torchgen/decompositions/gen_jit_decompositions.py"))) |
| 5366 | ;; ... rebuild their dependencies ... |
| 5367 | (add-after 'codegen2 'build2 |
| 5368 | (lambda _ |
| 5369 | (invoke "python3" "setup.py" "build"))) |
| 5370 | ;; ... and install again. |
| 5371 | (add-after 'build2 'install2 |
| 5372 | (lambda _ |
| 5373 | (invoke "python3" "setup.py" "install" (string-append "--prefix=" #$output) |
| 5374 | "--no-compile" "--single-version-externally-managed" "--root=/") |
| 5375 | (invoke "python" "-m" "compileall" |
| 5376 | "--invalidation-mode=unchecked-hash" #$output))) |
| 5377 | (replace 'check |
| 5378 | (lambda* (#:key tests? #:allow-other-keys) |
| 5379 | ;; Run the test suite following the instructions in |
| 5380 | ;; 'CONTRIBUTING.md'. Unfortunately this doesn't work, unless |
| 5381 | ;; you set PYTHONPATH or GUIX_PYTHONPATH, but this is done in |
| 5382 | ;; the codegen2 phase already. |
| 5383 | (when tests? |
| 5384 | (invoke "python3" "test/run_test.py" "--core")))) |
| 5385 | (add-after 'install2 'remove-test-executables |
| 5386 | (lambda* (#:key inputs outputs #:allow-other-keys) |
| 5387 | ;; Remove test executables, but keep other executables |
| 5388 | ;; such as 'torch_shm_manager' and and .so files such as |
| 5389 | ;; 'libtorch_global_deps.so'. |
| 5390 | (let ((python-site (site-packages inputs outputs))) |
| 5391 | (for-each delete-file |
| 5392 | (find-files python-site |
| 5393 | "(^test_cpp_rpc|_test)$"))))) |
| 5394 | (add-after 'install2 'remove-caffe2-onnx-scripts |
| 5395 | (lambda _ |
| 5396 | (let ((bin (string-append #$output "/bin"))) |
| 5397 | ;; Remove 'convert-caffe2-to-onnx' and |
| 5398 | ;; 'convert-onnx-to-caffe2': they seem to be |
| 5399 | ;; deprecated and they cause a failure of the |
| 5400 | ;; 'sanity-check' phase: |
| 5401 | ;; |
| 5402 | ;; ImportError: cannot import name 'metanet_pb2' from |
| 5403 | ;; partially initialized module 'caffe2.proto' (most likely |
| 5404 | ;; due to a circular import) |
| 5405 | (for-each delete-file |
| 5406 | (find-files bin "^convert-.*caffe2")) |
| 5407 | |
| 5408 | (substitute* (find-files #$output "^entry_points\\.txt$") |
| 5409 | (("^convert-.*" all) |
| 5410 | (string-append "# " all "\n"))))))))) |
| 5411 | (native-inputs |
| 5412 | (list cmake-minimal |
| 5413 | doxygen |
| 5414 | ideep-pytorch |
| 5415 | ninja |
| 5416 | nlohmann-json |
| 5417 | pocketfft-cpp |
| 5418 | python-expecttest |
| 5419 | python-pytest-flakefinder |
| 5420 | python-pytest-rerunfailures |
| 5421 | python-pytest-shard |
| 5422 | python-pytest-xdist |
| 5423 | python-hypothesis |
| 5424 | python-setuptools |
| 5425 | python-types-dataclasses |
| 5426 | shaderc |
| 5427 | valgrind/pinned)) |
| 5428 | (inputs |
| 5429 | (append |
| 5430 | (list brotli ; for cpp-httplib |
| 5431 | clog |
| 5432 | concurrentqueue |
| 5433 | cpp-httplib |
| 5434 | eigen |
| 5435 | flatbuffers |
| 5436 | fmt |
| 5437 | fp16 |
| 5438 | fxdiv |
| 5439 | gemmlowp |
| 5440 | gloo |
| 5441 | googletest-1.12 |
| 5442 | googlebenchmark |
| 5443 | libuv |
| 5444 | miniz-for-pytorch |
| 5445 | onednn |
| 5446 | openblas |
| 5447 | openmpi |
| 5448 | openssl ; for cpp-httplib |
| 5449 | pthreadpool |
| 5450 | protobuf |
| 5451 | pybind11-2 |
| 5452 | ;; qnnpack |
| 5453 | qnnpack-pytorch |
| 5454 | rdma-core |
| 5455 | sleef |
| 5456 | tensorpipe |
| 5457 | vulkan-headers |
| 5458 | vulkan-loader |
| 5459 | vulkan-memory-allocator |
| 5460 | xnnpack |
| 5461 | zlib ; for cpp-httplib |
| 5462 | zstd) |
| 5463 | ;; nnpack requires AVX2 for x86_64-linux |
| 5464 | (if (equal? (or (%current-target-system) |
| 5465 | (%current-system)) |
| 5466 | '("aarch64-linux")) |
| 5467 | (list nnpack) |
| 5468 | '()))) |
| 5469 | (propagated-inputs |
| 5470 | (append |
| 5471 | (list cpuinfo |
| 5472 | onnx ;propagated for its Python modules |
| 5473 | onnx-optimizer |
| 5474 | python-astunparse |
| 5475 | python-click |
| 5476 | python-filelock |
| 5477 | python-fsspec |
| 5478 | python-future |
| 5479 | python-jinja2 |
| 5480 | python-networkx |
| 5481 | python-numpy |
| 5482 | python-opt-einsum |
| 5483 | python-optree |
| 5484 | python-packaging |
| 5485 | python-psutil |
| 5486 | python-pyyaml |
| 5487 | python-requests |
| 5488 | python-sympy |
| 5489 | python-typing-extensions))) |
| 5490 | (home-page "https://pytorch.org/") |
| 5491 | (synopsis "Python library for tensor computation and deep neural networks") |
| 5492 | ;; TODO: Support other 64-bit systems. |
| 5493 | (supported-systems '("x86_64-linux" "aarch64-linux")) |
| 5494 | (description |
| 5495 | "PyTorch is a Python package that provides two high-level features: |
| 5496 | |
| 5497 | @itemize |
| 5498 | @item tensor computation (like NumPy) with strong GPU acceleration; |
| 5499 | @item deep neural networks (DNNs) built on a tape-based autograd system. |
| 5500 | @end itemize |
| 5501 | |
| 5502 | You can reuse Python packages such as NumPy, SciPy, and Cython to extend |
| 5503 | PyTorch when needed. |
| 5504 | |
| 5505 | Note: currently this package does not provide GPU support.") |
| 5506 | (license license:bsd-3))) |
| 5507 | |
| 5508 | (define-public python-torchaudio |
| 5509 | (package |
| 5510 | (name "python-torchaudio") |
| 5511 | (version %python-pytorch-version) |
| 5512 | (source |
| 5513 | (origin |
| 5514 | (method git-fetch) |
| 5515 | (uri (git-reference |
| 5516 | (url "https://github.com/pytorch/audio") |
| 5517 | (commit (string-append "v" version)))) |
| 5518 | (file-name (git-file-name name version)) |
| 5519 | (sha256 |
| 5520 | (base32 "1j2h722cvvwa4g4vq8gw9x214ara6qimjmyvdm1x2x2pa4fj6nvg")))) |
| 5521 | (build-system pyproject-build-system) |
| 5522 | (arguments |
| 5523 | (list |
| 5524 | #:test-flags |
| 5525 | #~(list |
| 5526 | ;; XXX: Those packages require additional inputs. |
| 5527 | "--ignore=\ |
| 5528 | test/torchaudio_unittest/prototype/hifi_gan/hifi_gan_cpu_test.py" |
| 5529 | "--ignore=\ |
| 5530 | test/torchaudio_unittest/prototype/hifi_gan/hifi_gan_gpu_test.py" |
| 5531 | "--ignore=test/torchaudio_unittest/kaldi_io_test.py" |
| 5532 | "--ignore=test/torchaudio_unittest/test_load_save_torchcodec.py" |
| 5533 | "--ignore-glob=test/torchaudio_unittest/backend/dispatcher/ffmpeg" |
| 5534 | ;; Dataset tests require torchcodec (used by 'torchaudio.load'). |
| 5535 | "--ignore-glob=test/torchaudio_unittest/datasets" |
| 5536 | ;; XXX: We don't really want to run those tests. |
| 5537 | "--ignore-glob=test/torchaudio_unittest/example" |
| 5538 | ;; XXX: Integration tests require additional data / models. |
| 5539 | "--ignore-glob=test/integration_tests" |
| 5540 | ;; XXX: Those tests are very costly. |
| 5541 | "--ignore-glob=test/torchaudio_unittest/functional" |
| 5542 | "--ignore-glob=test/torchaudio_unittest/models" |
| 5543 | "--ignore=test/torchaudio_unittest/models/models_test.py" |
| 5544 | "--ignore=test/torchaudio_unittest/transforms/autograd_cpu_test.py" |
| 5545 | "-k" (string-append |
| 5546 | "not test_torchscript_fails" ; requires BUILD_SOX=1 |
| 5547 | ;; XXX: Unmatching harmless warning message. |
| 5548 | " and not test_unknown_subtype_warning" |
| 5549 | ;; Skip TorchScript consistency tests that fail with 2.9.0 |
| 5550 | " and not test_FrequencyMasking" |
| 5551 | " and not test_TimeMasking" |
| 5552 | " and not test_deemphasis" |
| 5553 | " and not test_rnnt_loss")) |
| 5554 | #:phases |
| 5555 | #~(modify-phases %standard-phases |
| 5556 | (add-before 'build 'configure |
| 5557 | (lambda* (#:key inputs outputs #:allow-other-keys) |
| 5558 | (setenv "BUILD_SOX" "0") |
| 5559 | (setenv "USE_FFMPEG" "0") |
| 5560 | (substitute* "tools/setup_helpers/extension.py" |
| 5561 | (("^( *)\"-DCMAKE_VERBOSE_MAKEFILE=ON\"," all blank) |
| 5562 | (string-append |
| 5563 | blank "f\"-DCMAKE_INSTALL_RPATH=" |
| 5564 | (site-packages inputs |
| 5565 | `(("out" . |
| 5566 | ,(assoc-ref inputs "python-pytorch")))) |
| 5567 | "/torch/lib;" |
| 5568 | (site-packages inputs outputs) "/torchaudio/lib\",\n" |
| 5569 | blank "\"-DCMAKE_BUILD_WITH_INSTALL_RPATH=ON\",\n" |
| 5570 | all)))))))) |
| 5571 | (native-inputs |
| 5572 | (list cmake |
| 5573 | ninja |
| 5574 | onednn |
| 5575 | python-expecttest |
| 5576 | python-parameterized |
| 5577 | python-pytest |
| 5578 | python-scipy |
| 5579 | python-setuptools |
| 5580 | python-wheel |
| 5581 | sox)) |
| 5582 | (inputs (package-inputs python-pytorch)) |
| 5583 | (propagated-inputs (list python-soundfile python-pytorch)) |
| 5584 | (home-page "https://pytorch.org/audio") |
| 5585 | (synopsis |
| 5586 | "Data manipulation and transformation for audio signal processing") |
| 5587 | (description |
| 5588 | "This package provides a machine learning library of popular datasets, |
| 5589 | model architectures, and common transformations to apply @code{python-pytorch} |
| 5590 | in the audio domain.") |
| 5591 | (license license:bsd-2))) |
| 5592 | |
| 5593 | ;; This package variant includes the dependencies requiring at least AVX2 or |
| 5594 | ;; AVX-512. |
| 5595 | (define-public python-pytorch-avx |
| 5596 | (package/inherit python-pytorch |
| 5597 | (name "python-pytorch-avx") |
| 5598 | (inputs |
| 5599 | (modify-inputs inputs |
| 5600 | (append fbgemm nnpack))) |
| 5601 | (arguments |
| 5602 | (substitute-keyword-arguments arguments |
| 5603 | ((#:phases phases) |
| 5604 | #~(modify-phases #$phases |
| 5605 | (delete 'disable-avx-dependencies))))) |
| 5606 | (supported-systems '("x86_64-linux")))) |
| 5607 | |
| 5608 | (define %python-pytorch-for-r-torch-version "2.7.1") |
| 5609 | |
| 5610 | (define %python-pytorch-for-r-torch-src |
| 5611 | (origin |
| 5612 | (inherit %python-pytorch-src) |
| 5613 | (uri (git-reference |
| 5614 | (url "https://github.com/pytorch/pytorch") |
| 5615 | (commit (string-append "v" %python-pytorch-for-r-torch-version)))) |
| 5616 | (file-name (git-file-name "python-pytorch" |
| 5617 | %python-pytorch-for-r-torch-version)) |
| 5618 | (sha256 |
| 5619 | (base32 |
| 5620 | "0734kfm66hsqdzgs2s4wj5yagvifijbgb0c5wfmp3qcdrraa9x57")) |
| 5621 | (patches (search-patches |
| 5622 | "python-pytorch-for-r-torch-system-libraries.patch" |
| 5623 | "python-pytorch-runpath.patch" |
| 5624 | "python-pytorch-for-r-torch-without-kineto.patch" |
| 5625 | ;; Some autogeneration scripts depend on the |
| 5626 | ;; compile PyTorch library. Therefore, we create |
| 5627 | ;; dummy versions which are regenerated later. |
| 5628 | "python-pytorch-for-r-torch-fix-codegen.patch")))) |
| 5629 | |
| 5630 | ;; Keep in sync with r-torch |
| 5631 | (define-public python-pytorch-for-r-torch |
| 5632 | (package |
| 5633 | (inherit python-pytorch) |
| 5634 | (name "python-pytorch") |
| 5635 | (version %python-pytorch-for-r-torch-version) |
| 5636 | (source %python-pytorch-for-r-torch-src) |
| 5637 | (build-system pyproject-build-system) |
| 5638 | (inputs |
| 5639 | (modify-inputs inputs |
| 5640 | (replace "gloo" gloo-for-r-torch))) |
| 5641 | (arguments |
| 5642 | (substitute-keyword-arguments arguments |
| 5643 | ((#:phases phases) |
| 5644 | #~(modify-phases #$phases |
| 5645 | (replace 'use-system-libraries |
| 5646 | (lambda _ |
| 5647 | (for-each |
| 5648 | (lambda (file) |
| 5649 | ;; Check whether the files exist for the |
| 5650 | ;; python-pytorch-for-r-torch package |
| 5651 | (when (file-exists? file) |
| 5652 | (substitute* file |
| 5653 | (("\"miniz\\.h\"") "<miniz/miniz.h>") |
| 5654 | (("<miniz\\.h>") "<miniz/miniz.h>")))) |
| 5655 | '("caffe2/serialize/crc.cc" |
| 5656 | "caffe2/serialize/inline_container.cc" |
| 5657 | "torch/csrc/inductor/aoti_package/model_package_loader.cpp")) |
| 5658 | |
| 5659 | (substitute* "aten/src/ATen/native/vulkan/api/Allocator.h" |
| 5660 | (("<include/vk_mem_alloc.h>") |
| 5661 | "<vk_mem_alloc.h>")) |
| 5662 | ;; Fix missing <algorithm> header for std::for_each in Vulkan API |
| 5663 | (substitute* "aten/src/ATen/native/vulkan/api/QueryPool.cpp" |
| 5664 | (("#include <utility>" all) |
| 5665 | (string-append all "\n#include <algorithm>"))) |
| 5666 | ;; For Vulkan |
| 5667 | (substitute* "CMakeLists.txt" |
| 5668 | (("append_cxx_flag.*-Werror=(return-type|range-loop-construct).*") "")) |
| 5669 | ;; In this variant functorch is installed directly under |
| 5670 | ;; site-packages/functorch, so use a direct rpath to |
| 5671 | ;; site-packages/torch/lib. |
| 5672 | (substitute* "functorch/CMakeLists.txt" |
| 5673 | (("set_target_properties\\(\\$\\{PROJECT_NAME\\} PROPERTIES INSTALL_RPATH \"\\$\\{_rpath_portable_origin\\}/\\.\\./torch/lib\"\\)") |
| 5674 | "set_target_properties(${PROJECT_NAME} PROPERTIES INSTALL_RPATH \"$ORIGIN/../torch/lib\")")) |
| 5675 | (substitute* |
| 5676 | (cons* |
| 5677 | "torch/csrc/Module.cpp" |
| 5678 | (map |
| 5679 | (lambda (name) |
| 5680 | (string-append |
| 5681 | "torch/utils/benchmark/utils/valgrind_wrapper/" |
| 5682 | name)) |
| 5683 | '("compat_bindings.cpp" "timer_callgrind_template.cpp"))) |
| 5684 | (("<callgrind.h>") "<valgrind/callgrind.h>")) |
| 5685 | (setenv "USE_VULKAN" "1") |
| 5686 | ;; Tell 'setup.py' to let 'CMakeLists.txt' know that we |
| 5687 | ;; want to use "system libraries" instead of the bundled |
| 5688 | ;; ones. |
| 5689 | (setenv "USE_SYSTEM_LIBS" "1") |
| 5690 | ;; For oneDNN |
| 5691 | (setenv "USE_MKLDNN" "1") |
| 5692 | ;; Only works with CUPTI |
| 5693 | (setenv "USE_KINETO" "0") |
| 5694 | ;; Prevent CMake error by disabling explicitely |
| 5695 | (setenv "USE_ITT" "0") |
| 5696 | ;; Disable on unsupported systems |
| 5697 | (if #$(not (member |
| 5698 | (or (%current-target-system) |
| 5699 | (%current-system)) |
| 5700 | (package-transitive-supported-systems qnnpack))) |
| 5701 | (setenv "USE_QNNPACK" "0")) |
| 5702 | (substitute* '("requirements.txt" "setup.py") |
| 5703 | (("sympy>=1\\.13\\.3") |
| 5704 | "sympy>=1.13.1")) |
| 5705 | ;; Avoid ModuleNotFoundError while preserving setup.py flow. |
| 5706 | (substitute* "setup.py" |
| 5707 | (("from build_bundled import create_bundled") |
| 5708 | "create_bundled = lambda *args, **kwargs: None")))) |
| 5709 | (replace 'skip-nccl-call |
| 5710 | (lambda _ |
| 5711 | ;; Comment-out `checkout_nccl()` invokation in build_pytorch(). |
| 5712 | (substitute* "tools/build_pytorch_libs.py" |
| 5713 | (("^[[:blank:]]*checkout_nccl\\(\\)" all) |
| 5714 | (string-append "# " all "\n pass"))))))))))) |
| 5715 | |
| 5716 | (define-public python-pytorch-geometric |
| 5717 | (package |
| 5718 | (name "python-pytorch-geometric") |
| 5719 | (version "2.7.0") |
| 5720 | (source (origin |
| 5721 | (method git-fetch) |
| 5722 | (uri (git-reference |
| 5723 | (url "https://github.com/pyg-team/pytorch_geometric/") |
| 5724 | (commit version))) |
| 5725 | (file-name (git-file-name name version)) |
| 5726 | (sha256 |
| 5727 | (base32 |
| 5728 | "0w6gpw0wf9w6h317mbqcf3iidx9yqjba34348p44g80ihskb6ly6")))) |
| 5729 | (build-system pyproject-build-system) |
| 5730 | (arguments |
| 5731 | (list |
| 5732 | #:test-flags |
| 5733 | '(list ;; PyTorch 2.10 raises deprecation warnings during import. |
| 5734 | "-W" "ignore::DeprecationWarning" |
| 5735 | ;; Hangs with AttributeError: 'NoneType' object has no attribute 'rpc_async' |
| 5736 | "--ignore=test/distributed/test_rpc.py" |
| 5737 | ;; A message passing jinja template is missing |
| 5738 | "--ignore=test/nn/conv/test_message_passing.py" |
| 5739 | "--ignore=test/nn/test_sequential.py" |
| 5740 | "--ignore=test/nn/models/test_basic_gnn.py" |
| 5741 | ;; These all fail with a size mismatch error such as |
| 5742 | ;; RuntimeError: shape '[-1, 2, 1, 1]' is invalid for input of size 3 |
| 5743 | "--ignore=test/explain/algorithm/test_captum_explainer.py" |
| 5744 | ;; Requires the nonfree MKL on CPU. |
| 5745 | "--ignore=test/nn/models/test_graph_unet.py" |
| 5746 | "-k" (string-append |
| 5747 | ;; Permissions error |
| 5748 | "not test_packaging" |
| 5749 | ;; Unknown multiprocessing failure |
| 5750 | " and not test_data_share_memory" |
| 5751 | ;; This can fail due to accuracy problems |
| 5752 | " and not test_gdc" |
| 5753 | ;; These refuse to be run on CPU and really want a GPU |
| 5754 | " and not test_add_random_walk_pe" |
| 5755 | " and not test_asap" |
| 5756 | " and not test_two_hop" |
| 5757 | ;; Failed when switched to [email protected] |
| 5758 | ;; typing module internals |
| 5759 | " and not test_type_repr")) |
| 5760 | #:phases |
| 5761 | '(modify-phases %standard-phases |
| 5762 | (add-after 'unpack 'delete-top-level-directories |
| 5763 | (lambda _ |
| 5764 | ;; The presence of these directories confuses the pyproject build |
| 5765 | ;; system. |
| 5766 | (for-each delete-file-recursively |
| 5767 | '("docker" "graphgym"))))))) |
| 5768 | (propagated-inputs |
| 5769 | (list onnx |
| 5770 | python-captum |
| 5771 | python-graphviz |
| 5772 | python-h5py |
| 5773 | python-jinja2 |
| 5774 | python-matplotlib |
| 5775 | python-networkx |
| 5776 | python-numba |
| 5777 | python-numpy |
| 5778 | python-opt-einsum |
| 5779 | python-pandas |
| 5780 | python-protobuf |
| 5781 | python-psutil |
| 5782 | python-pyparsing |
| 5783 | python-pytorch-lightning |
| 5784 | python-rdflib |
| 5785 | python-requests |
| 5786 | python-scikit-image |
| 5787 | python-scikit-learn |
| 5788 | python-scipy |
| 5789 | python-statsmodels |
| 5790 | python-sympy |
| 5791 | python-tabulate |
| 5792 | python-torchmetrics |
| 5793 | python-tqdm |
| 5794 | python-xxhash)) |
| 5795 | (native-inputs |
| 5796 | (list openssl python-flit-core python-pytest)) |
| 5797 | (home-page "https://pyg.org") |
| 5798 | (synopsis "Graph Neural Network library for PyTorch") |
| 5799 | (description |
| 5800 | "PyG is a library built upon PyTorch to easily write and train Graph |
| 5801 | Neural Networks for a wide range of applications related to structured data.") |
| 5802 | (license license:expat))) |
| 5803 | |
| 5804 | (define-public python-lightning-cloud |
| 5805 | (package |
| 5806 | (name "python-lightning-cloud") |
| 5807 | (version "0.6.0") |
| 5808 | (source (origin |
| 5809 | (method url-fetch) |
| 5810 | (uri (pypi-uri "lightning_cloud" version)) |
| 5811 | (sha256 |
| 5812 | (base32 |
| 5813 | "1g3zwn8r33gkmx529mvyddbrx0wqwf6125qimj4xfchf53g0qkn4")))) |
| 5814 | (arguments (list #:tests? #f)) ; no tests in PyPI archive. |
| 5815 | (build-system pyproject-build-system) |
| 5816 | (propagated-inputs (list python-boto3 |
| 5817 | python-click |
| 5818 | python-fastapi |
| 5819 | python-multipart |
| 5820 | python-protobuf |
| 5821 | python-pyjwt |
| 5822 | python-requests |
| 5823 | python-rich |
| 5824 | python-six |
| 5825 | python-urllib3 |
| 5826 | python-uvicorn |
| 5827 | python-websocket-client)) |
| 5828 | (native-inputs (list python-setuptools)) |
| 5829 | (home-page "https://lightning.ai") |
| 5830 | (synopsis "Lightning Cloud command line client") |
| 5831 | (description "This package provides a command line interface for Lightning |
| 5832 | AI services.") |
| 5833 | (license license:asl2.0))) |
| 5834 | |
| 5835 | (define-public python-lightning-utilities |
| 5836 | (package |
| 5837 | (name "python-lightning-utilities") |
| 5838 | (version "0.15.3") |
| 5839 | (source |
| 5840 | (origin |
| 5841 | (method git-fetch) |
| 5842 | (uri (git-reference |
| 5843 | (url "https://github.com/Lightning-AI/utilities") |
| 5844 | (commit (string-append "v" version)))) |
| 5845 | (file-name (git-file-name name version)) |
| 5846 | (sha256 |
| 5847 | (base32 "0krdvabdi1lcix5jvjjrrqyh6ibjwabc0g724mpi32gsz6g7ppwg")))) |
| 5848 | (build-system pyproject-build-system) |
| 5849 | (arguments |
| 5850 | (list #:tests? #f)) ;XXX: cycles with python-pytorch-lightning |
| 5851 | (native-inputs |
| 5852 | (list python-setuptools)) |
| 5853 | (propagated-inputs |
| 5854 | (list python-packaging |
| 5855 | python-typing-extensions)) |
| 5856 | (home-page "https://github.com/Lightning-AI/utilities") |
| 5857 | (synopsis "PyTorch Lightning sample project") |
| 5858 | (description "This package provides common Python utilities and GitHub |
| 5859 | Actions for the Lightning suite of libraries.") |
| 5860 | (license license:asl2.0))) |
| 5861 | |
| 5862 | (define-public python-captum |
| 5863 | (package |
| 5864 | (name "python-captum") |
| 5865 | ;; The latest commit provides compatability with NumPy 2. |
| 5866 | (properties '((commit . "4fade3af8dc398c1f96b193c969565aad23aedf5") |
| 5867 | (revision . "0"))) |
| 5868 | (version (git-version "0.8.0" |
| 5869 | (assoc-ref properties 'revision) |
| 5870 | (assoc-ref properties 'commit))) |
| 5871 | (source (origin |
| 5872 | (method git-fetch) |
| 5873 | (uri (git-reference |
| 5874 | (url "https://github.com/pytorch/captum") |
| 5875 | (commit (assoc-ref properties 'commit)))) |
| 5876 | (file-name (git-file-name name version)) |
| 5877 | (sha256 |
| 5878 | (base32 |
| 5879 | "1v1di9kk2xinwwcll5q514fn9r6ivj3zzgg6l5h7dwlsn2vymsnd")))) |
| 5880 | (build-system pyproject-build-system) |
| 5881 | (arguments |
| 5882 | (list |
| 5883 | ;; tests: 17 failed, 1368 passed, 75 skipped, 3 deselected, 542 warnings |
| 5884 | #:test-flags |
| 5885 | #~(list "-k" (string-append |
| 5886 | ;; These two tests (out of more than 1000 tests) fail |
| 5887 | ;; because of accuracy problems. |
| 5888 | "not test_softmax_classification_batch_multi_target" |
| 5889 | " and not test_softmax_classification_batch_zero_baseline" |
| 5890 | ;; This test fails with PyTorch 2.7.0 due to stricter |
| 5891 | ;; torch.load weights_only behavior. |
| 5892 | " and not test_exp_sets_with_diffent_lengths") |
| 5893 | "tests"))) |
| 5894 | (native-inputs |
| 5895 | (list python-flask |
| 5896 | python-pytest |
| 5897 | python-flask-compress |
| 5898 | python-parameterized |
| 5899 | python-scikit-learn |
| 5900 | python-setuptools)) |
| 5901 | (propagated-inputs |
| 5902 | (list python-matplotlib |
| 5903 | python-numpy |
| 5904 | python-pytorch |
| 5905 | python-tqdm |
| 5906 | ;; [optional] |
| 5907 | python-openai)) |
| 5908 | (home-page "https://captum.ai") |
| 5909 | (synopsis "Model interpretability for PyTorch") |
| 5910 | (description "Captum is a model interpretability and understanding library |
| 5911 | for PyTorch. Captum contains general purpose implementations of integrated |
| 5912 | gradients, saliency maps, smoothgrad, vargrad and others for PyTorch models. |
| 5913 | It has quick integration for models built with domain-specific libraries such |
| 5914 | as torchvision, torchtext, and others.") |
| 5915 | (license license:bsd-3))) |
| 5916 | |
| 5917 | (define-public python-readchar |
| 5918 | (package |
| 5919 | (name "python-readchar") |
| 5920 | (version "4.2.2") |
| 5921 | (source |
| 5922 | (origin |
| 5923 | (method git-fetch) |
| 5924 | (uri (git-reference |
| 5925 | (url "https://github.com/magmax/python-readchar/") |
| 5926 | (commit (string-append "v" version)))) |
| 5927 | (file-name (git-file-name name version)) |
| 5928 | (sha256 |
| 5929 | (base32 "1m5knal63lg1pmhkkkcpvz1vq6k5rpar4yvq1rafc4q47m2cfaaq")))) |
| 5930 | (build-system pyproject-build-system) |
| 5931 | (native-inputs |
| 5932 | (list python-pytest python-setuptools)) |
| 5933 | (home-page "https://github.com/magmax/python-readchar") |
| 5934 | (synopsis "Library to easily read single chars and key strokes") |
| 5935 | (description "This package provides a Python library to easily read single |
| 5936 | characters and key strokes.") |
| 5937 | (license license:expat))) |
| 5938 | |
| 5939 | (define-public python-inquirer |
| 5940 | (package |
| 5941 | (name "python-inquirer") |
| 5942 | (version "3.4.1") |
| 5943 | ;; Pypi has no tests. |
| 5944 | (source (origin |
| 5945 | (method git-fetch) |
| 5946 | (uri (git-reference |
| 5947 | (url "https://github.com/magmax/python-inquirer") |
| 5948 | (commit (string-append "v" version)))) |
| 5949 | (file-name (git-file-name name version)) |
| 5950 | (sha256 |
| 5951 | (base32 |
| 5952 | "10ncdjz0z09m9piqxyk3yxqk1rd9icij46a9djqj7rwdj5sfclf5")))) |
| 5953 | (build-system pyproject-build-system) |
| 5954 | (propagated-inputs |
| 5955 | (list python-blessed python-editor python-readchar)) |
| 5956 | (native-inputs |
| 5957 | (list python-poetry-core python-pexpect python-pytest)) |
| 5958 | (home-page "https://github.com/magmax/python-inquirer") |
| 5959 | (synopsis "Collection of common interactive command line user interfaces") |
| 5960 | (description |
| 5961 | "Inquirer should ease the process of asking end user questions, parsing, |
| 5962 | validating answers, managing hierarchical prompts and providing error |
| 5963 | feedback.") |
| 5964 | (license license:expat))) |
| 5965 | |
| 5966 | (define-public python-pytorch-lightning |
| 5967 | (package |
| 5968 | (name "python-pytorch-lightning") |
| 5969 | (version "2.6.0") |
| 5970 | (source (origin |
| 5971 | (method git-fetch) |
| 5972 | (uri (git-reference |
| 5973 | (url "https://github.com/Lightning-AI/pytorch-lightning") |
| 5974 | (commit version))) |
| 5975 | (file-name (git-file-name name version)) |
| 5976 | (sha256 |
| 5977 | (base32 |
| 5978 | "0fqzjs3k55n2sp1kz8wxf82687d1vq0yydv45qx2vl3kzgvl0qff")))) |
| 5979 | (build-system pyproject-build-system) |
| 5980 | (arguments |
| 5981 | (list |
| 5982 | #:test-flags |
| 5983 | ;; The tests train a model. They are much too expensive for our |
| 5984 | ;; purposes, so we only run the core tests. |
| 5985 | '(list "-m" "not cloud and not tpu" |
| 5986 | "tests/tests_pytorch/core" |
| 5987 | "-k" |
| 5988 | (string-append |
| 5989 | ;; Some multiprocessing complaint. |
| 5990 | "not test_result_reduce_ddp" |
| 5991 | ;; FutureWarning is raised. |
| 5992 | " and not test_result_collection_restoration" |
| 5993 | ;; These need tensorboard |
| 5994 | " and not test_property_logger" |
| 5995 | " and not test_property_loggers")) |
| 5996 | #:phases |
| 5997 | '(modify-phases %standard-phases |
| 5998 | (add-after 'unpack 'patch-version-detection |
| 5999 | (lambda _ |
| 6000 | ;; We do have pytorch 2.4.0, but the version comparison fails. |
| 6001 | (substitute* "src/lightning/fabric/utilities/imports.py" |
| 6002 | (("_TORCH_GREATER_EQUAL_2_4 =.*") |
| 6003 | "_TORCH_GREATER_EQUAL_2_4 = True\n")))) |
| 6004 | (add-after 'patch-version-detection 'fix-leafspec-deprecation |
| 6005 | (lambda _ |
| 6006 | ;; `LeafSpec' is deprecated starting with PyTorch 2.10. |
| 6007 | (substitute* "src/lightning/pytorch/utilities/_pytree.py" |
| 6008 | (("SUPPORTED_NODES, LeafSpec, PyTree, TreeSpec, _get_node_type, tree_unflatten") |
| 6009 | "SUPPORTED_NODES, PyTree, TreeSpec, _get_node_type, tree_unflatten, treespec_leaf") |
| 6010 | (("LeafSpec\\(\\)") "treespec_leaf()")))) |
| 6011 | (add-before 'build 'pre-build |
| 6012 | (lambda _ (setenv "PACKAGE_NAME" "lightning"))) |
| 6013 | (add-after 'install 'pre-build-pytorch |
| 6014 | (lambda _ |
| 6015 | ;; pyproject-build-system only tolerates unicycles. |
| 6016 | (for-each delete-file (find-files "dist" "\\.whl")) |
| 6017 | (setenv "PACKAGE_NAME" "pytorch"))) |
| 6018 | (add-after 'pre-build-pytorch 'build-pytorch |
| 6019 | (assoc-ref %standard-phases 'build)) |
| 6020 | (add-after 'build-pytorch 'install-pytorch |
| 6021 | (assoc-ref %standard-phases 'install)) |
| 6022 | (add-before 'check 'pre-check |
| 6023 | (lambda _ |
| 6024 | ;; We don't have Tensorboard |
| 6025 | (substitute* "tests/tests_pytorch/test_cli.py" |
| 6026 | ((" TensorBoardLogger\\(\".\"\\)") ""))))))) |
| 6027 | (propagated-inputs |
| 6028 | (list python-arrow |
| 6029 | python-beautifulsoup4 |
| 6030 | python-croniter |
| 6031 | python-dateutils |
| 6032 | python-deepdiff |
| 6033 | python-fastapi |
| 6034 | python-fsspec |
| 6035 | python-inquirer |
| 6036 | python-jsonargparse |
| 6037 | python-lightning-cloud |
| 6038 | python-lightning-utilities |
| 6039 | python-numpy |
| 6040 | python-packaging |
| 6041 | python-pytorch |
| 6042 | python-pyyaml |
| 6043 | python-torchmetrics |
| 6044 | python-torchvision |
| 6045 | python-tqdm |
| 6046 | python-traitlets |
| 6047 | python-typing-extensions)) |
| 6048 | (native-inputs |
| 6049 | (list python-aiohttp |
| 6050 | python-cloudpickle |
| 6051 | python-flaky |
| 6052 | python-pympler |
| 6053 | python-pytest |
| 6054 | python-psutil |
| 6055 | python-requests-mock |
| 6056 | python-scikit-learn |
| 6057 | python-setuptools)) |
| 6058 | (home-page "https://lightning.ai/") |
| 6059 | (synopsis "Deep learning framework to train, deploy, and ship AI products") |
| 6060 | (description |
| 6061 | "PyTorch Lightning is just organized PyTorch; Lightning disentangles |
| 6062 | PyTorch code to decouple the science from the engineering.") |
| 6063 | (license license:asl2.0))) |
| 6064 | |
| 6065 | (define-public python-torchmetrics |
| 6066 | (package |
| 6067 | (name "python-torchmetrics") |
| 6068 | (version "1.9.0") |
| 6069 | (source (origin |
| 6070 | (method git-fetch) |
| 6071 | (uri (git-reference |
| 6072 | (url "https://github.com/Lightning-AI/torchmetrics") |
| 6073 | (commit (string-append "v" version)))) |
| 6074 | (file-name (git-file-name name version)) |
| 6075 | (sha256 |
| 6076 | (base32 |
| 6077 | "076x59493wym57lrvmlk8xsbfhgs4rlcw4gx3bxlv55ldd0ijllf")))) |
| 6078 | (build-system pyproject-build-system) |
| 6079 | (arguments |
| 6080 | (list |
| 6081 | ;; Tests require many additional dependencies |
| 6082 | #:tests? #f)) |
| 6083 | (propagated-inputs |
| 6084 | (list python-numpy python-packaging python-pytorch |
| 6085 | python-typing-extensions |
| 6086 | python-lightning-utilities)) |
| 6087 | (native-inputs |
| 6088 | (list python-cloudpickle |
| 6089 | python-fire |
| 6090 | python-mir-eval |
| 6091 | python-pandas |
| 6092 | python-psutil |
| 6093 | python-pytest |
| 6094 | python-pytest-doctestplus |
| 6095 | python-pytest-rerunfailures |
| 6096 | python-pytest-timeout |
| 6097 | python-requests |
| 6098 | python-scikit-image |
| 6099 | python-scikit-learn |
| 6100 | python-scipy |
| 6101 | python-types-protobuf |
| 6102 | python-types-setuptools)) |
| 6103 | (home-page "https://github.com/Lightning-AI/torchmetrics") |
| 6104 | (synopsis "Machine learning metrics for PyTorch applications") |
| 6105 | (description "TorchMetrics is a collection of 100+ PyTorch metrics |
| 6106 | implementations and an easy-to-use API to create custom metrics. It offers: |
| 6107 | |
| 6108 | @itemize |
| 6109 | @item A standardized interface to increase reproducibility |
| 6110 | @item Reduces boilerplate |
| 6111 | @item Automatic accumulation over batches |
| 6112 | @item Metrics optimized for distributed-training |
| 6113 | @item Automatic synchronization between multiple devices |
| 6114 | @end itemize |
| 6115 | ") |
| 6116 | (license license:asl2.0))) |
| 6117 | |
| 6118 | ;; Keep this in sync with python-pytorch |
| 6119 | (define-public python-torchvision |
| 6120 | (package |
| 6121 | (name "python-torchvision") |
| 6122 | (version "0.25.0") |
| 6123 | (source (origin |
| 6124 | (method git-fetch) |
| 6125 | (uri (git-reference |
| 6126 | (url "https://github.com/pytorch/vision") |
| 6127 | (commit (string-append "v" version)) |
| 6128 | (recursive? #t))) |
| 6129 | (file-name (git-file-name name version)) |
| 6130 | (sha256 |
| 6131 | (base32 |
| 6132 | "0bikvjiqgl659vjk1pw7hka95g55h5sgh3lmybwqfkzsqhfljjx2")) |
| 6133 | (modules '((guix build utils))) |
| 6134 | (snippet |
| 6135 | '(begin |
| 6136 | (delete-file-recursively "android") |
| 6137 | (delete-file-recursively "ios"))))) |
| 6138 | (build-system pyproject-build-system) |
| 6139 | (arguments |
| 6140 | (list |
| 6141 | ;; The test suite is expensive and there is no easy way to subset it. |
| 6142 | #:tests? #f |
| 6143 | #:phases |
| 6144 | #~(modify-phases %standard-phases |
| 6145 | (add-after 'unpack 'setenv |
| 6146 | (lambda _ |
| 6147 | (let ((jpegdir #$(this-package-input "libjpeg-turbo"))) |
| 6148 | (setenv "TORCHVISION_INCLUDE" |
| 6149 | (string-append jpegdir "/include/")) |
| 6150 | (setenv "TORCHVISION_LIBRARY" |
| 6151 | (string-append jpegdir "/lib/")))))))) |
| 6152 | (inputs |
| 6153 | (list ffmpeg-6 |
| 6154 | libpng |
| 6155 | libjpeg-turbo)) |
| 6156 | (propagated-inputs |
| 6157 | (list python-numpy |
| 6158 | python-typing-extensions |
| 6159 | python-requests |
| 6160 | python-pillow |
| 6161 | python-pytorch)) |
| 6162 | (native-inputs |
| 6163 | (list pybind11-2 |
| 6164 | python-pytest |
| 6165 | python-setuptools |
| 6166 | python-wheel |
| 6167 | which)) |
| 6168 | (home-page "https://pytorch.org/vision/stable/index.html") |
| 6169 | (synopsis "Datasets, transforms and models specific to computer vision") |
| 6170 | (description |
| 6171 | "The torchvision package consists of popular datasets, model architectures, |
| 6172 | and common image transformations for computer vision.") |
| 6173 | (license license:bsd-3))) |
| 6174 | |
| 6175 | (define-public python-torchfile |
| 6176 | ;; Latest release is nine years old. |
| 6177 | (let ((commit "fbd434a5b5562c88b91a95e6476e11dbb7735436") |
| 6178 | (revision "0")) |
| 6179 | (package |
| 6180 | (name "python-torchfile") |
| 6181 | (version (git-version "0.1.0" revision commit)) |
| 6182 | (source |
| 6183 | (origin |
| 6184 | (method git-fetch) |
| 6185 | (uri (git-reference |
| 6186 | (url "https://github.com/bshillingford/python-torchfile/") |
| 6187 | (commit commit))) |
| 6188 | (file-name (git-file-name name version)) |
| 6189 | (sha256 |
| 6190 | (base32 "1f03ks5n3i3cdh16wx5ysxsxh0ai9vpf0k5pdx759vf31f4niz36")))) |
| 6191 | (build-system pyproject-build-system) |
| 6192 | (arguments |
| 6193 | (list |
| 6194 | #:phases |
| 6195 | #~(modify-phases %standard-phases |
| 6196 | (replace 'check |
| 6197 | (lambda* (#:key tests? #:allow-other-keys) |
| 6198 | (when tests? |
| 6199 | (invoke "python3" "tests.py"))))))) |
| 6200 | (propagated-inputs |
| 6201 | (list python-numpy python-setuptools)) |
| 6202 | (home-page "https://github.com/bshillingford/python-torchfile") |
| 6203 | (synopsis "Torch7 binary serialized file parser") |
| 6204 | (description "This package enables you to deserialize Lua |
| 6205 | torch-serialized objects from Python.") |
| 6206 | (license license:bsd-3)))) |
| 6207 | |
| 6208 | (define-public python-geomloss |
| 6209 | (package |
| 6210 | (name "python-geomloss") |
| 6211 | (version "0.3.1") |
| 6212 | (source |
| 6213 | (origin |
| 6214 | (method git-fetch) |
| 6215 | (uri (git-reference |
| 6216 | (url "https://github.com/jeanfeydy/geomloss") |
| 6217 | (commit (string-append "v" version)))) |
| 6218 | (file-name (git-file-name name version)) |
| 6219 | (sha256 |
| 6220 | (base32 "0sx851d54iagvfmdlpwh4559hms4gr2kabsirbd4ib2m7118p7jb")))) |
| 6221 | (build-system pyproject-build-system) |
| 6222 | ;; There are no automated tests. |
| 6223 | (arguments (list #:tests? #false)) |
| 6224 | (propagated-inputs (list python-numpy python-pytorch)) |
| 6225 | (native-inputs (list python-setuptools)) |
| 6226 | (home-page "https://www.kernel-operations.io/geomloss/") |
| 6227 | (synopsis |
| 6228 | "Geometric loss functions between point clouds, images and volumes") |
| 6229 | (description |
| 6230 | "The GeomLoss library provides efficient GPU implementations for: |
| 6231 | |
| 6232 | @itemize |
| 6233 | @item Kernel norms (also known as Maximum Mean Discrepancies). |
| 6234 | @item Hausdorff divergences, which are positive definite generalizations of |
| 6235 | the Chamfer-ICP loss and are analogous to log-likelihoods of Gaussian Mixture |
| 6236 | Models. |
| 6237 | @item Debiased Sinkhorn divergences, which are affordable yet positive and |
| 6238 | definite approximations of Optimal Transport (Wasserstein) distances. |
| 6239 | @end itemize") |
| 6240 | (license license:expat))) |
| 6241 | |
| 6242 | (define-public python-tokenizers |
| 6243 | (package |
| 6244 | (name "python-tokenizers") |
| 6245 | (version "0.22.2") |
| 6246 | (source |
| 6247 | (origin |
| 6248 | (method git-fetch) |
| 6249 | (uri (git-reference |
| 6250 | (url "https://github.com/huggingface/tokenizers/") |
| 6251 | (commit (string-append "v" version)))) |
| 6252 | (file-name (git-file-name name version)) |
| 6253 | (sha256 |
| 6254 | (base32 "0mcivkmhdpqg872yvp77505apaylml808lbq2i3i5pln2rhgf1jw")) |
| 6255 | (modules '((guix build utils) |
| 6256 | (ice-9 ftw))) |
| 6257 | (snippet |
| 6258 | #~(begin ;; Only keeping bindings. |
| 6259 | (for-each (lambda (file) |
| 6260 | (unless (member file |
| 6261 | '("." ".." "bindings" "tokenizers")) |
| 6262 | (delete-file-recursively file))) |
| 6263 | (scandir ".")) |
| 6264 | (delete-file-recursively "bindings/node") |
| 6265 | (delete-file "bindings/python/Cargo.lock"))))) |
| 6266 | (build-system pyproject-build-system) |
| 6267 | (arguments |
| 6268 | (list |
| 6269 | #:imported-modules (append %cargo-build-system-modules |
| 6270 | %pyproject-build-system-modules) |
| 6271 | #:modules '(((guix build cargo-build-system) #:prefix cargo:) |
| 6272 | (guix build pyproject-build-system) |
| 6273 | (guix build utils)) |
| 6274 | #:test-flags |
| 6275 | #~(list |
| 6276 | "--ignore=benches" |
| 6277 | ;; This test requires the python-datasets package. |
| 6278 | "--ignore=tests/documentation/test_tutorial_train_from_iterators.py" |
| 6279 | ;; XXX: Tests below fail because they require a network connection. |
| 6280 | ;; "-m" "not network" is respected from version 0.23 onwards |
| 6281 | ;; and makes the flags below unecessary. |
| 6282 | "-k" (string-join |
| 6283 | (list "not TestEncoding" |
| 6284 | "TestBPE" |
| 6285 | "TestWordLevel" |
| 6286 | "TestByteLevelProcessing" |
| 6287 | "TestAsyncTokenizer" |
| 6288 | "TestUnigram" |
| 6289 | "TestPipeline" |
| 6290 | "TestCharBPETokenizer" |
| 6291 | "TestSerialization" |
| 6292 | "TestByteLevelBPE" |
| 6293 | "TestBertWordPieceTokenizer" |
| 6294 | "TestQuicktour" |
| 6295 | "test_encode_formats" |
| 6296 | "test_encode_add_special_tokens" |
| 6297 | "test_decode_stream_fallback" |
| 6298 | "test_decode_skip_special_tokens" |
| 6299 | "test_from_pretrained" |
| 6300 | "test_from_pretrained_revision" |
| 6301 | "test_encode_special_tokens" |
| 6302 | "test_instantiate" |
| 6303 | "test_splitting") |
| 6304 | " and not ")) |
| 6305 | #:phases |
| 6306 | #~(modify-phases %standard-phases |
| 6307 | (add-after 'unpack 'prepare-cargo-build-system |
| 6308 | (lambda args |
| 6309 | (for-each (lambda (phase) |
| 6310 | (format #t "Running cargo phase: ~a~%" phase) |
| 6311 | (apply (assoc-ref cargo:%standard-phases phase) |
| 6312 | #:cargo-target #$(cargo-triplet) |
| 6313 | args)) |
| 6314 | '(prepare-rust-crates |
| 6315 | unpack-rust-crates |
| 6316 | configure |
| 6317 | check-for-pregenerated-files |
| 6318 | patch-cargo-checksums)))) |
| 6319 | (add-after 'prepare-cargo-build-system 'chdir |
| 6320 | (lambda _ |
| 6321 | (chdir "bindings/python")))))) |
| 6322 | (native-inputs |
| 6323 | (append (list python-httpx |
| 6324 | python-huggingface-hub |
| 6325 | python-numpy |
| 6326 | python-pytest |
| 6327 | python-pytest-asyncio |
| 6328 | python-pytest-timeout |
| 6329 | python-requests |
| 6330 | python-tqdm) |
| 6331 | (list maturin |
| 6332 | nss-certs-for-test |
| 6333 | pkg-config |
| 6334 | rust |
| 6335 | `(,rust "cargo")) |
| 6336 | (or (and=> (%current-target-system) |
| 6337 | (compose list make-rust-sysroot)) |
| 6338 | '()))) |
| 6339 | (inputs |
| 6340 | (cons oniguruma (cargo-inputs 'tokenizers-python))) |
| 6341 | (home-page "https://huggingface.co/docs/tokenizers") |
| 6342 | (synopsis "Implementation of various popular tokenizers") |
| 6343 | (description |
| 6344 | "This package provides an implementation of today’s most used tokenizers, |
| 6345 | with a focus on performance and versatility.") |
| 6346 | (license license:asl2.0))) |
| 6347 | |
| 6348 | (define-public python-transformers |
| 6349 | (package |
| 6350 | (name "python-transformers") |
| 6351 | (version "4.57.6") |
| 6352 | (source |
| 6353 | (origin |
| 6354 | (method url-fetch) |
| 6355 | (uri (pypi-uri "transformers" version)) |
| 6356 | (sha256 |
| 6357 | (base32 "1lw90svznnmbp48acf33fvr6wbjp5d4ybdr12llhmp79xhk43r2m")))) |
| 6358 | (build-system pyproject-build-system) |
| 6359 | (arguments |
| 6360 | ;; Missing inputs. |
| 6361 | (list #:test-flags |
| 6362 | '(list "--ignore=tests/test_modeling_tf_common.py" |
| 6363 | "--ignore=tests/test_modeling_common.py" ; ImportError |
| 6364 | "--ignore=tests/test_processing_common.py" ; ImportError |
| 6365 | "--ignore=tests/test_configuration_common.py" |
| 6366 | "--ignore=tests/test_pipeline_mixin.py" |
| 6367 | "--ignore=tests/test_sequence_feature_extraction_common.py" |
| 6368 | "--ignore=tests/test_tokenization_mistral_common.py" |
| 6369 | ;; Requires python-accelerate |
| 6370 | "--ignore=tests/test_training_args.py" |
| 6371 | ;; Requires network access |
| 6372 | "--ignore=tests/test_executorch.py"))) |
| 6373 | ;; The imported package contains ~60 more inputs, but they don't seem |
| 6374 | ;; necessary to build a minimal version of the package. |
| 6375 | (propagated-inputs |
| 6376 | (list python-filelock |
| 6377 | python-huggingface-hub |
| 6378 | python-numpy |
| 6379 | python-packaging |
| 6380 | python-pytorch |
| 6381 | python-pyyaml |
| 6382 | python-regex |
| 6383 | python-requests |
| 6384 | python-safetensors |
| 6385 | python-tokenizers |
| 6386 | python-tqdm)) |
| 6387 | (native-inputs |
| 6388 | (list python-parameterized |
| 6389 | python-pytest |
| 6390 | python-pytest-asyncio |
| 6391 | python-setuptools)) |
| 6392 | (home-page "https://github.com/huggingface/transformers") |
| 6393 | (synopsis "Machine Learning for PyTorch and TensorFlow") |
| 6394 | (description |
| 6395 | "This package provides easy download of thousands of pretrained models to |
| 6396 | perform tasks on different modalities such as text, vision, and audio. |
| 6397 | |
| 6398 | These models can be applied on: |
| 6399 | @itemize |
| 6400 | @item Text, for tasks like text classification, information extraction, |
| 6401 | question answering, summarization, translation, and text generation, in over |
| 6402 | 100 languages. |
| 6403 | @item Images, for tasks like image classification, object detection, and |
| 6404 | segmentation. |
| 6405 | @item Audio, for tasks like speech recognition and audio classification. |
| 6406 | @end itemize |
| 6407 | |
| 6408 | Transformer models can also perform tasks on several modalities combined, such |
| 6409 | as table question answering, optical character recognition, information |
| 6410 | extraction from scanned documents, video classification, and visual question |
| 6411 | answering. |
| 6412 | |
| 6413 | This package provides APIs to quickly download and use those pretrained models |
| 6414 | on a given text, fine-tune them on your own datasets and then share them with |
| 6415 | the community. At the same time, each Python module defining an architecture |
| 6416 | is fully standalone and can be modified to enable quick research experiments. |
| 6417 | |
| 6418 | Transformers is backed by the three most popular deep learning libraries — |
| 6419 | Jax, PyTorch and TensorFlow — with a seamless integration between them.") |
| 6420 | (license license:asl2.0))) |
| 6421 | |
| 6422 | (define-public ctranslate2 |
| 6423 | (package |
| 6424 | (name "ctranslate2") |
| 6425 | (version "4.8.0") |
| 6426 | (source |
| 6427 | (origin |
| 6428 | (method git-fetch) |
| 6429 | (uri (git-reference |
| 6430 | (url "https://github.com/OpenNMT/CTranslate2") |
| 6431 | (commit (string-append "v" version)))) |
| 6432 | (file-name (git-file-name "CTranslate2" version)) |
| 6433 | (sha256 (base32 "0lpykbp9jvkds2dy3lfyn699d0kvz9kzqzfzg4vh5cgn99d7k4g2")) |
| 6434 | (patches (search-patches "ctranslate2-local-build.patch")))) |
| 6435 | (build-system cmake-build-system) |
| 6436 | (arguments |
| 6437 | (list |
| 6438 | ;; XXX: mkl and openblas seem incompatible. |
| 6439 | #:configure-flags #~(list "-DBUILD_TESTS=ON" |
| 6440 | "-DWITH_ACCELERATE=OFF" |
| 6441 | "-DWITH_RUY=OFF" |
| 6442 | "-DWITH_DNNL=ON" |
| 6443 | "-DWITH_CUDA=OFF" |
| 6444 | "-DWITH_CUDNN=OFF" |
| 6445 | "-DWITH_MKL=OFF" |
| 6446 | "-DWITH_OPENBLAS=ON") |
| 6447 | #:modules '((guix build cmake-build-system) |
| 6448 | ((guix build gnu-build-system) #:prefix gnu:) |
| 6449 | (guix build utils)) |
| 6450 | #:phases |
| 6451 | #~(modify-phases %standard-phases |
| 6452 | (replace 'check |
| 6453 | (lambda* (#:rest args) |
| 6454 | (apply (assoc-ref gnu:%standard-phases 'check) |
| 6455 | #:test-target "ctranslate2_test" args)))))) |
| 6456 | (native-inputs (list libomp-13 |
| 6457 | cxxopts |
| 6458 | spdlog |
| 6459 | googletest |
| 6460 | cpu-features |
| 6461 | onednn |
| 6462 | openblas)) |
| 6463 | (home-page "https://opennmt.net/CTranslate2/") |
| 6464 | (synopsis "Fast inference engine for Transformer models") |
| 6465 | (description "CTranslate2 is a C++ and Python library for efficient |
| 6466 | inference with Transformer models. |
| 6467 | |
| 6468 | The project implements a custom runtime that applies many performance |
| 6469 | optimization techniques such as weights quantization, layers fusion, batch |
| 6470 | reordering, etc., to accelerate and reduce the memory usage of Transformer |
| 6471 | models on CPU and GPU.") |
| 6472 | (license license:expat))) |
| 6473 | |
| 6474 | (define-public python-ctranslate2 |
| 6475 | (package/inherit ctranslate2 |
| 6476 | (name "python-ctranslate2") |
| 6477 | (build-system pyproject-build-system) |
| 6478 | (arguments |
| 6479 | (list |
| 6480 | ;; XXX: tests require the eole python project. |
| 6481 | #:tests? #f |
| 6482 | #:phases |
| 6483 | #~(modify-phases %standard-phases |
| 6484 | (add-before 'build 'cd-to-build-dir |
| 6485 | (lambda _ |
| 6486 | (chdir "python")))))) |
| 6487 | (native-inputs (list ctranslate2 |
| 6488 | python-setuptools |
| 6489 | python-wheel |
| 6490 | pybind11 |
| 6491 | python-pyyaml |
| 6492 | python-numpy)))) |
| 6493 | |
| 6494 | (define-public python-xx-sent-ud-sm |
| 6495 | ;; For Argos Translate CLI tool |
| 6496 | (hidden-package |
| 6497 | (package |
| 6498 | (name "python-xx-sent-ud-sm") |
| 6499 | (version "3.8.0") |
| 6500 | (source |
| 6501 | (origin |
| 6502 | (method url-fetch) |
| 6503 | (uri (string-append "https://github.com/explosion/spacy-models" |
| 6504 | "/releases/download/xx_sent_ud_sm-3.8.0" |
| 6505 | "/xx_sent_ud_sm-" version ".tar.gz")) |
| 6506 | (sha256 |
| 6507 | (base32 "1pwyl23nvmmkpvq5wgafdqq0mlnq14l2svh9m6gi6w0skv7dd5fq")))) |
| 6508 | (build-system pyproject-build-system) |
| 6509 | (arguments '(#:tests? #f)) ;no test |
| 6510 | (native-inputs (list python-setuptools)) |
| 6511 | (propagated-inputs (list python-spacy)) |
| 6512 | (home-page "https://spacy.io/models/xx#xx_sent_ud_sm") |
| 6513 | (synopsis "Small multi-language sentence boundary model for spaCy") |
| 6514 | (description "@code{xx_sent_ud_sm} is a spaCy model optimized for CPU. |
| 6515 | It includes a sentence boundary component of a relatively small size.") |
| 6516 | (license license:cc-by-sa3.0)))) |
| 6517 | |
| 6518 | (define-public python-argostranslate |
| 6519 | (package |
| 6520 | (name "python-argostranslate") |
| 6521 | (version "1.10.0") |
| 6522 | (source |
| 6523 | (origin |
| 6524 | (method git-fetch) |
| 6525 | (uri (git-reference |
| 6526 | (url "https://github.com/argosopentech/argos-translate") |
| 6527 | (commit (string-append "v" version)))) |
| 6528 | (file-name (git-file-name name version)) |
| 6529 | (sha256 |
| 6530 | (base32 "18fhqabdyh66ymri9xi9xc3q619f5bd3yb9izwhvwyladx65xcfn")))) |
| 6531 | (arguments |
| 6532 | (list |
| 6533 | #:phases |
| 6534 | #~(modify-phases %standard-phases |
| 6535 | (add-after 'unpack 'disable-spacy-download |
| 6536 | (lambda _ |
| 6537 | (substitute* "argostranslate/networking.py" |
| 6538 | ((".*spacy_download.*") "")))) |
| 6539 | (add-before 'build 'set-home-env |
| 6540 | (lambda _ |
| 6541 | (setenv "HOME" "/tmp")))))) |
| 6542 | (build-system pyproject-build-system) |
| 6543 | (native-inputs (list python-pytest python-setuptools)) |
| 6544 | (propagated-inputs (list python-ctranslate2 |
| 6545 | python-packaging |
| 6546 | python-sacremoses |
| 6547 | python-sentencepiece |
| 6548 | python-spacy |
| 6549 | python-stanza |
| 6550 | python-xx-sent-ud-sm)) |
| 6551 | (home-page "https://www.argosopentech.com") |
| 6552 | (synopsis "Offline translation library and CLI tool") |
| 6553 | (description |
| 6554 | "Argos Translate uses OpenNMT for translations and can be used |
| 6555 | as either a Python library, command-line, or GUI application. |
| 6556 | It supports installing language model packages which are ZIP archives |
| 6557 | with a @code{.argosmodel} extension. This package provides |
| 6558 | the Python library and the command-line program.") |
| 6559 | (license license:expat))) |
| 6560 | |
| 6561 | (define-public argos-translate-gui |
| 6562 | ;; The Git repository is not tagged. |
| 6563 | (let ((commit "06aafde3faf1709080471c94914382a566f6f775") |
| 6564 | (revision "0")) |
| 6565 | (package |
| 6566 | (name "argos-translate-gui") |
| 6567 | (version (git-version "1.6.5" revision commit)) |
| 6568 | (source |
| 6569 | (origin |
| 6570 | (method git-fetch) |
| 6571 | (uri (git-reference |
| 6572 | (url "https://github.com/argosopentech/argos-translate-gui") |
| 6573 | (commit commit))) |
| 6574 | (file-name (git-file-name name version)) |
| 6575 | (sha256 |
| 6576 | (base32 "0g251fgwwbk96f8n11c8gzidqgfkrkwpldxk9zq0kkhm729b9yyw")))) |
| 6577 | (build-system pyproject-build-system) |
| 6578 | (arguments '(#:tests? #f)) ;no test |
| 6579 | (native-inputs (list python-setuptools)) |
| 6580 | (inputs (list python-argostranslate python-pyqt)) |
| 6581 | (home-page "https://www.argosopentech.com") |
| 6582 | (synopsis "Graphical user interface for Argos Translate") |
| 6583 | (description |
| 6584 | "Argos Translate uses OpenNMT for translations and can be used |
| 6585 | as either a Python library, command-line, or GUI application. |
| 6586 | It supports installing language model packages which are ZIP archives |
| 6587 | with a @code{.argosmodel} extension. This package provides |
| 6588 | a Qt interface for Argos Translate.") |
| 6589 | (license license:expat)))) |
| 6590 | |
| 6591 | (define-public python-argos-translate-files |
| 6592 | (package |
| 6593 | (name "python-argos-translate-files") |
| 6594 | (version "1.4.3") |
| 6595 | (source |
| 6596 | (origin |
| 6597 | (method git-fetch) |
| 6598 | (uri (git-reference |
| 6599 | (url "https://github.com/LibreTranslate/argos-translate-files") |
| 6600 | (commit (string-append "v" version)))) |
| 6601 | (file-name (git-file-name name version)) |
| 6602 | (sha256 |
| 6603 | (base32 "1vz5v2gnd5wxwq6cjnhrvhflndl55c7ffy21w05m8sy3ll8iaiij")))) |
| 6604 | (build-system pyproject-build-system) |
| 6605 | (arguments (list #:phases #~(modify-phases %standard-phases |
| 6606 | (add-after 'unpack 'unpin-dependencies |
| 6607 | (lambda _ |
| 6608 | (substitute* "requirements.txt" |
| 6609 | (("argos-translate-lt.*") |
| 6610 | "argostranslate\n")))) |
| 6611 | (add-before 'build 'set-home-env |
| 6612 | (lambda _ |
| 6613 | (setenv "HOME" "/tmp")))))) |
| 6614 | (propagated-inputs (list python-argostranslate |
| 6615 | python-beautifulsoup4 |
| 6616 | python-lxml |
| 6617 | python-pymupdf |
| 6618 | python-pysrt)) |
| 6619 | (native-inputs (list python-pytest python-setuptools)) |
| 6620 | (home-page "https://github.com/LibreTranslate/argos-translate-files") |
| 6621 | (synopsis "Translate files with Argos Translate") |
| 6622 | (description "This package provides a Python library |
| 6623 | for translating plain- and rich-text documents with Argos Translate.") |
| 6624 | (license license:agpl3))) |
| 6625 | |
| 6626 | (define-public libretranslate |
| 6627 | (package |
| 6628 | (name "libretranslate") |
| 6629 | (version "1.9.6") |
| 6630 | (source |
| 6631 | (origin |
| 6632 | (method git-fetch) |
| 6633 | (uri (git-reference |
| 6634 | (url "https://github.com/LibreTranslate/LibreTranslate") |
| 6635 | (commit (string-append "v" version)))) |
| 6636 | (file-name (git-file-name name version)) |
| 6637 | (sha256 (base32 "0gyhahlyz06nr7h8nb7nrw866z44kw6kaijczk6jk8wn3p1xp200")) |
| 6638 | (patches (search-patches "libretranslate-use-flasgger.patch")))) |
| 6639 | (build-system pyproject-build-system) |
| 6640 | (arguments |
| 6641 | (list |
| 6642 | #:tests? #f ;all tests depend on Argos models |
| 6643 | #:phases |
| 6644 | #~(modify-phases %standard-phases |
| 6645 | (add-after 'unpack 'unpin-dependencies |
| 6646 | (lambda _ |
| 6647 | (substitute* "pyproject.toml" |
| 6648 | ;; Unpin numpy for transitive dependencies. |
| 6649 | (("^ \"numpy .+") "") |
| 6650 | ;; Unpin dependencies. |
| 6651 | (("^( \"[[:alnum:]-]+) ==.+((;.+)?\",)" all left right) |
| 6652 | (string-append left right)) |
| 6653 | (("argos-translate-lt") "argostranslate")))) |
| 6654 | (add-before 'build 'set-home-env |
| 6655 | (lambda _ |
| 6656 | (setenv "HOME" "/tmp")))))) |
| 6657 | (inputs (list python-apscheduler |
| 6658 | python-argos-translate-files |
| 6659 | python-argostranslate |
| 6660 | python-expiringdict |
| 6661 | python-flasgger |
| 6662 | python-flask |
| 6663 | python-flask-babel |
| 6664 | python-flask-limiter |
| 6665 | python-itsdangerous |
| 6666 | python-langdetect |
| 6667 | python-lexilang |
| 6668 | python-minisbd |
| 6669 | python-numpy |
| 6670 | python-packaging |
| 6671 | python-polib |
| 6672 | python-prometheus-client |
| 6673 | python-redis |
| 6674 | python-requests |
| 6675 | python-waitress |
| 6676 | python-werkzeug)) |
| 6677 | (native-inputs (list python-hatchling)) |
| 6678 | (home-page "https://docs.libretranslate.com") |
| 6679 | (synopsis "Machine translation API and web interface") |
| 6680 | (description |
| 6681 | "LibreTranslate is a machine translation API and web interface, |
| 6682 | powered by the Argos Translate library.") |
| 6683 | (license license:agpl3))) |
| 6684 | |
| 6685 | (define-public python-hmmlearn |
| 6686 | (package |
| 6687 | (name "python-hmmlearn") |
| 6688 | (version "0.3.3") |
| 6689 | (source |
| 6690 | (origin |
| 6691 | (method git-fetch) |
| 6692 | (uri (git-reference |
| 6693 | (url "https://github.com/hmmlearn/hmmlearn") |
| 6694 | (commit version))) |
| 6695 | (file-name (git-file-name name version)) |
| 6696 | (sha256 |
| 6697 | (base32 "1d1gcw2xb1i7zvkcz2vmrk4c2iqsfy71pgrc0r5p65r51w9p97n1")))) |
| 6698 | (build-system pyproject-build-system) |
| 6699 | (arguments |
| 6700 | (list |
| 6701 | #:test-flags |
| 6702 | #~(list "--pyargs" "hmmlearn") |
| 6703 | #:phases |
| 6704 | #~(modify-phases %standard-phases |
| 6705 | (add-after 'unpack 'set-core-count |
| 6706 | (lambda _ |
| 6707 | ;; "Could not find the number of physical cores", so we tell it |
| 6708 | ;; how many cores to use. |
| 6709 | (setenv "LOKY_MAX_CPU_COUNT" "1")))))) |
| 6710 | (native-inputs |
| 6711 | (list pybind11 |
| 6712 | python-pytest |
| 6713 | python-setuptools |
| 6714 | python-setuptools-scm |
| 6715 | util-linux)) ;for lscpu |
| 6716 | (propagated-inputs |
| 6717 | (list python-numpy |
| 6718 | python-scikit-learn |
| 6719 | python-scipy)) |
| 6720 | (home-page "https://github.com/hmmlearn/hmmlearn") |
| 6721 | (synopsis "Hidden Markov Models with scikit-learn like API") |
| 6722 | (description |
| 6723 | "Hmmlearn is a set of algorithms for unsupervised learning and inference |
| 6724 | of Hidden Markov Models.") |
| 6725 | (license license:bsd-3))) |
| 6726 | |
| 6727 | (define-public llm-proxy |
| 6728 | (package |
| 6729 | (name "llm-proxy") |
| 6730 | (version "0.23.0") |
| 6731 | (source |
| 6732 | (origin |
| 6733 | (method git-fetch) |
| 6734 | (uri (git-reference |
| 6735 | (url "https://codeberg.org/jbousquie/llm-proxy") |
| 6736 | (commit (string-append "v" version)))) |
| 6737 | (file-name (git-file-name name version)) |
| 6738 | (sha256 |
| 6739 | (base32 "125lfhx28fs8s3fzhcqbs87b7crswn9x0br3dnv6qnbybrg0bacd")))) |
| 6740 | (build-system cargo-build-system) |
| 6741 | (arguments (list #:install-source? #f)) |
| 6742 | (inputs (cargo-inputs 'llm-proxy)) |
| 6743 | (home-page "https://jbousquie.codeberg.page/llm-proxy/") |
| 6744 | (synopsis "Network proxy for LLM API vendors") |
| 6745 | (native-inputs (list nss-certs-for-test)) |
| 6746 | (description |
| 6747 | "This package provides a local HTTP proxy written in Rust. It sits |
| 6748 | between agentic clients and the Albert API (Etalab).") |
| 6749 | (license license:etalab2.0))) |
| 6750 | |
| 6751 | ;; Keep this in sync with the r-torch package. |
| 6752 | (define-public liblantern |
| 6753 | (package |
| 6754 | (name "liblantern") |
| 6755 | (version "0.16.3") |
| 6756 | (source |
| 6757 | (origin |
| 6758 | (method git-fetch) |
| 6759 | (uri (git-reference |
| 6760 | (url "https://github.com/mlverse/torch") |
| 6761 | (commit (string-append "v" version)))) |
| 6762 | (file-name (git-file-name name version)) |
| 6763 | (sha256 |
| 6764 | (base32 "13gdrj9aklqgfx10jyylr3jh2ilk8j27fzgv708yv9jn2ckmfdhn")))) |
| 6765 | (build-system cmake-build-system) |
| 6766 | (arguments |
| 6767 | (list |
| 6768 | #:tests? #false ;no test target |
| 6769 | #:phases |
| 6770 | (let ((python-version (version-major+minor (package-version python)))) |
| 6771 | #~(modify-phases %standard-phases |
| 6772 | (add-after 'unpack 'chdir |
| 6773 | (lambda _ (chdir "src/lantern"))) |
| 6774 | (add-after 'chdir 'do-not-download-binaries |
| 6775 | (lambda* (#:key inputs #:allow-other-keys) |
| 6776 | (substitute* "CMakeLists.txt" |
| 6777 | (("find_package\\(Torch.*") "set(TORCH_CXX_FLAGS \"-ltorch\")\n") |
| 6778 | (("retrieve_lib\\(.*") "")) |
| 6779 | (let ((site-packages (string-append "/lib/python" |
| 6780 | #$python-version |
| 6781 | "/site-packages"))) |
| 6782 | (setenv "LIBRARY_PATH" |
| 6783 | (string-append |
| 6784 | (search-input-directory |
| 6785 | inputs (string-append site-packages "/torch/lib")) |
| 6786 | ":" (or (getenv "LIBRARY_PATH") ""))) |
| 6787 | (setenv "CPLUS_INCLUDE_PATH" |
| 6788 | (string-append |
| 6789 | (search-input-directory |
| 6790 | inputs (string-append |
| 6791 | site-packages "/torch/include/torch/csrc/api/include/")) |
| 6792 | ":" |
| 6793 | (search-input-directory |
| 6794 | inputs (string-append site-packages "/torch/include/")) |
| 6795 | ":" |
| 6796 | (or (getenv "CPLUS_INCLUDE_PATH") ""))) |
| 6797 | (setenv "C_INCLUDE_PATH" |
| 6798 | (string-append |
| 6799 | (search-input-directory |
| 6800 | inputs (string-append site-packages "/torch/include/")) |
| 6801 | ":" |
| 6802 | (or (getenv "C_INCLUDE_PATH") "")))))) |
| 6803 | (replace 'install |
| 6804 | (lambda _ |
| 6805 | (install-file |
| 6806 | "../build/liblantern.so" |
| 6807 | (string-append #$output "/lib")) |
| 6808 | (copy-recursively |
| 6809 | "../lantern/include" |
| 6810 | (string-append #$output "/include")))))))) |
| 6811 | (inputs (list python-pytorch-for-r-torch)) |
| 6812 | (home-page "https://github.com/mlverse/torch/") |
| 6813 | (synopsis "C API to libtorch") |
| 6814 | (description |
| 6815 | "Lantern provides a C API to the libtorch machine learning library.") |
| 6816 | (license license:expat))) |
| 6817 | |
| 6818 | (define-public python-lap |
| 6819 | (package |
| 6820 | (name "python-lap") |
| 6821 | (version "0.5.12") |
| 6822 | (source (origin |
| 6823 | (method url-fetch) |
| 6824 | (uri (pypi-uri "lap" version)) |
| 6825 | (sha256 |
| 6826 | (base32 |
| 6827 | "1za4mf5nd7vzwd24sy2mfxrk8mnwq7d8rv6h96yh8v5flx7422sp")))) |
| 6828 | (build-system pyproject-build-system) |
| 6829 | (arguments |
| 6830 | (list |
| 6831 | #:test-flags #~(list "-v" #$output) |
| 6832 | #:phases |
| 6833 | #~(modify-phases %standard-phases |
| 6834 | (add-after 'check 'check-cleanup |
| 6835 | (lambda* (#:key tests? #:allow-other-keys) |
| 6836 | (when tests? |
| 6837 | (for-each |
| 6838 | delete-file-recursively |
| 6839 | (find-files #$output |
| 6840 | (lambda (file stat) |
| 6841 | (or (member (basename file) |
| 6842 | '("tests" ".pytest_cache")))) |
| 6843 | #:directories? #t)))))))) |
| 6844 | (propagated-inputs |
| 6845 | (list python-numpy)) |
| 6846 | (native-inputs |
| 6847 | (list python-cython python-pytest python-setuptools python-wheel)) |
| 6848 | (home-page "https://github.com/gatagat/lap") |
| 6849 | (synopsis "Linear Assignment Problem solver (LAPJV/LAPMOD)") |
| 6850 | (description "Lap is a linear assignment problem solver using Jonker-Volgenant |
| 6851 | algorithm for dense (LAPJV) or sparse (LAPMOD) matrices.") |
| 6852 | (license license:bsd-2))) |
| 6853 | |
| 6854 | (define-public python-visdom |
| 6855 | (package |
| 6856 | (name "python-visdom") |
| 6857 | (version "0.2.4") |
| 6858 | (source |
| 6859 | (origin |
| 6860 | (method url-fetch) |
| 6861 | (uri (pypi-uri "visdom" version)) |
| 6862 | (sha256 |
| 6863 | (base32 "0x05faxh45bh1zz8afjhzjy49jrqv4dkhaw1ainhajl1r39i3ac4")))) |
| 6864 | (build-system pyproject-build-system) |
| 6865 | (arguments (list #:tests? #f)) ;tests require launching the server |
| 6866 | (native-inputs |
| 6867 | (list python-setuptools)) |
| 6868 | (propagated-inputs |
| 6869 | (list python-jsonpatch |
| 6870 | python-networkx |
| 6871 | python-numpy |
| 6872 | python-pillow |
| 6873 | python-requests |
| 6874 | python-scipy |
| 6875 | python-six |
| 6876 | python-tornado |
| 6877 | python-websocket-client)) |
| 6878 | (home-page "https://github.com/fossasia/visdom") |
| 6879 | (synopsis "Visualizations of live, rich data for Torch and Numpy") |
| 6880 | (description |
| 6881 | "This package provides a tool for visualizing live, rich data for Torch |
| 6882 | and Numpy.") |
| 6883 | (license license:asl2.0))) |
| 6884 | |
| 6885 | (define-public python-pyro-api |
| 6886 | (package |
| 6887 | (name "python-pyro-api") |
| 6888 | (version "0.1.2") |
| 6889 | (source |
| 6890 | (origin |
| 6891 | (method git-fetch) |
| 6892 | (uri (git-reference |
| 6893 | (url "https://github.com/pyro-ppl/pyro-api") |
| 6894 | (commit version))) |
| 6895 | (file-name (git-file-name name version)) |
| 6896 | (sha256 |
| 6897 | (base32 "17x7niagx43cajqq67dxmssr7q94db6axyg154y7vqdxzp25hf7g")))) |
| 6898 | (build-system pyproject-build-system) |
| 6899 | (arguments '(#:tests? #false)) ;requires pyro |
| 6900 | (native-inputs |
| 6901 | (list python-ipython |
| 6902 | python-pytest |
| 6903 | python-setuptools |
| 6904 | python-sphinx |
| 6905 | python-sphinx-rtd-theme |
| 6906 | python-wheel)) |
| 6907 | (home-page "https://github.com/pyro-ppl/pyro-api") |
| 6908 | (synopsis "Generic API for dispatch to Pyro backends") |
| 6909 | (description "This package provides a generic API for dispatch to Pyro backends.") |
| 6910 | (license license:asl2.0))) |
| 6911 | |
| 6912 | (define-public python-pyro-ppl |
| 6913 | (package |
| 6914 | (name "python-pyro-ppl") |
| 6915 | (version "1.9.1") |
| 6916 | ;; The sources on pypi don't include tests. |
| 6917 | (source |
| 6918 | (origin |
| 6919 | (method git-fetch) |
| 6920 | (uri (git-reference |
| 6921 | (url "https://github.com/pyro-ppl/pyro") |
| 6922 | (commit version))) |
| 6923 | (file-name (git-file-name name version)) |
| 6924 | (sha256 |
| 6925 | (base32 "0q87d0frgzn1ljnpbyxmj582yfn6zy3m960m3ab826h4rpzybxhf")))) |
| 6926 | (build-system pyproject-build-system) |
| 6927 | (arguments |
| 6928 | ;; Tests take too long. |
| 6929 | ;; XXX: Maybe select the most important test modules. |
| 6930 | (list #:tests? #f)) |
| 6931 | (propagated-inputs |
| 6932 | (list python-numpy |
| 6933 | python-opt-einsum |
| 6934 | python-pyro-api |
| 6935 | python-pytorch |
| 6936 | python-tqdm)) |
| 6937 | (native-inputs (list python-setuptools python-wheel)) |
| 6938 | (home-page "https://pyro.ai") |
| 6939 | (synopsis "Python library for probabilistic modeling and inference") |
| 6940 | (description |
| 6941 | "This package provides a Python library for probabilistic modeling and |
| 6942 | inference.") |
| 6943 | (license license:asl2.0))) |
| 6944 | |
| 6945 | (define-public python-linear-operator |
| 6946 | (package |
| 6947 | (name "python-linear-operator") |
| 6948 | (version "0.6.1") |
| 6949 | (source |
| 6950 | (origin |
| 6951 | (method git-fetch) |
| 6952 | (uri (git-reference |
| 6953 | (url "https://github.com/cornellius-gp/linear_operator/") |
| 6954 | (commit (string-append "v" version)))) |
| 6955 | (file-name (git-file-name name version)) |
| 6956 | (sha256 |
| 6957 | (base32 |
| 6958 | "1pfk4mdkvm75lyyb7zmq9pxclhica7fkk4cyvkpz8jncgimvh5qs")))) |
| 6959 | (build-system pyproject-build-system) |
| 6960 | (propagated-inputs (list python-jaxtyping |
| 6961 | python-mpmath |
| 6962 | python-pytorch |
| 6963 | python-scipy)) |
| 6964 | (native-inputs (list python-pytest |
| 6965 | python-setuptools |
| 6966 | python-setuptools-scm |
| 6967 | python-typeguard)) |
| 6968 | (home-page "https://github.com/cornellius-gp/linear_operator/") |
| 6969 | (synopsis "Linear operator implementation") |
| 6970 | (description "LinearOperator is a PyTorch package for abstracting away the |
| 6971 | linear algebra routines needed for structured matrices (or operators).") |
| 6972 | (license license:expat))) |
| 6973 | |
| 6974 | (define-public python-gpytorch |
| 6975 | (package |
| 6976 | (name "python-gpytorch") |
| 6977 | (version "1.15.2") |
| 6978 | (source (origin |
| 6979 | (method url-fetch) |
| 6980 | (uri (pypi-uri "gpytorch" version)) |
| 6981 | (sha256 |
| 6982 | (base32 |
| 6983 | "0qakfyw9akxbcvjz2rzn5vmy7ln6lsq5yp5jfbvqa6w57zlja1iq")))) |
| 6984 | (build-system pyproject-build-system) |
| 6985 | (arguments |
| 6986 | (list #:test-flags |
| 6987 | #~(list "-k" (string-append |
| 6988 | ;; test_deprecated_methods fails with an AssertionError. |
| 6989 | "not test_deprecated_methods" |
| 6990 | ;; This test is flaky: Expects gradients of 0 exactly, |
| 6991 | ;; can get negligible ones (e-10 to e-16). |
| 6992 | " and not test_optimization_optimal_error") |
| 6993 | ;; Ignore lenghty tests of little relevance. |
| 6994 | "--ignore=test/examples/"))) |
| 6995 | (propagated-inputs (list python-jaxtyping |
| 6996 | python-linear-operator |
| 6997 | python-mpmath |
| 6998 | python-scikit-learn |
| 6999 | python-scipy)) |
| 7000 | (native-inputs (list python-nbval |
| 7001 | python-pytest |
| 7002 | python-setuptools |
| 7003 | python-setuptools-scm)) |
| 7004 | (home-page "https://gpytorch.ai") |
| 7005 | (synopsis "Implementation of Gaussian Processes in PyTorch") |
| 7006 | (description |
| 7007 | "GPyTorch is a Gaussian process library implemented using PyTorch.") |
| 7008 | (license license:expat))) |
| 7009 | |
| 7010 | (define-public python-botorch |
| 7011 | (package |
| 7012 | (name "python-botorch") |
| 7013 | (version "0.17.2") |
| 7014 | (source (origin |
| 7015 | (method git-fetch) ;no tests in PyPI |
| 7016 | (uri (git-reference |
| 7017 | (url "https://github.com/meta-pytorch/botorch") |
| 7018 | (commit (string-append "v" version)))) |
| 7019 | (file-name (git-file-name name version)) |
| 7020 | (sha256 |
| 7021 | (base32 |
| 7022 | "1flb6fv2m6l8bbmbq9n7lgf1fd45r9xx2rw08qanm1c00dv12s99")))) |
| 7023 | (build-system pyproject-build-system) |
| 7024 | (arguments |
| 7025 | ;; 7 failed, 1502 passed, 1 skipped, 1 deselected, 807 warnings |
| 7026 | (list #:test-flags #~(list "-k" |
| 7027 | (string-append |
| 7028 | "not test_all_cases_covered" |
| 7029 | " and not test_input_constructors" |
| 7030 | " and not test_gen" |
| 7031 | " and not test_mock" |
| 7032 | " and not test_evaluation" |
| 7033 | ;; SciPy<1.13: gen.py doesn't import |
| 7034 | ;; fmin_l_bfgs_b_batched; mock patch fails. |
| 7035 | " and not test_emsemble_map_saas" |
| 7036 | " and not test_negative_fixed_features") |
| 7037 | ;; Requires optional 'pfns' dependency. |
| 7038 | "--ignore=test_community/") |
| 7039 | #:phases |
| 7040 | #~(modify-phases %standard-phases |
| 7041 | (add-after 'unpack 'fix-hypervolume-initial-hvs |
| 7042 | (lambda _ |
| 7043 | ;; PyTorch 2.10 rejects passing a torch.Size plus extra args to `view'. |
| 7044 | (substitute* "botorch/utils/multi_objective/hypervolume.py" |
| 7045 | (("self\\._batch_sample_shape, *\\*obj\\.shape\\[-2:\\]") |
| 7046 | "self._batch_sample_shape"))))))) |
| 7047 | (propagated-inputs (list python-gpytorch |
| 7048 | python-linear-operator |
| 7049 | python-multipledispatch |
| 7050 | python-pyre-extensions |
| 7051 | python-pyro-ppl |
| 7052 | python-pytorch |
| 7053 | python-scipy |
| 7054 | python-threadpoolctl |
| 7055 | python-typing-extensions)) |
| 7056 | (native-inputs (list python-pytest |
| 7057 | python-setuptools |
| 7058 | python-setuptools-scm)) |
| 7059 | (home-page "https://botorch.org") |
| 7060 | (synopsis "Bayesian Optimization in PyTorch") |
| 7061 | (description |
| 7062 | "BoTorch is a library for Bayesian Optimization built on PyTorch.") |
| 7063 | (license license:expat))) |
| 7064 | |
| 7065 | (define-public vosk-api |
| 7066 | (let* ((openfst openfst-for-vosk) |
| 7067 | (kaldi kaldi-for-vosk)) |
| 7068 | (package |
| 7069 | (name "vosk-api") |
| 7070 | (version "0.3.50") |
| 7071 | (source |
| 7072 | (origin |
| 7073 | (method git-fetch) |
| 7074 | (uri (git-reference |
| 7075 | (url "https://github.com/alphacep/vosk-api") |
| 7076 | (commit (string-append "v" version)))) |
| 7077 | (file-name (git-file-name name version)) |
| 7078 | (sha256 |
| 7079 | (base32 "0rm7c1n9iv4y9q6a860rqiy2bdawxjhbfd993lms1ly86vwyai8k")))) |
| 7080 | (build-system gnu-build-system) |
| 7081 | (arguments |
| 7082 | (list |
| 7083 | #:tests? #f |
| 7084 | #:phases |
| 7085 | #~(modify-phases %standard-phases |
| 7086 | (add-after 'unpack 'chdir |
| 7087 | (lambda _ (chdir "src"))) |
| 7088 | (replace 'configure |
| 7089 | (lambda _ |
| 7090 | (let* ((openfst #$(this-package-input "openfst")) |
| 7091 | (openblas #$(this-package-input "openblas")) |
| 7092 | (kaldi #$(this-package-input "kaldi"))) |
| 7093 | (substitute* "./Makefile" |
| 7094 | (("USE_SHARED\\?=0") |
| 7095 | "USE_SHARED?=1") |
| 7096 | (("-DFST_NO_DYNAMIC_LINKING") |
| 7097 | "") |
| 7098 | (("-lopenblas -llapack -lblas -lf2c") |
| 7099 | (string-append |
| 7100 | "-L" openblas "/lib " "-lopenblas ")) |
| 7101 | (("-lfst -lfstngram") |
| 7102 | (string-append |
| 7103 | "-L" openfst "/lib " "-lfst -lfstngram ")) |
| 7104 | (("\\$\\(HOME\\)\\/travis\\/kaldi") |
| 7105 | (string-append kaldi "/include")) |
| 7106 | (("\\$\\(KALDI_ROOT\\)\\/tools\\/openfst") |
| 7107 | openfst) |
| 7108 | (("\\$\\(KALDI_ROOT\\)\\/tools\\/OpenBLAS\\/install") |
| 7109 | openblas) |
| 7110 | (("\\$\\(KALDI_ROOT\\)\\/libs") |
| 7111 | (string-append kaldi "/lib")))))) |
| 7112 | (replace 'install |
| 7113 | (lambda _ |
| 7114 | (let* ((lib (string-append #$output "/lib")) |
| 7115 | (src (string-append #$output "/src"))) |
| 7116 | (mkdir-p lib) |
| 7117 | (mkdir-p src) |
| 7118 | (install-file "libvosk.so" lib) |
| 7119 | (for-each |
| 7120 | (lambda (x) (install-file x src)) |
| 7121 | (find-files "." "\\.h$")))))))) |
| 7122 | (inputs (list kaldi openfst openblas)) |
| 7123 | (home-page "https://alphacephei.com/vosk/") |
| 7124 | (synopsis "Speech recognition toolkit based on @code{kaldi}") |
| 7125 | (description "This package provides a speech recognition toolkit based |
| 7126 | on @code{kaldi}. It supports more than 20 languages and dialects - English, |
| 7127 | Indian English, German, French, Spanish, Portuguese, Chinese, Russian, |
| 7128 | Turkish, Vietnamese, Italian, Dutch, Catalan, Arabic, Greek, Farsi, Filipino, |
| 7129 | Ukrainian, Kazakh, Swedish, Japanese, Esperanto, Hindi, Czech, Polish. The |
| 7130 | program works offline, even on lightweight devices. Portable per-language |
| 7131 | models are about 50Mb each, and there are much bigger and precise models |
| 7132 | available. |
| 7133 | |
| 7134 | Vosk API provides a streaming API allowing to use it @emph{on-the-fly} and |
| 7135 | bindings for different programming languages. It allows quick reconfiguration |
| 7136 | of vocabulary for better accuracy, and supports speaker identification beside |
| 7137 | simple speech recognition.") |
| 7138 | (license license:asl2.0)))) |
| 7139 | |
| 7140 | (define-public python-vosk |
| 7141 | (package |
| 7142 | (inherit vosk-api) |
| 7143 | (name "python-vosk") |
| 7144 | (build-system pyproject-build-system) |
| 7145 | (propagated-inputs |
| 7146 | (list python-cffi python-requests python-tqdm python-srt python-websockets)) |
| 7147 | (inputs (list vosk-api)) |
| 7148 | (native-inputs |
| 7149 | (list python-setuptools python-wheel)) |
| 7150 | (arguments |
| 7151 | (list |
| 7152 | #:tests? #f ;; TODO There are tests but not run through Makefile. |
| 7153 | #:phases |
| 7154 | #~(modify-phases %standard-phases |
| 7155 | (add-after 'unpack 'config |
| 7156 | (lambda _ |
| 7157 | (chdir "python") |
| 7158 | (setenv "VOSK_SOURCE" #$vosk-api))) |
| 7159 | (add-before 'build 'from-abi-to-api |
| 7160 | (lambda* (#:key inputs #:allow-other-keys) |
| 7161 | (substitute* "vosk_builder.py" |
| 7162 | (("ffibuilder\\.set_source\\(\"vosk.vosk_cffi\", None\\)") |
| 7163 | (string-append |
| 7164 | "ffibuilder.set_source(\"vosk.vosk_cffi\", " |
| 7165 | "r\"\"\"\n#include<vosk_api.h>\n#include<Python.h>\"\"\",\n\t" |
| 7166 | "library_dirs=[" |
| 7167 | "'" #$vosk-api "/lib'" |
| 7168 | "],\n\t" |
| 7169 | "libraries=['vosk', 'python" |
| 7170 | (python-version |
| 7171 | (dirname (dirname (search-input-file inputs "bin/python")))) |
| 7172 | "'],\n\t" |
| 7173 | "include_dirs=[" |
| 7174 | "'" #$vosk-api "/src'" "])"))) |
| 7175 | (substitute* "vosk/__init__.py" |
| 7176 | (("_c = open_dll\\(\\)") |
| 7177 | "") |
| 7178 | (("_ffi") |
| 7179 | "ffi") |
| 7180 | (("from \\.vosk_cffi import ffi as ffi") |
| 7181 | "from .vosk_cffi import ffi, lib") |
| 7182 | (("_c\\.") |
| 7183 | "lib."))))))))) |
| 7184 | |
| 7185 | (define-public nerd-dictation |
| 7186 | (let* ((commit "03ce043a6d569a5bb9a715be6a8e45d8ba0930fd") |
| 7187 | (revision "2")) |
| 7188 | (package |
| 7189 | (name "nerd-dictation") |
| 7190 | (version (git-version "0" revision commit)) |
| 7191 | (source |
| 7192 | (origin |
| 7193 | (method git-fetch) |
| 7194 | (uri (git-reference |
| 7195 | (url "https://github.com/ideasman42/nerd-dictation") |
| 7196 | (commit commit))) |
| 7197 | (file-name (git-file-name name version)) |
| 7198 | (sha256 |
| 7199 | (base32 "0ksnlpy2jcldxamq7bsbcgxgafcmgxkbjccd4fwmmn0y814kkz9k")))) |
| 7200 | (build-system pyproject-build-system) |
| 7201 | (arguments |
| 7202 | (list |
| 7203 | #:test-backend #~'custom |
| 7204 | #:test-flags #~(list "../../tests/from_words_to_digits.py") |
| 7205 | #:phases |
| 7206 | #~(modify-phases %standard-phases |
| 7207 | (add-after 'unpack 'chdir |
| 7208 | (lambda _ |
| 7209 | (chdir "package/python")))))) |
| 7210 | (native-inputs (list python-setuptools)) |
| 7211 | (propagated-inputs (list python-vosk)) |
| 7212 | (home-page "https://github.com/ideasman42/nerd-dictation") |
| 7213 | (synopsis "Offline speech-to-text for desktop Linux") |
| 7214 | (description |
| 7215 | "This package provides simple access speech to text for using in |
| 7216 | Linux without being tied to a desktop environment, using the @code{vosk-api}. |
| 7217 | The user configuration lets you manipulate text using Python string |
| 7218 | operations. It has zero overhead, as this relies on manual activation and |
| 7219 | there are no background processes. Dictation is accessed manually with |
| 7220 | @code{nerd-dictation begin} and @code{nerd-dictation end} commands.") |
| 7221 | (license license:gpl2+)))) |
| 7222 | |
| 7223 | (define (nerd-dictation-gexp input-name output-name bash nerd-dictation) |
| 7224 | (with-imported-modules '((guix build utils)) |
| 7225 | #~(begin |
| 7226 | (use-modules (guix build utils)) |
| 7227 | |
| 7228 | (let* ((exe (string-append #$output "/bin/nerd-dictation")) |
| 7229 | (nerd-dictation-exe |
| 7230 | #$(file-append nerd-dictation "/bin/nerd-dictation"))) |
| 7231 | (mkdir-p (dirname exe)) |
| 7232 | (call-with-output-file exe |
| 7233 | (lambda (port) |
| 7234 | (format port "#!~a |
| 7235 | if [ \"$1\" = begin ] |
| 7236 | then |
| 7237 | exec ~a $@ --input=~a --simulate-input-tool=~a |
| 7238 | else |
| 7239 | exec ~a $@ |
| 7240 | fi" |
| 7241 | #$(file-append bash "/bin/bash") |
| 7242 | nerd-dictation-exe |
| 7243 | #$input-name |
| 7244 | #$output-name |
| 7245 | nerd-dictation-exe))) |
| 7246 | (chmod exe #o555))))) |
| 7247 | |
| 7248 | (define-public nerd-dictation/xdotool |
| 7249 | (package |
| 7250 | (inherit nerd-dictation) |
| 7251 | (name "nerd-dictation-xdotool") |
| 7252 | (build-system trivial-build-system) |
| 7253 | (arguments |
| 7254 | (list #:builder |
| 7255 | (nerd-dictation-gexp "PAREC" "XDOTOOL" |
| 7256 | (this-package-input "bash-minimal") |
| 7257 | (this-package-input "nerd-dictation")))) |
| 7258 | (inputs (list bash-minimal nerd-dictation)) |
| 7259 | (propagated-inputs (list pulseaudio xdotool)))) |
| 7260 | |
| 7261 | (define-public nerd-dictation/sox-xdotool |
| 7262 | (package |
| 7263 | (inherit nerd-dictation/xdotool) |
| 7264 | (name "nerd-dictation-sox-xdotool") |
| 7265 | (arguments |
| 7266 | (list #:builder |
| 7267 | (nerd-dictation-gexp "SOX" "XDOTOOL" |
| 7268 | (this-package-input "bash-minimal") |
| 7269 | (this-package-input "nerd-dictation")))) |
| 7270 | (propagated-inputs (list sox xdotool)))) |
| 7271 | |
| 7272 | (define-public nerd-dictation/sox-ydotool |
| 7273 | (package |
| 7274 | (inherit nerd-dictation/xdotool) |
| 7275 | (name "nerd-dictation-sox-ydotool") |
| 7276 | (arguments |
| 7277 | (list #:builder |
| 7278 | (nerd-dictation-gexp "SOX" "YDOTOOL" |
| 7279 | (this-package-input "bash-minimal") |
| 7280 | (this-package-input "nerd-dictation")))) |
| 7281 | (propagated-inputs (list sox ydotool)))) |
| 7282 | |
| 7283 | (define-public nerd-dictation/sox-wtype |
| 7284 | (package |
| 7285 | (inherit nerd-dictation/xdotool) |
| 7286 | (name "nerd-dictation-sox-wtype") |
| 7287 | (arguments |
| 7288 | (list #:builder |
| 7289 | (nerd-dictation-gexp "SOX" "WTYPE" |
| 7290 | (this-package-input "bash-minimal") |
| 7291 | (this-package-input "nerd-dictation")))) |
| 7292 | (propagated-inputs (list sox wtype)))) |
| 7293 | |
| 7294 | (define-public python-brian2 |
| 7295 | (package |
| 7296 | (name "python-brian2") |
| 7297 | (version "2.10.1") |
| 7298 | (source (origin |
| 7299 | (method url-fetch) |
| 7300 | (uri (pypi-uri "brian2" version)) |
| 7301 | (sha256 |
| 7302 | (base32 |
| 7303 | "0kqvw716q26liifjds807nxz17xbzrqwhvyp34j3d4i115xivln1")))) |
| 7304 | (build-system pyproject-build-system) |
| 7305 | (arguments |
| 7306 | (list |
| 7307 | ;; tests: 104 passed, 2 skipped, 16 warnings |
| 7308 | #:test-flags |
| 7309 | ;; Tests hang during collection with higher number of threads. |
| 7310 | #~(list "--numprocesses" (number->string (min 8 (parallel-job-count))) |
| 7311 | ;; XXX: The complete test suite is massive, run a simple test |
| 7312 | ;; that uses the main simulation elements and force code |
| 7313 | ;; generation to use Cython. |
| 7314 | "tests/test_synapses.py") |
| 7315 | #:phases |
| 7316 | #~(modify-phases %standard-phases |
| 7317 | (add-after 'unpack 'relax-requirements |
| 7318 | (lambda _ |
| 7319 | (substitute* "pyproject.toml" |
| 7320 | ;; 'cython>=0.29.21,<3.1.4', |
| 7321 | ((">=0.29.21,<3.1.4") "")))) |
| 7322 | (add-before 'check 'pre-check |
| 7323 | (lambda _ |
| 7324 | (setenv "HOME" "/tmp") |
| 7325 | (copy-recursively "brian2/tests" "tests") |
| 7326 | (delete-file-recursively "brian2")))))) |
| 7327 | (native-inputs |
| 7328 | (list python-pytest |
| 7329 | python-pytest-timeout |
| 7330 | python-pytest-xdist |
| 7331 | python-setuptools-scm)) |
| 7332 | (propagated-inputs |
| 7333 | (list python-cython |
| 7334 | python-jinja2 |
| 7335 | python-numpy |
| 7336 | python-packaging |
| 7337 | python-pyparsing |
| 7338 | python-setuptools |
| 7339 | python-sympy)) |
| 7340 | (home-page "https://briansimulator.org/") |
| 7341 | (synopsis "Clock-driven simulator for spiking neural networks") |
| 7342 | (description |
| 7343 | "Brian is a simulator for spiking neural networks written in Python. It |
| 7344 | is therefore designed to be easy to learn and use, highly flexible and |
| 7345 | easily extensible.") |
| 7346 | (license license:cecill))) |
| 7347 | |
| 7348 | (define-public python-brian2tools |
| 7349 | (package |
| 7350 | (name "python-brian2tools") |
| 7351 | (version "0.3") |
| 7352 | (source (origin |
| 7353 | (method url-fetch) |
| 7354 | (uri (pypi-uri "brian2tools" version)) |
| 7355 | (sha256 |
| 7356 | (base32 |
| 7357 | "0fn028mfy3qlzjkadd0wr5d7rcplijd5jphln414xifvvsb9jcc2")))) |
| 7358 | (build-system pyproject-build-system) |
| 7359 | ;; Both pypi tarball and git repo lack test files. |
| 7360 | (arguments (list #:tests? #f)) |
| 7361 | (propagated-inputs (list python-brian2 |
| 7362 | python-libneuroml |
| 7363 | python-markdown-strings |
| 7364 | python-matplotlib |
| 7365 | python-pylems |
| 7366 | python-tables)) |
| 7367 | (native-inputs (list python-setuptools python-setuptools-scm)) |
| 7368 | (home-page "https://github.com/brian-team/brian2tools") |
| 7369 | (synopsis "Tools for the Brian 2 simulator") |
| 7370 | (description "Visualization and NeuroML import/export tools for the |
| 7371 | Brian 2 simulator.") |
| 7372 | (license license:cecill))) |
| 7373 | |
| 7374 | (define-public mnn |
| 7375 | (package |
| 7376 | (name "mnn") |
| 7377 | (version "3.5.0") |
| 7378 | (source |
| 7379 | (origin |
| 7380 | (method git-fetch) |
| 7381 | (uri (git-reference |
| 7382 | (url "https://github.com/alibaba/MNN/") |
| 7383 | (commit version))) |
| 7384 | (snippet |
| 7385 | #~(begin |
| 7386 | (use-modules (guix build utils)) |
| 7387 | ;; Remove failing tests. |
| 7388 | (delete-file-recursively "test/expr/ModuleTest.cpp") |
| 7389 | (delete-file-recursively "test/op/ConvInt8Test.cpp") |
| 7390 | ;; Remove external libraries. Almost all of them are |
| 7391 | ;; available in Guix. |
| 7392 | (with-directory-excursion "3rd_party" |
| 7393 | (for-each delete-file-recursively |
| 7394 | '(;;"OpenCLHeaders"; for now does not build if in tree copy removed |
| 7395 | ;; "flatbuffers" ; for now does not build if in tree copy removed |
| 7396 | "protobuf" |
| 7397 | "rapidjson"))))) |
| 7398 | (file-name (git-file-name name version)) |
| 7399 | (sha256 |
| 7400 | (base32 "1jllq0zik55kznnks0kjqkilk362msgbz65qc3az5hvcrz34m8j7")))) |
| 7401 | (build-system cmake-build-system) |
| 7402 | (native-inputs (list googletest pkg-config python-minimal-wrapper)) |
| 7403 | (inputs (list ;; flatbuffers ;requires 1.10.2 |
| 7404 | fp16 |
| 7405 | glew |
| 7406 | glslang |
| 7407 | glu |
| 7408 | mesa |
| 7409 | mesa-headers |
| 7410 | mesa-opencl |
| 7411 | opencl-headers |
| 7412 | opencl-clhpp |
| 7413 | opencl-icd-loader |
| 7414 | opencv |
| 7415 | protobuf |
| 7416 | pthreadpool |
| 7417 | rapidjson ;requires 1.1.0 |
| 7418 | shaderc |
| 7419 | vulkan-headers |
| 7420 | vulkan-loader)) |
| 7421 | (arguments |
| 7422 | (list |
| 7423 | #:phases |
| 7424 | #~(modify-phases %standard-phases |
| 7425 | (replace 'check |
| 7426 | (lambda* (#:key tests? #:allow-other-keys) |
| 7427 | (when tests? |
| 7428 | (invoke "./run_test.out"))))) |
| 7429 | #:configure-flags |
| 7430 | #~(list "-DMNN_USE_SYSTEM_LIB=ON" |
| 7431 | "-DMNN_OPENMP=ON" |
| 7432 | ;; "-DMNN_OPENGL=ON" ; TO DO |
| 7433 | "-DMNN_OPENCL=ON" |
| 7434 | "-DMNN_VULKAN=ON" |
| 7435 | "-DMNN_SUPPORT_BF16=ON" |
| 7436 | "-DMNN_BUILD_TEST=ON" |
| 7437 | "-DMNN_BUILD_AUDIO=ON" |
| 7438 | "-DMNN_BUILD_OPENCV=ON" |
| 7439 | "-DMNN_BUILD_PROTOBUFFER=OFF" |
| 7440 | "-DMNN_IMGCODECS=ON" |
| 7441 | ;; "-DMNN_AUDIO_TEST=ON" |
| 7442 | ;; tests trying to fetch googletest and cannot build |
| 7443 | ;; "-DMNN_OPENCV_TEST=ON" |
| 7444 | ;; TO DO: remove the fetch in cmake and use system googletest |
| 7445 | "-DMNN_OPENCV_BENCH=ON" |
| 7446 | "-DMNN_BUILD_QUANTOOLS=ON" |
| 7447 | "-DMNN_EVALUATION=ON" |
| 7448 | "-DMNN_BUILD_BENCHMARK=ON" |
| 7449 | "-DMNN_BUILD_TOOLS=ON" |
| 7450 | "-DMNN_BUILD_CONVERTER=ON" |
| 7451 | "-DMNN_USE_SSE=ON" |
| 7452 | "-DMNN_AVX512=ON"))) ;TO DO: add Pytorch support |
| 7453 | (home-page "http://www.mnn.zone/") |
| 7454 | (synopsis "Fast, lightweight deep learning framework") |
| 7455 | (description |
| 7456 | "MNN is a deep learning framework. It supports inference and |
| 7457 | training of deep learning models for inference and training on-device.") |
| 7458 | (license license:asl2.0))) |
| 7459 | |
| 7460 | ;; This package needs to be in sync with llama-cpp version in use to avoid |
| 7461 | ;; incompatibilities. |
| 7462 | (define-public python-gguf |
| 7463 | (package |
| 7464 | (inherit llama-cpp) |
| 7465 | (name "python-gguf") |
| 7466 | (version "0.19.0") ;from gguf-py/pyproject.toml |
| 7467 | (build-system pyproject-build-system) |
| 7468 | (arguments |
| 7469 | (list |
| 7470 | #:phases |
| 7471 | #~(modify-phases %standard-phases |
| 7472 | (add-after 'unpack 'chdir |
| 7473 | (lambda _ |
| 7474 | (chdir "gguf-py")))))) |
| 7475 | (inputs '()) |
| 7476 | (propagated-inputs (list python-numpy python-pyyaml python-pyside-6 |
| 7477 | python-sentencepiece python-tqdm)) |
| 7478 | (native-inputs (list python-poetry-core python-pytest)) |
| 7479 | (properties '()) |
| 7480 | (home-page "https://ggml.ai") |
| 7481 | (synopsis "Read and write ML models in GGUF for GGML") |
| 7482 | (description "A Python library for reading and writing GGUF & GGML format |
| 7483 | ML models.") |
| 7484 | (license license:expat))) |
| 7485 | |
| 7486 | (define-public python-gymnasium |
| 7487 | (package |
| 7488 | (name "python-gymnasium") |
| 7489 | (version "0.29.1") |
| 7490 | (source |
| 7491 | (origin |
| 7492 | (method url-fetch) |
| 7493 | (uri (pypi-uri "gymnasium" version)) |
| 7494 | (sha256 |
| 7495 | (base32 "1cab4wsnlsxn2z90qmymv8ppmsq8yq2amiqwid3r0xfbxx92flqs")))) |
| 7496 | (build-system pyproject-build-system) |
| 7497 | (arguments |
| 7498 | (list |
| 7499 | #:phases |
| 7500 | #~(modify-phases %standard-phases |
| 7501 | (add-after 'unpack 'create-tests-module |
| 7502 | (lambda _ |
| 7503 | (with-output-to-file "tests/__init__.py" |
| 7504 | (lambda _ (display "")))))))) |
| 7505 | (propagated-inputs (list python-cloudpickle python-farama-notifications |
| 7506 | python-numpy python-typing-extensions)) |
| 7507 | (native-inputs |
| 7508 | (list python-pytest python-scipy python-setuptools python-wheel)) |
| 7509 | (home-page "https://gymnasium.farama.org/") |
| 7510 | (synopsis |
| 7511 | "Standard API for reinforcement learning and a set of reference environments") |
| 7512 | (description |
| 7513 | "This package provides a standard API for reinforcement learning and a |
| 7514 | diverse set of reference environments (formerly Gym).") |
| 7515 | (license license:expat))) |
| 7516 | |
| 7517 | ;; This will build dlib in the process of building python-dlib--and that |
| 7518 | ;; seems to be intended by upstream. Well, at least it probably optimizes |
| 7519 | ;; better that way. |
| 7520 | (define-public python-dlib |
| 7521 | (package |
| 7522 | (inherit dlib) |
| 7523 | (name "python-dlib") |
| 7524 | (build-system pyproject-build-system) |
| 7525 | (arguments |
| 7526 | (list |
| 7527 | #:phases |
| 7528 | #~(modify-phases %standard-phases |
| 7529 | (add-after 'unpack 'subst |
| 7530 | (lambda _ |
| 7531 | (substitute* "tools/python/CMakeLists.txt" |
| 7532 | (("\ |
| 7533 | add_subdirectory[(][.][.]/[.][.]/dlib/external/pybind11 pybind11_build[)]") |
| 7534 | "find_package(pybind11 CONFIG)"))))))) |
| 7535 | (native-inputs |
| 7536 | (list cmake-minimal |
| 7537 | perl |
| 7538 | pkg-config |
| 7539 | pybind11-2 |
| 7540 | python-pytest |
| 7541 | python-setuptools)))) |
| 7542 | |
| 7543 | ;; Version 0.16 breaks miopen |
| 7544 | (define-public frugally-deep |
| 7545 | (package |
| 7546 | (name "frugally-deep") |
| 7547 | (version "0.15.31") |
| 7548 | (source (origin |
| 7549 | (method git-fetch) |
| 7550 | (uri (git-reference |
| 7551 | (url "https://github.com/Dobiasd/frugally-deep") |
| 7552 | (commit (string-append "v" version)))) |
| 7553 | (file-name (git-file-name name version)) |
| 7554 | (sha256 |
| 7555 | (base32 |
| 7556 | "0gzkv03kz9p94a9apmds77npvxhmqyahhah4bq5yz3v9pvw6kx10")))) |
| 7557 | (build-system cmake-build-system) |
| 7558 | ;; Tests require tensorflow |
| 7559 | (arguments (list #:tests? #f)) |
| 7560 | (native-inputs (list eigen functionalplus nlohmann-json)) |
| 7561 | (home-page "https://github.com/Dobiasd/frugally-deep") |
| 7562 | (synopsis "Header-only library for using Keras (TensorFlow) models |
| 7563 | in C++") |
| 7564 | (description "@code{frugally-deep} is a header-only C++ library for |
| 7565 | inference of Keras machine learning models on a single CPU core.") |
| 7566 | (license license:expat))) |