gnu: emacs-sly: Update to 20200228.
[jackhill/guix/guix.git] / gnu / packages / machine-learning.scm
1 ;;; GNU Guix --- Functional package management for GNU
2 ;;; Copyright © 2015, 2016, 2017, 2018, 2019 Ricardo Wurmus <rekado@elephly.net>
3 ;;; Copyright © 2016 Efraim Flashner <efraim@flashner.co.il>
4 ;;; Copyright © 2016, 2017 Marius Bakke <mbakke@fastmail.com>
5 ;;; Copyright © 2016 Hartmut Goebel <h.goebel@crazy-compilers.com>
6 ;;; Copyright © 2018, 2019 Tobias Geerinckx-Rice <me@tobias.gr>
7 ;;; Copyright © 2018 Kei Kebreau <kkebreau@posteo.net>
8 ;;; Copyright © 2018 Mark Meyer <mark@ofosos.org>
9 ;;; Copyright © 2018 Ben Woodcroft <donttrustben@gmail.com>
10 ;;; Copyright © 2018 Fis Trivial <ybbs.daans@hotmail.com>
11 ;;; Copyright © 2018 Julien Lepiller <julien@lepiller.eu>
12 ;;; Copyright © 2018 Björn Höfling <bjoern.hoefling@bjoernhoefling.de>
13 ;;; Copyright © 2019 Nicolas Goaziou <mail@nicolasgoaziou.fr>
14 ;;; Copyright © 2019 Guillaume Le Vaillant <glv@posteo.net>
15 ;;; Copyright © 2019 Brett Gilio <brettg@gnu.org>
16 ;;;
17 ;;; This file is part of GNU Guix.
18 ;;;
19 ;;; GNU Guix is free software; you can redistribute it and/or modify it
20 ;;; under the terms of the GNU General Public License as published by
21 ;;; the Free Software Foundation; either version 3 of the License, or (at
22 ;;; your option) any later version.
23 ;;;
24 ;;; GNU Guix is distributed in the hope that it will be useful, but
25 ;;; WITHOUT ANY WARRANTY; without even the implied warranty of
26 ;;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
27 ;;; GNU General Public License for more details.
28 ;;;
29 ;;; You should have received a copy of the GNU General Public License
30 ;;; along with GNU Guix. If not, see <http://www.gnu.org/licenses/>.
31
32 (define-module (gnu packages machine-learning)
33 #:use-module ((guix licenses) #:prefix license:)
34 #:use-module (guix packages)
35 #:use-module (guix utils)
36 #:use-module (guix download)
37 #:use-module (guix svn-download)
38 #:use-module (guix build-system asdf)
39 #:use-module (guix build-system cmake)
40 #:use-module (guix build-system gnu)
41 #:use-module (guix build-system ocaml)
42 #:use-module (guix build-system python)
43 #:use-module (guix build-system r)
44 #:use-module (guix git-download)
45 #:use-module (gnu packages)
46 #:use-module (gnu packages adns)
47 #:use-module (gnu packages algebra)
48 #:use-module (gnu packages audio)
49 #:use-module (gnu packages autotools)
50 #:use-module (gnu packages base)
51 #:use-module (gnu packages bash)
52 #:use-module (gnu packages boost)
53 #:use-module (gnu packages check)
54 #:use-module (gnu packages compression)
55 #:use-module (gnu packages cran)
56 #:use-module (gnu packages databases)
57 #:use-module (gnu packages dejagnu)
58 #:use-module (gnu packages gcc)
59 #:use-module (gnu packages glib)
60 #:use-module (gnu packages graphviz)
61 #:use-module (gnu packages gstreamer)
62 #:use-module (gnu packages image)
63 #:use-module (gnu packages linux)
64 #:use-module (gnu packages lisp-xyz)
65 #:use-module (gnu packages maths)
66 #:use-module (gnu packages mpi)
67 #:use-module (gnu packages ocaml)
68 #:use-module (gnu packages onc-rpc)
69 #:use-module (gnu packages perl)
70 #:use-module (gnu packages pkg-config)
71 #:use-module (gnu packages protobuf)
72 #:use-module (gnu packages python)
73 #:use-module (gnu packages python-science)
74 #:use-module (gnu packages python-web)
75 #:use-module (gnu packages python-xyz)
76 #:use-module (gnu packages serialization)
77 #:use-module (gnu packages sphinx)
78 #:use-module (gnu packages statistics)
79 #:use-module (gnu packages sqlite)
80 #:use-module (gnu packages swig)
81 #:use-module (gnu packages tls)
82 #:use-module (gnu packages web)
83 #:use-module (gnu packages xml)
84 #:use-module (gnu packages xorg)
85 #:use-module (ice-9 match))
86
87 (define-public fann
88 ;; The last release is >100 commits behind, so we package from git.
89 (let ((commit "d71d54788bee56ba4cf7522801270152da5209d7"))
90 (package
91 (name "fann")
92 (version (string-append "2.2.0-1." (string-take commit 8)))
93 (source (origin
94 (method git-fetch)
95 (uri (git-reference
96 (url "https://github.com/libfann/fann.git")
97 (commit commit)))
98 (file-name (string-append name "-" version "-checkout"))
99 (sha256
100 (base32
101 "0ibwpfrjs6q2lijs8slxjgzb2llcl6rk3v2ski4r6215g5jjhg3x"))))
102 (build-system cmake-build-system)
103 (arguments
104 `(#:phases
105 (modify-phases %standard-phases
106 (replace 'check
107 (lambda* (#:key outputs #:allow-other-keys)
108 (let* ((out (assoc-ref outputs "out")))
109 (with-directory-excursion (string-append (getcwd) "/tests")
110 (invoke "./fann_tests"))))))))
111 (home-page "http://leenissen.dk/fann/wp/")
112 (synopsis "Fast Artificial Neural Network")
113 (description
114 "FANN is a neural network library, which implements multilayer
115 artificial neural networks in C with support for both fully connected and
116 sparsely connected networks.")
117 (license license:lgpl2.1))))
118
119 (define-public libsvm
120 (package
121 (name "libsvm")
122 (version "3.23")
123 (source
124 (origin
125 (method url-fetch)
126 (uri (string-append "https://www.csie.ntu.edu.tw/~cjlin/libsvm/"
127 name "-" version ".tar.gz"))
128 (sha256
129 (base32 "0jpaq0rr92x38p4nk3gjan79ip67m6p80anb28z1d8601miysyi5"))))
130 (build-system gnu-build-system)
131 (arguments
132 `(#:tests? #f ; no "check" target
133 #:phases (modify-phases %standard-phases
134 (delete 'configure)
135 (replace
136 'install ; no ‘install’ target
137 (lambda* (#:key outputs #:allow-other-keys)
138 (let* ((out (assoc-ref outputs "out"))
139 (bin (string-append out "/bin/")))
140 (mkdir-p bin)
141 (for-each (lambda (file)
142 (copy-file file (string-append bin file)))
143 '("svm-train"
144 "svm-predict"
145 "svm-scale")))
146 #t)))))
147 (home-page "https://www.csie.ntu.edu.tw/~cjlin/libsvm/")
148 (synopsis "Library for Support Vector Machines")
149 (description
150 "LIBSVM is a machine learning library for support vector
151 classification, (C-SVC, nu-SVC), regression (epsilon-SVR, nu-SVR) and
152 distribution estimation (one-class SVM). It supports multi-class
153 classification.")
154 (license license:bsd-3)))
155
156 (define-public python-libsvm
157 (package (inherit libsvm)
158 (name "python-libsvm")
159 (build-system gnu-build-system)
160 (arguments
161 `(#:tests? #f ; no "check" target
162 #:make-flags '("-C" "python")
163 #:phases
164 (modify-phases %standard-phases
165 (delete 'configure)
166 (replace
167 'install ; no ‘install’ target
168 (lambda* (#:key inputs outputs #:allow-other-keys)
169 (let ((site (string-append (assoc-ref outputs "out")
170 "/lib/python"
171 (string-take
172 (string-take-right
173 (assoc-ref inputs "python") 5) 3)
174 "/site-packages/")))
175 (substitute* "python/svm.py"
176 (("../libsvm.so.2") "libsvm.so.2"))
177 (mkdir-p site)
178 (for-each (lambda (file)
179 (copy-file file (string-append site (basename file))))
180 (find-files "python" "\\.py"))
181 (copy-file "libsvm.so.2"
182 (string-append site "libsvm.so.2")))
183 #t)))))
184 (inputs
185 `(("python" ,python)))
186 (synopsis "Python bindings of libSVM")))
187
188 (define-public ghmm
189 ;; The latest release candidate is several years and a couple of fixes have
190 ;; been published since. This is why we download the sources from the SVN
191 ;; repository.
192 (let ((svn-revision 2341))
193 (package
194 (name "ghmm")
195 (version (string-append "0.9-rc3-0." (number->string svn-revision)))
196 (source (origin
197 (method svn-fetch)
198 (uri (svn-reference
199 (url "http://svn.code.sf.net/p/ghmm/code/trunk")
200 (revision svn-revision)))
201 (file-name (string-append name "-" version "-checkout"))
202 (sha256
203 (base32
204 "0qbq1rqp94l530f043qzp8aw5lj7dng9wq0miffd7spd1ff638wq"))))
205 (build-system gnu-build-system)
206 (arguments
207 `(#:imported-modules (,@%gnu-build-system-modules
208 (guix build python-build-system))
209 #:modules ((guix build python-build-system)
210 ,@%gnu-build-system-modules)
211 #:phases
212 (modify-phases %standard-phases
213 (add-after 'unpack 'enter-dir
214 (lambda _ (chdir "ghmm") #t))
215 (delete 'check)
216 (add-after 'install 'check
217 (assoc-ref %standard-phases 'check))
218 (add-before 'check 'fix-PYTHONPATH
219 (lambda* (#:key inputs outputs #:allow-other-keys)
220 (let ((python-version (python-version
221 (assoc-ref inputs "python"))))
222 (setenv "PYTHONPATH"
223 (string-append (getenv "PYTHONPATH")
224 ":" (assoc-ref outputs "out")
225 "/lib/python" python-version
226 "/site-packages")))
227 #t))
228 (add-after 'enter-dir 'fix-runpath
229 (lambda* (#:key outputs #:allow-other-keys)
230 (substitute* "ghmmwrapper/setup.py"
231 (("^(.*)extra_compile_args = \\[" line indent)
232 (string-append indent
233 "extra_link_args = [\"-Wl,-rpath="
234 (assoc-ref outputs "out") "/lib\"],\n"
235 line
236 "\"-Wl,-rpath="
237 (assoc-ref outputs "out")
238 "/lib\", ")))
239 #t))
240 (add-after 'enter-dir 'disable-broken-tests
241 (lambda _
242 (substitute* "tests/Makefile.am"
243 ;; GHMM_SILENT_TESTS is assumed to be a command.
244 (("TESTS_ENVIRONMENT.*") "")
245 ;; Do not build broken tests.
246 (("chmm .*") "")
247 (("read_fa .*") "")
248 (("mcmc .*") "")
249 (("label_higher_order_test.*$")
250 "label_higher_order_test\n"))
251
252 ;; These Python unittests are broken as there is no gato.
253 ;; See https://sourceforge.net/p/ghmm/support-requests/3/
254 (substitute* "ghmmwrapper/ghmmunittests.py"
255 (("^(.*)def (testNewXML|testMultipleTransitionClasses|testNewXML)"
256 line indent)
257 (string-append indent
258 "@unittest.skip(\"Disabled by Guix\")\n"
259 line)))
260 #t)))))
261 (inputs
262 `(("python" ,python-2) ; only Python 2 is supported
263 ("libxml2" ,libxml2)))
264 (native-inputs
265 `(("pkg-config" ,pkg-config)
266 ("dejagnu" ,dejagnu)
267 ("swig" ,swig)
268 ("autoconf" ,autoconf)
269 ("automake" ,automake)
270 ("libtool" ,libtool)))
271 (home-page "http://ghmm.org")
272 (synopsis "Hidden Markov Model library")
273 (description
274 "The General Hidden Markov Model library (GHMM) is a C library with
275 additional Python bindings implementing a wide range of types of @dfn{Hidden
276 Markov Models} (HMM) and algorithms: discrete, continuous emissions, basic
277 training, HMM clustering, HMM mixtures.")
278 (license license:lgpl2.0+))))
279
280 (define-public mcl
281 (package
282 (name "mcl")
283 (version "14.137")
284 (source (origin
285 (method url-fetch)
286 (uri (string-append
287 "http://micans.org/mcl/src/mcl-"
288 (string-replace-substring version "." "-")
289 ".tar.gz"))
290 (sha256
291 (base32
292 "15xlax3z31lsn62vlg94hkm75nm40q4679amnfg13jm8m2bnhy5m"))))
293 (build-system gnu-build-system)
294 (arguments
295 `(#:configure-flags (list "--enable-blast")))
296 (inputs
297 `(("perl" ,perl)))
298 (home-page "http://micans.org/mcl/")
299 (synopsis "Clustering algorithm for graphs")
300 (description
301 "The MCL algorithm is short for the @dfn{Markov Cluster Algorithm}, a
302 fast and scalable unsupervised cluster algorithm for graphs (also known as
303 networks) based on simulation of (stochastic) flow in graphs.")
304 ;; In the LICENCE file and web page it says "The software is licensed
305 ;; under the GNU General Public License, version 3.", but in several of
306 ;; the source code files it suggests GPL3 or later.
307 ;; http://listserver.ebi.ac.uk/pipermail/mcl-users/2016/000376.html
308 (license license:gpl3)))
309
310 (define-public ocaml-mcl
311 (package
312 (name "ocaml-mcl")
313 (version "12-068oasis4")
314 (source
315 (origin
316 (method git-fetch)
317 (uri (git-reference
318 (url "https://github.com/fhcrc/mcl.git")
319 (commit version)))
320 (file-name (git-file-name name version))
321 (sha256
322 (base32
323 "0009dc3h2jp3qg5val452wngpqnbfyhbcxylghq0mrjqxx0jdq5p"))))
324 (build-system ocaml-build-system)
325 (arguments
326 `(#:phases
327 (modify-phases %standard-phases
328 (add-before 'configure 'patch-paths
329 (lambda _
330 (substitute* "configure"
331 (("/bin/sh") (which "sh")))
332 (substitute* "setup.ml"
333 (("LDFLAGS=-fPIC")
334 (string-append "LDFLAGS=-fPIC\"; \"SHELL=" (which "sh")))
335 (("-std=c89") "-std=gnu99")
336
337 ;; This is a mutable string, which is no longer supported. Use
338 ;; a byte buffer instead.
339 (("String.make \\(String.length s\\)")
340 "Bytes.make (String.length s)")
341
342 ;; These two belong together.
343 (("OASISString.replace_chars")
344 "Bytes.to_string (OASISString.replace_chars")
345 ((" s;")
346 " s);"))
347 (substitute* "myocamlbuild.ml"
348 (("std=c89") "std=gnu99"))
349 ;; Since we build with a more recent OCaml, we have to use C99 or
350 ;; later. This causes problems with the old C code.
351 (substitute* "src/impala/matrix.c"
352 (("restrict") "restrict_"))
353 #t)))))
354 (native-inputs
355 `(("ocamlbuild" ,ocamlbuild)))
356 (home-page "https://github.com/fhcrc/mcl")
357 (synopsis "OCaml wrappers around MCL")
358 (description
359 "This package provides OCaml bindings for the MCL graph clustering
360 algorithm.")
361 (license license:gpl3)))
362
363 (define-public randomjungle
364 (package
365 (name "randomjungle")
366 (version "2.1.0")
367 (source
368 (origin
369 (method url-fetch)
370 (uri (string-append
371 "https://www.imbs.uni-luebeck.de/fileadmin/files/Software"
372 "/randomjungle/randomjungle-" version ".tar_.gz"))
373 (patches (search-patches "randomjungle-disable-static-build.patch"))
374 (sha256
375 (base32
376 "12c8rf30cla71swx2mf4ww9mfd8jbdw5lnxd7dxhyw1ygrvg6y4w"))))
377 (build-system gnu-build-system)
378 (arguments
379 `(#:configure-flags
380 (list "--disable-static"
381 (string-append "--with-boost="
382 (assoc-ref %build-inputs "boost")))
383 #:phases
384 (modify-phases %standard-phases
385 (add-before
386 'configure 'set-CXXFLAGS
387 (lambda _
388 (setenv "CXXFLAGS" "-fpermissive ")
389 #t)))))
390 (inputs
391 `(("boost" ,boost)
392 ("gsl" ,gsl)
393 ("libxml2" ,libxml2)
394 ("zlib" ,zlib)))
395 (native-inputs
396 `(("gfortran" ,gfortran)
397 ("gfortran:lib" ,gfortran "lib")))
398 ;; Non-portable assembly instructions are used so building fails on
399 ;; platforms other than x86_64 or i686.
400 (supported-systems '("x86_64-linux" "i686-linux"))
401 (home-page "https://www.imbs.uni-luebeck.de/forschung/software/details.html#c224")
402 (synopsis "Implementation of the Random Forests machine learning method")
403 (description
404 "Random Jungle is an implementation of Random Forests. It is supposed to
405 analyse high dimensional data. In genetics, it can be used for analysing big
406 Genome Wide Association (GWA) data. Random Forests is a powerful machine
407 learning method. Most interesting features are variable selection, missing
408 value imputation, classifier creation, generalization error estimation and
409 sample proximities between pairs of cases.")
410 (license license:gpl3+)))
411
412 (define-public openfst
413 (package
414 (name "openfst")
415 (version "1.7.2")
416 (source (origin
417 (method url-fetch)
418 (uri (string-append "http://www.openfst.org/twiki/pub/FST/"
419 "FstDownload/openfst-" version ".tar.gz"))
420 (sha256
421 (base32
422 "0fqgk8195kz21is09gwzwnrg7fr9526bi9mh4apyskapz27pbhr1"))))
423 (build-system gnu-build-system)
424 (home-page "http://www.openfst.org")
425 (synopsis "Library for weighted finite-state transducers")
426 (description "OpenFst is a library for constructing, combining,
427 optimizing, and searching weighted finite-state transducers (FSTs).")
428 (license license:asl2.0)))
429
430 (define-public shogun
431 (package
432 (name "shogun")
433 (version "6.1.3")
434 (source
435 (origin
436 (method url-fetch)
437 (uri (string-append
438 "ftp://shogun-toolbox.org/shogun/releases/"
439 (version-major+minor version)
440 "/sources/shogun-" version ".tar.bz2"))
441 (sha256
442 (base32
443 "1rn9skm3nw6hr7mr3lgp2gfqhi7ii0lyxck7qmqnf8avq349s5jp"))
444 (modules '((guix build utils)
445 (ice-9 rdelim)))
446 (snippet
447 '(begin
448 ;; Remove non-free sources and files referencing them
449 (for-each delete-file
450 (find-files "src/shogun/classifier/svm/"
451 "SVMLight\\.(cpp|h)"))
452 (for-each delete-file
453 (find-files "examples/undocumented/libshogun/"
454 (string-append
455 "(classifier_.*svmlight.*|"
456 "evaluation_cross_validation_locked_comparison).cpp")))
457 ;; Remove non-free functions.
458 (define (delete-ifdefs file)
459 (with-atomic-file-replacement file
460 (lambda (in out)
461 (let loop ((line (read-line in 'concat))
462 (skipping? #f))
463 (if (eof-object? line)
464 #t
465 (let ((skip-next?
466 (or (and skipping?
467 (not (string-prefix?
468 "#endif //USE_SVMLIGHT" line)))
469 (string-prefix?
470 "#ifdef USE_SVMLIGHT" line))))
471 (when (or (not skipping?)
472 (and skipping? (not skip-next?)))
473 (display line out))
474 (loop (read-line in 'concat) skip-next?)))))))
475 (for-each delete-ifdefs
476 (append
477 (find-files "src/shogun/classifier/mkl"
478 "^MKLClassification\\.cpp")
479 (find-files "src/shogun/classifier/svm"
480 "^SVMLightOneClass\\.(cpp|h)")
481 (find-files "src/shogun/multiclass"
482 "^ScatterSVM\\.(cpp|h)")
483 (find-files "src/shogun/kernel/"
484 "^(Kernel|CombinedKernel|ProductKernel)\\.(cpp|h)")
485 (find-files "src/shogun/regression/svr"
486 "^(MKLRegression|SVRLight)\\.(cpp|h)")
487 (find-files "src/shogun/transfer/domain_adaptation"
488 "^DomainAdaptationSVM\\.(cpp|h)")))
489 #t))))
490 (build-system cmake-build-system)
491 (arguments
492 '(#:tests? #f ;no check target
493 #:phases
494 (modify-phases %standard-phases
495 (add-after 'unpack 'delete-broken-symlinks
496 (lambda _
497 (for-each delete-file '("applications/arts/data"
498 "applications/asp/data"
499 "applications/easysvm/data"
500 "applications/msplicer/data"
501 "applications/ocr/data"
502 "examples/meta/data"
503 "examples/undocumented/data"))
504 #t))
505 (add-after 'unpack 'change-R-target-path
506 (lambda* (#:key outputs #:allow-other-keys)
507 (substitute* '("src/interfaces/r/CMakeLists.txt"
508 "examples/meta/r/CMakeLists.txt")
509 (("\\$\\{R_COMPONENT_LIB_PATH\\}")
510 (string-append (assoc-ref outputs "out")
511 "/lib/R/library/")))
512 #t))
513 (add-after 'unpack 'fix-octave-modules
514 (lambda* (#:key outputs #:allow-other-keys)
515 (substitute* "src/interfaces/octave/CMakeLists.txt"
516 (("^include_directories\\(\\$\\{OCTAVE_INCLUDE_DIRS\\}")
517 "include_directories(${OCTAVE_INCLUDE_DIRS} ${OCTAVE_INCLUDE_DIRS}/octave")
518 ;; change target directory
519 (("\\$\\{OCTAVE_OCT_LOCAL_API_FILE_DIR\\}")
520 (string-append (assoc-ref outputs "out")
521 "/share/octave/packages")))
522 (substitute* '("src/interfaces/octave/swig_typemaps.i"
523 "src/interfaces/octave/sg_print_functions.cpp")
524 ;; "octave/config.h" and "octave/oct-obj.h" deprecated in Octave.
525 (("octave/config\\.h") "octave/octave-config.h")
526 (("octave/oct-obj.h") "octave/ovl.h"))
527 #t))
528 (add-after 'unpack 'move-rxcpp
529 (lambda* (#:key inputs #:allow-other-keys)
530 (let ((rxcpp-dir "shogun/third-party/rxcpp"))
531 (mkdir-p rxcpp-dir)
532 (install-file (assoc-ref inputs "rxcpp") rxcpp-dir)
533 #t)))
534 (add-before 'build 'set-HOME
535 ;; $HOME needs to be set at some point during the build phase
536 (lambda _ (setenv "HOME" "/tmp") #t)))
537 #:configure-flags
538 (list "-DCMAKE_BUILD_WITH_INSTALL_RPATH=TRUE"
539 "-DUSE_SVMLIGHT=OFF" ;disable proprietary SVMLIGHT
540 "-DBUILD_META_EXAMPLES=OFF" ;requires unpackaged ctags
541 ;;"-DINTERFACE_JAVA=ON" ;requires unpackaged jblas
542 ;;"-DINTERFACE_RUBY=ON" ;requires unpackaged ruby-narray
543 ;;"-DINTERFACE_PERL=ON" ;"FindPerlLibs" does not exist
544 ;;"-DINTERFACE_LUA=ON" ;fails because lua doesn't build pkgconfig file
545 "-DINTERFACE_OCTAVE=ON"
546 "-DINTERFACE_PYTHON=ON"
547 "-DINTERFACE_R=ON")))
548 (inputs
549 `(("python" ,python)
550 ("numpy" ,python-numpy)
551 ("r-minimal" ,r-minimal)
552 ("octave" ,octave-cli)
553 ("swig" ,swig)
554 ("eigen" ,eigen)
555 ("hdf5" ,hdf5)
556 ("atlas" ,atlas)
557 ("arpack" ,arpack-ng)
558 ("lapack" ,lapack)
559 ("glpk" ,glpk)
560 ("libxml2" ,libxml2)
561 ("lzo" ,lzo)
562 ("zlib" ,zlib)))
563 (native-inputs
564 `(("pkg-config" ,pkg-config)
565 ("rxcpp" ,rxcpp)))
566 ;; Non-portable SSE instructions are used so building fails on platforms
567 ;; other than x86_64.
568 (supported-systems '("x86_64-linux"))
569 (home-page "http://shogun-toolbox.org/")
570 (synopsis "Machine learning toolbox")
571 (description
572 "The Shogun Machine learning toolbox provides a wide range of unified and
573 efficient Machine Learning (ML) methods. The toolbox seamlessly allows to
574 combine multiple data representations, algorithm classes, and general purpose
575 tools. This enables both rapid prototyping of data pipelines and extensibility
576 in terms of new algorithms.")
577 (license license:gpl3+)))
578
579 (define-public rxcpp
580 (package
581 (name "rxcpp")
582 (version "4.1.0")
583 (source
584 (origin
585 (method git-fetch)
586 (uri (git-reference
587 (url "https://github.com/ReactiveX/RxCpp.git")
588 (commit (string-append "v" version))))
589 (sha256
590 (base32 "1rdpa3jlc181jd08nk437aar085h28i45s6nzrv65apb3xyyz0ij"))
591 (file-name (git-file-name name version))))
592 (build-system cmake-build-system)
593 (arguments
594 `(#:phases
595 (modify-phases %standard-phases
596 (add-after 'unpack 'remove-werror
597 (lambda _
598 (substitute* (find-files ".")
599 (("-Werror") ""))
600 #t))
601 (replace 'check
602 (lambda _
603 (invoke "ctest"))))))
604 (native-inputs
605 `(("catch" ,catch-framework)))
606 (home-page "http://reactivex.io/")
607 (synopsis "Reactive Extensions for C++")
608 (description
609 "The Reactive Extensions for C++ (RxCpp) is a library of algorithms for
610 values-distributed-in-time. ReactiveX is a library for composing asynchronous
611 and event-based programs by using observable sequences.
612
613 It extends the observer pattern to support sequences of data and/or events and
614 adds operators that allow you to compose sequences together declaratively while
615 abstracting away concerns about things like low-level threading,
616 synchronization, thread-safety, concurrent data structures, and non-blocking
617 I/O.")
618 (license license:asl2.0)))
619
620 (define-public r-adaptivesparsity
621 (package
622 (name "r-adaptivesparsity")
623 (version "1.6")
624 (source (origin
625 (method url-fetch)
626 (uri (cran-uri "AdaptiveSparsity" version))
627 (sha256
628 (base32
629 "0imr5m8mll9j6n4icsv6z9rl5kbnwsp9wvzrg7n90nnmcxq2cz91"))))
630 (properties
631 `((upstream-name . "AdaptiveSparsity")))
632 (build-system r-build-system)
633 (arguments
634 `(#:phases
635 (modify-phases %standard-phases
636 (add-after 'unpack 'link-against-armadillo
637 (lambda _
638 (substitute* "src/Makevars"
639 (("PKG_LIBS=" prefix)
640 (string-append prefix "-larmadillo"))))))))
641 (propagated-inputs
642 `(("r-mass" ,r-mass)
643 ("r-matrix" ,r-matrix)
644 ("r-rcpp" ,r-rcpp)
645 ("r-rcpparmadillo" ,r-rcpparmadillo)))
646 (inputs
647 `(("armadillo" ,armadillo)))
648 (home-page "https://cran.r-project.org/web/packages/AdaptiveSparsity")
649 (synopsis "Adaptive sparsity models")
650 (description
651 "This package implements the Figueiredo machine learning algorithm for
652 adaptive sparsity and the Wong algorithm for adaptively sparse gaussian
653 geometric models.")
654 (license license:lgpl3+)))
655
656 (define-public gemmlowp-for-tensorflow
657 ;; The commit hash is taken from "tensorflow/workspace.bzl".
658 (let ((commit "38ebac7b059e84692f53e5938f97a9943c120d98")
659 (revision "2"))
660 (package
661 (name "gemmlowp")
662 (version (git-version "0" revision commit))
663 (source (origin
664 (method url-fetch)
665 (uri (string-append "https://mirror.bazel.build/"
666 "github.com/google/gemmlowp/archive/"
667 commit ".zip"))
668 (file-name (string-append "gemmlowp-" version ".zip"))
669 (sha256
670 (base32
671 "0n56s2g8hrssm4w8qj1v58gfm56a04n9v992ixkmvk6zjiralzxq"))))
672 (build-system cmake-build-system)
673 (arguments
674 `(#:configure-flags
675 (list ,@(match (%current-system)
676 ((or "x86_64-linux" "i686-linux")
677 '("-DCMAKE_CXX_FLAGS=-msse2"))
678 (_ '())))
679 #:phases
680 (modify-phases %standard-phases
681 ;; This directory contains the CMakeLists.txt.
682 (add-after 'unpack 'chdir
683 (lambda _ (chdir "contrib") #t))
684 ;; There is no install target
685 (replace 'install
686 (lambda* (#:key outputs #:allow-other-keys)
687 (let* ((out (assoc-ref outputs "out"))
688 (lib (string-append out "/lib/"))
689 (inc (string-append out "/include/")))
690 (install-file "../build/libeight_bit_int_gemm.so" lib)
691 (for-each (lambda (dir)
692 (let ((target (string-append inc "/" dir)))
693 (mkdir-p target)
694 (for-each (lambda (h)
695 (install-file h target))
696 (find-files (string-append "../" dir)
697 "\\.h$"))))
698 '("meta" "profiling" "public" "fixedpoint"
699 "eight_bit_int_gemm" "internal"))
700 #t))))))
701 (native-inputs
702 `(("unzip" ,unzip)))
703 (home-page "https://github.com/google/gemmlowp")
704 (synopsis "Small self-contained low-precision GEMM library")
705 (description
706 "This is a small self-contained low-precision @dfn{general matrix
707 multiplication} (GEMM) library. It is not a full linear algebra library.
708 Low-precision means that the input and output matrix entries are integers on
709 at most 8 bits. To avoid overflow, results are internally accumulated on more
710 than 8 bits, and at the end only some significant 8 bits are kept.")
711 (license license:asl2.0))))
712
713 (define-public dlib
714 (package
715 (name "dlib")
716 (version "19.7")
717 (source (origin
718 (method url-fetch)
719 (uri (string-append
720 "http://dlib.net/files/dlib-" version ".tar.bz2"))
721 (sha256
722 (base32
723 "1mljz02kwkrbggyncxv5fpnyjdybw2qihaacb3js8yfkw12vwpc2"))
724 (modules '((guix build utils)))
725 (snippet
726 '(begin
727 ;; Delete ~13MB of bundled dependencies.
728 (delete-file-recursively "dlib/external")
729 (delete-file-recursively "docs/dlib/external")
730 #t))))
731 (build-system cmake-build-system)
732 (arguments
733 `(#:phases
734 (modify-phases %standard-phases
735 (add-after 'unpack 'disable-asserts
736 (lambda _
737 ;; config.h recommends explicitly enabling or disabling asserts
738 ;; when building as a shared library. By default neither is set.
739 (substitute* "dlib/config.h"
740 (("^//#define DLIB_DISABLE_ASSERTS") "#define DLIB_DISABLE_ASSERTS"))
741 #t))
742 (add-after 'disable-asserts 'disable-failing-tests
743 (lambda _
744 ;; One test times out on MIPS, so we need to disable it.
745 ;; Others are flaky on some platforms.
746 (let* ((system ,(or (%current-target-system)
747 (%current-system)))
748 (disabled-tests (cond
749 ((string-prefix? "mips64" system)
750 '("object_detector" ; timeout
751 "data_io"))
752 ((string-prefix? "armhf" system)
753 '("learning_to_track"))
754 ((string-prefix? "i686" system)
755 '("optimization"))
756 (else '()))))
757 (for-each
758 (lambda (test)
759 (substitute* "dlib/test/makefile"
760 (((string-append "SRC \\+= " test "\\.cpp")) "")))
761 disabled-tests)
762 #t)))
763 (replace 'check
764 (lambda _
765 ;; No test target, so we build and run the unit tests here.
766 (let ((test-dir (string-append "../dlib-" ,version "/dlib/test")))
767 (with-directory-excursion test-dir
768 (invoke "make" "-j" (number->string (parallel-job-count)))
769 (invoke "./dtest" "--runall"))
770 #t)))
771 (add-after 'install 'delete-static-library
772 (lambda* (#:key outputs #:allow-other-keys)
773 (delete-file (string-append (assoc-ref outputs "out")
774 "/lib/libdlib.a"))
775 #t)))))
776 (native-inputs
777 `(("pkg-config" ,pkg-config)
778 ;; For tests.
779 ("libnsl" ,libnsl)))
780 (inputs
781 `(("giflib" ,giflib)
782 ("lapack" ,lapack)
783 ("libjpeg" ,libjpeg)
784 ("libpng" ,libpng)
785 ("libx11" ,libx11)
786 ("openblas" ,openblas)
787 ("zlib" ,zlib)))
788 (synopsis
789 "Toolkit for making machine learning and data analysis applications in C++")
790 (description
791 "Dlib is a modern C++ toolkit containing machine learning algorithms and
792 tools. It is used in both industry and academia in a wide range of domains
793 including robotics, embedded devices, mobile phones, and large high performance
794 computing environments.")
795 (home-page "http://dlib.net")
796 (license license:boost1.0)))
797
798 (define-public python-scikit-learn
799 (package
800 (name "python-scikit-learn")
801 (version "0.22.1")
802 (source
803 (origin
804 (method git-fetch)
805 (uri (git-reference
806 (url "https://github.com/scikit-learn/scikit-learn.git")
807 (commit version)))
808 (file-name (git-file-name name version))
809 (sha256
810 (base32
811 "1xqxv210gsmjw094vc5ghq2y9lmm74qkk22pq6flcjzj51b86jxf"))))
812 (build-system python-build-system)
813 (arguments
814 `(#:phases
815 (modify-phases %standard-phases
816 (add-after 'build 'build-ext
817 (lambda _ (invoke "python" "setup.py" "build_ext" "--inplace") #t))
818 (replace 'check
819 (lambda _
820 ;; Restrict OpenBLAS threads to prevent segfaults while testing!
821 (setenv "OPENBLAS_NUM_THREADS" "1")
822
823 ;; Some tests require write access to $HOME.
824 (setenv "HOME" "/tmp")
825
826 (invoke "pytest" "sklearn" "-m" "not network")))
827 (add-before 'reset-gzip-timestamps 'make-files-writable
828 (lambda* (#:key outputs #:allow-other-keys)
829 ;; Make sure .gz files are writable so that the
830 ;; 'reset-gzip-timestamps' phase can do its work.
831 (let ((out (assoc-ref outputs "out")))
832 (for-each make-file-writable
833 (find-files out "\\.gz$"))
834 #t))))))
835 (inputs
836 `(("openblas" ,openblas)))
837 (native-inputs
838 `(("python-joblib" ,python-joblib)
839 ("python-pytest" ,python-pytest)
840 ("python-pandas" ,python-pandas) ;for tests
841 ("python-cython" ,python-cython)))
842 (propagated-inputs
843 `(("python-numpy" ,python-numpy)
844 ("python-scipy" ,python-scipy)))
845 (home-page "http://scikit-learn.org/")
846 (synopsis "Machine Learning in Python")
847 (description
848 "Scikit-learn provides simple and efficient tools for data mining and
849 data analysis.")
850 (license license:bsd-3)))
851
852 (define-public python2-scikit-learn
853 (package-with-python2 python-scikit-learn))
854
855 (define-public python-autograd
856 (let* ((commit "442205dfefe407beffb33550846434baa90c4de7")
857 (revision "0")
858 (version (git-version "0.0.0" revision commit)))
859 (package
860 (name "python-autograd")
861 (home-page "https://github.com/HIPS/autograd")
862 (source (origin
863 (method git-fetch)
864 (uri (git-reference
865 (url home-page)
866 (commit commit)))
867 (sha256
868 (base32
869 "189sv2xb0mwnjawa9z7mrgdglc1miaq93pnck26r28fi1jdwg0z4"))
870 (file-name (git-file-name name version))))
871 (version version)
872 (build-system python-build-system)
873 (native-inputs
874 `(("python-nose" ,python-nose)
875 ("python-pytest" ,python-pytest)))
876 (propagated-inputs
877 `(("python-future" ,python-future)
878 ("python-numpy" ,python-numpy)))
879 (arguments
880 `(#:phases (modify-phases %standard-phases
881 (replace 'check
882 (lambda _
883 (invoke "py.test" "-v"))))))
884 (synopsis "Efficiently computes derivatives of NumPy code")
885 (description "Autograd can automatically differentiate native Python and
886 NumPy code. It can handle a large subset of Python's features, including loops,
887 ifs, recursion and closures, and it can even take derivatives of derivatives
888 of derivatives. It supports reverse-mode differentiation
889 (a.k.a. backpropagation), which means it can efficiently take gradients of
890 scalar-valued functions with respect to array-valued arguments, as well as
891 forward-mode differentiation, and the two can be composed arbitrarily. The
892 main intended application of Autograd is gradient-based optimization.")
893 (license license:expat))))
894
895 (define-public python2-autograd
896 (package-with-python2 python-autograd))
897
898 (define-public lightgbm
899 (package
900 (name "lightgbm")
901 (version "2.0.12")
902 (source (origin
903 (method url-fetch)
904 (uri (string-append
905 "https://github.com/Microsoft/LightGBM/archive/v"
906 version ".tar.gz"))
907 (sha256
908 (base32
909 "132zf0yk0545mg72hyzxm102g3hpb6ixx9hnf8zd2k55gas6cjj1"))
910 (file-name (string-append name "-" version ".tar.gz"))))
911 (native-inputs
912 `(("python-pytest" ,python-pytest)
913 ("python-nose" ,python-nose)))
914 (inputs
915 `(("openmpi" ,openmpi)))
916 (propagated-inputs
917 `(("python-numpy" ,python-numpy)
918 ("python-scipy" ,python-scipy)))
919 (arguments
920 `(#:configure-flags
921 '("-DUSE_MPI=ON")
922 #:phases
923 (modify-phases %standard-phases
924 (replace 'check
925 (lambda* (#:key outputs #:allow-other-keys)
926 (with-directory-excursion ,(string-append "../LightGBM-" version)
927 (invoke "pytest" "tests/c_api_test/test_.py")))))))
928 (build-system cmake-build-system)
929 (home-page "https://github.com/Microsoft/LightGBM")
930 (synopsis "Gradient boosting framework based on decision tree algorithms")
931 (description "LightGBM is a gradient boosting framework that uses tree
932 based learning algorithms. It is designed to be distributed and efficient with
933 the following advantages:
934
935 @itemize
936 @item Faster training speed and higher efficiency
937 @item Lower memory usage
938 @item Better accuracy
939 @item Parallel and GPU learning supported (not enabled in this package)
940 @item Capable of handling large-scale data
941 @end itemize\n")
942 (license license:expat)))
943
944 (define-public vowpal-wabbit
945 ;; Language bindings not included.
946 (package
947 (name "vowpal-wabbit")
948 (version "8.5.0")
949 (source (origin
950 (method url-fetch)
951 (uri (string-append
952 "https://github.com/JohnLangford/vowpal_wabbit/archive/"
953 version ".tar.gz"))
954 (sha256
955 (base32
956 "0clp2kb7rk5sckhllxjr5a651awf4s8dgzg4659yh4hf5cqnf0gr"))
957 (file-name (string-append name "-" version ".tar.gz"))))
958 (inputs
959 `(("boost" ,boost)
960 ("zlib" ,zlib)))
961 (arguments
962 `(#:configure-flags
963 (list (string-append "--with-boost="
964 (assoc-ref %build-inputs "boost")))))
965 (build-system gnu-build-system)
966 (home-page "https://github.com/JohnLangford/vowpal_wabbit")
967 (synopsis "Fast machine learning library for online learning")
968 (description "Vowpal Wabbit is a machine learning system with techniques
969 such as online, hashing, allreduce, reductions, learning2search, active, and
970 interactive learning.")
971 (license license:bsd-3)))
972
973 (define-public python2-fastlmm
974 (package
975 (name "python2-fastlmm")
976 (version "0.2.21")
977 (source
978 (origin
979 (method url-fetch)
980 (uri (pypi-uri "fastlmm" version ".zip"))
981 (sha256
982 (base32
983 "1q8c34rpmwkfy3r4d5172pzdkpfryj561897z9r3x22gq7813x1m"))))
984 (build-system python-build-system)
985 (arguments
986 `(#:tests? #f ; some test files are missing
987 #:python ,python-2)) ; only Python 2.7 is supported
988 (propagated-inputs
989 `(("python2-numpy" ,python2-numpy)
990 ("python2-scipy" ,python2-scipy)
991 ("python2-matplotlib" ,python2-matplotlib)
992 ("python2-pandas" ,python2-pandas)
993 ("python2-scikit-learn" ,python2-scikit-learn)
994 ("python2-pysnptools" ,python2-pysnptools)))
995 (native-inputs
996 `(("unzip" ,unzip)
997 ("python2-cython" ,python2-cython)
998 ("python2-mock" ,python2-mock)
999 ("python2-nose" ,python2-nose)))
1000 (home-page "http://research.microsoft.com/en-us/um/redmond/projects/mscompbio/fastlmm/")
1001 (synopsis "Perform genome-wide association studies on large data sets")
1002 (description
1003 "FaST-LMM, which stands for Factored Spectrally Transformed Linear Mixed
1004 Models, is a program for performing both single-SNP and SNP-set genome-wide
1005 association studies (GWAS) on extremely large data sets.")
1006 (license license:asl2.0)))
1007
1008 ;; There have been no proper releases yet.
1009 (define-public kaldi
1010 (let ((commit "2f95609f0bb085bd3a1dc5eb0a39f3edea59e606")
1011 (revision "1"))
1012 (package
1013 (name "kaldi")
1014 (version (git-version "0" revision commit))
1015 (source (origin
1016 (method git-fetch)
1017 (uri (git-reference
1018 (url "https://github.com/kaldi-asr/kaldi.git")
1019 (commit commit)))
1020 (file-name (git-file-name name version))
1021 (sha256
1022 (base32
1023 "082qh3pfi7hvncylp4xsmkfahbd7gb0whdfa4rwrx7fxk9rdh3kz"))))
1024 (build-system gnu-build-system)
1025 (arguments
1026 `(#:test-target "test"
1027 #:phases
1028 (modify-phases %standard-phases
1029 (add-after 'unpack 'chdir
1030 (lambda _ (chdir "src") #t))
1031 (replace 'configure
1032 (lambda* (#:key build system inputs outputs #:allow-other-keys)
1033 (when (not (or (string-prefix? "x86_64" system)
1034 (string-prefix? "i686" system)))
1035 (substitute* "makefiles/linux_openblas.mk"
1036 (("-msse -msse2") "")))
1037 (substitute* "makefiles/default_rules.mk"
1038 (("/bin/bash") (which "bash")))
1039 (substitute* "Makefile"
1040 (("ext_depend: check_portaudio")
1041 "ext_depend:"))
1042 (substitute* '("online/Makefile"
1043 "onlinebin/Makefile"
1044 "gst-plugin/Makefile")
1045 (("../../tools/portaudio/install")
1046 (assoc-ref inputs "portaudio")))
1047
1048 ;; This `configure' script doesn't support variables passed as
1049 ;; arguments, nor does it support "prefix".
1050 (let ((out (assoc-ref outputs "out"))
1051 (openblas (assoc-ref inputs "openblas"))
1052 (openfst (assoc-ref inputs "openfst")))
1053 (substitute* "configure"
1054 (("check_for_slow_expf;") "")
1055 ;; This affects the RPATH and also serves as the installation
1056 ;; directory.
1057 (("KALDILIBDIR=`pwd`/lib")
1058 (string-append "KALDILIBDIR=" out "/lib")))
1059 (mkdir-p out) ; must exist
1060 (setenv "CONFIG_SHELL" (which "bash"))
1061 (setenv "OPENFST_VER" ,(package-version openfst))
1062 (invoke "./configure"
1063 "--use-cuda=no"
1064 "--shared"
1065 (string-append "--openblas-root=" openblas)
1066 (string-append "--fst-root=" openfst)))))
1067 (add-after 'build 'build-ext-and-gstreamer-plugin
1068 (lambda _
1069 (invoke "make" "-C" "online" "depend")
1070 (invoke "make" "-C" "online")
1071 (invoke "make" "-C" "onlinebin" "depend")
1072 (invoke "make" "-C" "onlinebin")
1073 (invoke "make" "-C" "gst-plugin" "depend")
1074 (invoke "make" "-C" "gst-plugin")
1075 #t))
1076 ;; TODO: also install the executables.
1077 (replace 'install
1078 (lambda* (#:key outputs #:allow-other-keys)
1079 (let* ((out (assoc-ref outputs "out"))
1080 (inc (string-append out "/include"))
1081 (lib (string-append out "/lib")))
1082 (mkdir-p lib)
1083 ;; The build phase installed symlinks to the actual
1084 ;; libraries. Install the actual targets.
1085 (for-each (lambda (file)
1086 (let ((target (readlink file)))
1087 (delete-file file)
1088 (install-file target lib)))
1089 (find-files lib "\\.so"))
1090 ;; Install headers
1091 (for-each (lambda (file)
1092 (let ((target-dir (string-append inc "/" (dirname file))))
1093 (install-file file target-dir)))
1094 (find-files "." "\\.h"))
1095 (install-file "gst-plugin/libgstonlinegmmdecodefaster.so"
1096 (string-append lib "/gstreamer-1.0"))
1097 #t))))))
1098 (inputs
1099 `(("alsa-lib" ,alsa-lib)
1100 ("gfortran" ,gfortran "lib")
1101 ("glib" ,glib)
1102 ("gstreamer" ,gstreamer)
1103 ("jack" ,jack-1)
1104 ("openblas" ,openblas)
1105 ("openfst" ,openfst)
1106 ("portaudio" ,portaudio)
1107 ("python" ,python)))
1108 (native-inputs
1109 `(("glib" ,glib "bin") ; glib-genmarshal
1110 ("grep" ,grep)
1111 ("sed" ,sed)
1112 ("pkg-config" ,pkg-config)
1113 ("which" ,which)))
1114 (home-page "https://kaldi-asr.org/")
1115 (synopsis "Speech recognition toolkit")
1116 (description "Kaldi is an extensible toolkit for speech recognition
1117 written in C++.")
1118 (license license:asl2.0))))
1119
1120 (define-public gst-kaldi-nnet2-online
1121 (let ((commit "617e43e73c7cc45eb9119028c02bd4178f738c4a")
1122 (revision "1"))
1123 (package
1124 (name "gst-kaldi-nnet2-online")
1125 (version (git-version "0" revision commit))
1126 (source (origin
1127 (method git-fetch)
1128 (uri (git-reference
1129 (url "https://github.com/alumae/gst-kaldi-nnet2-online.git")
1130 (commit commit)))
1131 (file-name (git-file-name name version))
1132 (sha256
1133 (base32
1134 "0xh3w67b69818s6ib02ara4lw7wamjdmh4jznvkpzrs4skbs9jx9"))))
1135 (build-system gnu-build-system)
1136 (arguments
1137 `(#:tests? #f ; there are none
1138 #:make-flags
1139 (list (string-append "SHELL="
1140 (assoc-ref %build-inputs "bash") "/bin/bash")
1141 (string-append "KALDI_ROOT="
1142 (assoc-ref %build-inputs "kaldi-src"))
1143 (string-append "KALDILIBDIR="
1144 (assoc-ref %build-inputs "kaldi") "/lib")
1145 "KALDI_FLAVOR=dynamic")
1146 #:phases
1147 (modify-phases %standard-phases
1148 (add-after 'unpack 'chdir
1149 (lambda _ (chdir "src") #t))
1150 (replace 'configure
1151 (lambda* (#:key inputs #:allow-other-keys)
1152 (let ((glib (assoc-ref inputs "glib")))
1153 (setenv "CXXFLAGS" "-fPIC")
1154 (setenv "CPLUS_INCLUDE_PATH"
1155 (string-append glib "/include/glib-2.0:"
1156 glib "/lib/glib-2.0/include:"
1157 (assoc-ref inputs "gstreamer")
1158 "/include/gstreamer-1.0")))
1159 (substitute* "Makefile"
1160 (("include \\$\\(KALDI_ROOT\\)/src/kaldi.mk") "")
1161 (("\\$\\(error Cannot find") "#"))
1162 #t))
1163 (add-before 'build 'build-depend
1164 (lambda* (#:key make-flags #:allow-other-keys)
1165 (apply invoke "make" "depend" make-flags)))
1166 (replace 'install
1167 (lambda* (#:key outputs #:allow-other-keys)
1168 (let* ((out (assoc-ref outputs "out"))
1169 (lib (string-append out "/lib/gstreamer-1.0")))
1170 (install-file "libgstkaldinnet2onlinedecoder.so" lib)
1171 #t))))))
1172 (inputs
1173 `(("glib" ,glib)
1174 ("gstreamer" ,gstreamer)
1175 ("jansson" ,jansson)
1176 ("openfst" ,openfst)
1177 ("kaldi" ,kaldi)))
1178 (native-inputs
1179 `(("bash" ,bash)
1180 ("glib:bin" ,glib "bin") ; glib-genmarshal
1181 ("kaldi-src" ,(package-source kaldi))
1182 ("pkg-config" ,pkg-config)))
1183 (home-page "https://kaldi-asr.org/")
1184 (synopsis "Gstreamer plugin for decoding speech")
1185 (description "This package provides a GStreamer plugin that wraps
1186 Kaldi's @code{SingleUtteranceNnet2Decoder}. It requires iVector-adapted DNN
1187 acoustic models. The iVectors are adapted to the current audio stream
1188 automatically.")
1189 (license license:asl2.0))))
1190
1191 (define-public kaldi-gstreamer-server
1192 (let ((commit "1735ba49c5dc0ebfc184e45105fc600cd9f1f508")
1193 (revision "1"))
1194 (package
1195 (name "kaldi-gstreamer-server")
1196 (version (git-version "0" revision commit))
1197 (source (origin
1198 (method git-fetch)
1199 (uri (git-reference
1200 (url "https://github.com/alumae/kaldi-gstreamer-server.git")
1201 (commit commit)))
1202 (file-name (git-file-name name version))
1203 (sha256
1204 (base32
1205 "0j701m7lbwmzqxsfanj882v7881hrbmpqybbczbxqpcbg8q34w0k"))))
1206 (build-system gnu-build-system)
1207 (arguments
1208 `(#:tests? #f ; there are no tests that can be run automatically
1209 #:modules ((guix build utils)
1210 (guix build gnu-build-system)
1211 (srfi srfi-26))
1212 #:phases
1213 (modify-phases %standard-phases
1214 (delete 'configure)
1215 (replace 'build
1216 (lambda* (#:key outputs #:allow-other-keys)
1217 ;; Disable hash randomization to ensure the generated .pycs
1218 ;; are reproducible.
1219 (setenv "PYTHONHASHSEED" "0")
1220 (with-directory-excursion "kaldigstserver"
1221 (for-each (lambda (file)
1222 (apply invoke
1223 `("python"
1224 "-m" "compileall"
1225 "-f" ; force rebuild
1226 ,file)))
1227 (find-files "." "\\.py$")))
1228 #t))
1229 (replace 'install
1230 (lambda* (#:key inputs outputs #:allow-other-keys)
1231 (let* ((out (assoc-ref outputs "out"))
1232 (bin (string-append out "/bin"))
1233 (share (string-append out "/share/kaldi-gstreamer-server/")))
1234 ;; Install Python files
1235 (with-directory-excursion "kaldigstserver"
1236 (for-each (cut install-file <> share)
1237 (find-files "." ".*")))
1238
1239 ;; Install sample configuration files
1240 (for-each (cut install-file <> share)
1241 (find-files "." "\\.yaml"))
1242
1243 ;; Install executables
1244 (mkdir-p bin)
1245 (let* ((server (string-append bin "/kaldi-gst-server"))
1246 (client (string-append bin "/kaldi-gst-client"))
1247 (worker (string-append bin "/kaldi-gst-worker"))
1248 (PYTHONPATH (getenv "PYTHONPATH"))
1249 (GST_PLUGIN_PATH (string-append
1250 (assoc-ref inputs "gst-kaldi-nnet2-online")
1251 "/lib/gstreamer-1.0:${GST_PLUGIN_PATH}"))
1252 (wrap (lambda (wrapper what)
1253 (with-output-to-file wrapper
1254 (lambda _
1255 (format #t
1256 "#!~a
1257 export PYTHONPATH=~a
1258 export GST_PLUGIN_PATH=~a
1259 exec ~a ~a/~a \"$@\"~%"
1260 (which "bash") PYTHONPATH GST_PLUGIN_PATH
1261 (which "python") share what)))
1262 (chmod wrapper #o555))))
1263 (for-each wrap
1264 (list server client worker)
1265 (list "master_server.py"
1266 "client.py"
1267 "worker.py")))
1268 #t))))))
1269 (inputs
1270 `(("gst-kaldi-nnet2-online" ,gst-kaldi-nnet2-online)
1271 ("python2" ,python-2)
1272 ("python2-futures" ,python2-futures)
1273 ("python2-pygobject" ,python2-pygobject)
1274 ("python2-pyyaml" ,python2-pyyaml)
1275 ("python2-tornado" ,python2-tornado)
1276 ("python2-ws4py" ,python2-ws4py-for-kaldi-gstreamer-server)))
1277 (home-page "https://github.com/alumae/kaldi-gstreamer-server")
1278 (synopsis "Real-time full-duplex speech recognition server")
1279 (description "This is a real-time full-duplex speech recognition server,
1280 based on the Kaldi toolkit and the GStreamer framework and implemented in
1281 Python.")
1282 (license license:bsd-2))))
1283
1284 (define-public grpc
1285 (package
1286 (name "grpc")
1287 (version "1.16.1")
1288 (source (origin
1289 (method git-fetch)
1290 (uri (git-reference
1291 (url "https://github.com/grpc/grpc.git")
1292 (commit (string-append "v" version))))
1293 (file-name (git-file-name name version))
1294 (sha256
1295 (base32
1296 "1jimqz3115f9pli5w6ik9wi7mjc7ix6y7yrq4a1ab9fc3dalj7p2"))))
1297 (build-system cmake-build-system)
1298 (arguments
1299 `(#:tests? #f ; no test target
1300 #:configure-flags
1301 (list "-DgRPC_ZLIB_PROVIDER=package"
1302 "-DgRPC_CARES_PROVIDER=package"
1303 "-DgRPC_SSL_PROVIDER=package"
1304 "-DgRPC_PROTOBUF_PROVIDER=package")))
1305 (inputs
1306 `(("c-ares" ,c-ares/cmake)
1307 ("openssl" ,openssl)
1308 ("zlib" ,zlib)))
1309 (native-inputs
1310 `(("protobuf" ,protobuf)
1311 ("python" ,python-wrapper)))
1312 (home-page "https://grpc.io")
1313 (synopsis "High performance universal RPC framework")
1314 (description "gRPC is a modern high performance @dfn{Remote Procedure Call}
1315 (RPC) framework that can run in any environment. It can efficiently connect
1316 services in and across data centers with pluggable support for load balancing,
1317 tracing, health checking and authentication. It is also applicable in last
1318 mile of distributed computing to connect devices, mobile applications and
1319 browsers to backend services.")
1320 (license license:asl2.0)))
1321
1322 ;; Note that Tensorflow includes a "third_party" directory, which seems to not
1323 ;; only contain modified subsets of upstream library source code, but also
1324 ;; adapter headers provided by Google (such as the fft.h header, which is not
1325 ;; part of the upstream project code). The Tensorflow code includes headers
1326 ;; from the "third_party" directory. It does not look like we can replace
1327 ;; these headers with unmodified upstream files, so we keep them.
1328 (define-public tensorflow
1329 (package
1330 (name "tensorflow")
1331 (version "1.9.0")
1332 (source
1333 (origin
1334 (method git-fetch)
1335 (uri (git-reference
1336 (url "https://github.com/tensorflow/tensorflow.git")
1337 (commit (string-append "v" version))))
1338 (file-name (string-append "tensorflow-" version "-checkout"))
1339 (sha256
1340 (base32
1341 "0a9kwha395g3wgxfwln5j8vn9nkspmd75xldrlqdq540w996g8xa"))))
1342 (build-system cmake-build-system)
1343 (arguments
1344 `(#:tests? #f ; no "check" target
1345 #:build-type "Release"
1346 #:configure-flags
1347 (let ((protobuf (assoc-ref %build-inputs "protobuf"))
1348 (protobuf:native (assoc-ref %build-inputs "protobuf:native"))
1349 (jsoncpp (assoc-ref %build-inputs "jsoncpp"))
1350 (snappy (assoc-ref %build-inputs "snappy"))
1351 (sqlite (assoc-ref %build-inputs "sqlite")))
1352 (list
1353 ;; Use protobuf from Guix
1354 (string-append "-Dprotobuf_STATIC_LIBRARIES="
1355 protobuf "/lib/libprotobuf.so")
1356 (string-append "-DPROTOBUF_PROTOC_EXECUTABLE="
1357 protobuf:native "/bin/protoc")
1358
1359 ;; Use snappy from Guix
1360 (string-append "-Dsnappy_STATIC_LIBRARIES="
1361 snappy "/lib/libsnappy.so")
1362 ;; Yes, this is not actually the include directory but a prefix...
1363 (string-append "-Dsnappy_INCLUDE_DIR=" snappy)
1364
1365 ;; Use jsoncpp from Guix
1366 (string-append "-Djsoncpp_STATIC_LIBRARIES="
1367 jsoncpp "/lib/libjsoncpp.so")
1368 ;; Yes, this is not actually the include directory but a prefix...
1369 (string-append "-Djsoncpp_INCLUDE_DIR=" jsoncpp)
1370
1371 ;; Use sqlite from Guix
1372 (string-append "-Dsqlite_STATIC_LIBRARIES="
1373 sqlite "/lib/libsqlite.a")
1374
1375 ;; Use system libraries wherever possible. Currently, this
1376 ;; only affects zlib.
1377 "-Dsystemlib_ALL=ON"
1378 "-Dtensorflow_ENABLE_POSITION_INDEPENDENT_CODE=ON"
1379 "-Dtensorflow_BUILD_SHARED_LIB=ON"
1380 "-Dtensorflow_OPTIMIZE_FOR_NATIVE_ARCH=OFF"
1381 "-Dtensorflow_ENABLE_SSL_SUPPORT=OFF"
1382 "-Dtensorflow_BUILD_CONTRIB_KERNELS=OFF"))
1383 #:make-flags
1384 (list "CC=gcc")
1385 #:modules ((ice-9 ftw)
1386 (guix build utils)
1387 (guix build cmake-build-system))
1388 #:phases
1389 (modify-phases %standard-phases
1390 (add-after 'unpack 'set-source-file-times-to-1980
1391 ;; At the end of the tf_python_build_pip_package target, a ZIP
1392 ;; archive should be generated via bdist_wheel, but it fails with
1393 ;; "ZIP does not support timestamps before 1980". Luckily,
1394 ;; SOURCE_DATE_EPOCH is respected, which we set to some time in
1395 ;; 1980.
1396 (lambda _ (setenv "SOURCE_DATE_EPOCH" "315532800") #t))
1397 ;; See https://github.com/tensorflow/tensorflow/issues/20517#issuecomment-406373913
1398 (add-after 'unpack 'python3.7-compatibility
1399 (lambda _
1400 (substitute* '("tensorflow/python/eager/pywrap_tfe_src.cc"
1401 "tensorflow/python/lib/core/ndarray_tensor.cc"
1402 "tensorflow/python/lib/core/py_func.cc")
1403 (("PyUnicode_AsUTF8") "(char *)PyUnicode_AsUTF8"))
1404 (substitute* "tensorflow/c/eager/c_api.h"
1405 (("unsigned char async")
1406 "unsigned char is_async"))
1407
1408 ;; Remove dependency on tensorboard, a complicated but probably
1409 ;; optional package.
1410 (substitute* "tensorflow/tools/pip_package/setup.py"
1411 ((".*'tensorboard >.*") ""))
1412 #t))
1413 (add-after 'python3.7-compatibility 'chdir
1414 (lambda _ (chdir "tensorflow/contrib/cmake") #t))
1415 (add-after 'chdir 'disable-downloads
1416 (lambda* (#:key inputs #:allow-other-keys)
1417 (substitute* (find-files "external" "\\.cmake$")
1418 (("GIT_REPOSITORY.*") "")
1419 (("GIT_TAG.*") "")
1420 (("PREFIX ")
1421 "DOWNLOAD_COMMAND \"\"\nPREFIX "))
1422
1423 ;; Use packages from Guix
1424 (let ((grpc (assoc-ref inputs "grpc")))
1425 (substitute* "CMakeLists.txt"
1426 ;; Sqlite
1427 (("include\\(sqlite\\)") "")
1428 (("\\$\\{sqlite_STATIC_LIBRARIES\\}")
1429 (string-append (assoc-ref inputs "sqlite")
1430 "/lib/libsqlite3.so"))
1431 (("sqlite_copy_headers_to_destination") "")
1432
1433 ;; PNG
1434 (("include\\(png\\)") "")
1435 (("\\$\\{png_STATIC_LIBRARIES\\}")
1436 (string-append (assoc-ref inputs "libpng")
1437 "/lib/libpng16.so"))
1438 (("png_copy_headers_to_destination") "")
1439
1440 ;; JPEG
1441 (("include\\(jpeg\\)") "")
1442 (("\\$\\{jpeg_STATIC_LIBRARIES\\}")
1443 (string-append (assoc-ref inputs "libjpeg")
1444 "/lib/libjpeg.so"))
1445 (("jpeg_copy_headers_to_destination") "")
1446
1447 ;; GIF
1448 (("include\\(gif\\)") "")
1449 (("\\$\\{gif_STATIC_LIBRARIES\\}")
1450 (string-append (assoc-ref inputs "giflib")
1451 "/lib/libgif.so"))
1452 (("gif_copy_headers_to_destination") "")
1453
1454 ;; lmdb
1455 (("include\\(lmdb\\)") "")
1456 (("\\$\\{lmdb_STATIC_LIBRARIES\\}")
1457 (string-append (assoc-ref inputs "lmdb")
1458 "/lib/liblmdb.so"))
1459 (("lmdb_copy_headers_to_destination") "")
1460
1461 ;; Protobuf
1462 (("include\\(protobuf\\)") "")
1463 (("protobuf_copy_headers_to_destination") "")
1464 (("^ +protobuf") "")
1465
1466 ;; gRPC
1467 (("include\\(grpc\\)")
1468 "find_package(grpc REQUIRED NAMES gRPC)")
1469 (("list\\(APPEND tensorflow_EXTERNAL_DEPENDENCIES grpc\\)") "")
1470
1471 ;; Eigen
1472 (("include\\(eigen\\)")
1473 (string-append "find_package(eigen REQUIRED NAMES Eigen3)
1474 set(eigen_INCLUDE_DIRS ${CMAKE_CURRENT_BINARY_DIR}/external/eigen_archive "
1475 (assoc-ref inputs "eigen") "/include/eigen3)"))
1476 (("^ +eigen") "")
1477
1478 ;; snappy
1479 (("include\\(snappy\\)")
1480 "add_definitions(-DTF_USE_SNAPPY)")
1481 (("list\\(APPEND tensorflow_EXTERNAL_DEPENDENCIES snappy\\)") "")
1482
1483 ;; jsoncpp
1484 (("include\\(jsoncpp\\)") "")
1485 (("^ +jsoncpp") ""))
1486
1487 (substitute* "tf_core_framework.cmake"
1488 ((" grpc") "")
1489 (("\\$\\{GRPC_BUILD\\}/grpc_cpp_plugin")
1490 (which "grpc_cpp_plugin"))
1491 ;; Link with gRPC libraries
1492 (("add_library\\(tf_protos_cc.*" m)
1493 (string-append m
1494 (format #f "\ntarget_link_libraries(tf_protos_cc PRIVATE \
1495 ~a/lib/libgrpc++_unsecure.a \
1496 ~a/lib/libgrpc_unsecure.a \
1497 ~a/lib/libaddress_sorting.a \
1498 ~a/lib/libgpr.a \
1499 ~a//lib/libcares.so
1500 )\n"
1501 grpc grpc grpc grpc
1502 (assoc-ref inputs "c-ares"))))))
1503 (substitute* "tf_tools.cmake"
1504 (("add_dependencies\\(\\$\\{proto_text.*") ""))
1505 ;; Remove dependency on bundled grpc
1506 (substitute* "tf_core_distributed_runtime.cmake"
1507 (("tf_core_cpu grpc") "tf_core_cpu"))
1508
1509 ;; This directory is a dependency of many targets.
1510 (mkdir-p "protobuf")
1511 #t))
1512 (add-after 'configure 'unpack-third-party-sources
1513 (lambda* (#:key inputs #:allow-other-keys)
1514 ;; This is needed to configure bundled packages properly.
1515 (setenv "CONFIG_SHELL" (which "bash"))
1516 (for-each
1517 (lambda (name)
1518 (let* ((what (assoc-ref inputs (string-append name "-src")))
1519 (name* (string-map (lambda (c)
1520 (if (char=? c #\-)
1521 #\_ c)) name))
1522 (where (string-append "../build/" name* "/src/" name*)))
1523 (cond
1524 ((string-suffix? ".zip" what)
1525 (mkdir-p where)
1526 (with-directory-excursion where
1527 (invoke "unzip" what)))
1528 ((string-suffix? ".tar.gz" what)
1529 (mkdir-p where)
1530 (invoke "tar" "xf" what
1531 "-C" where "--strip-components=1"))
1532 (else
1533 (let ((parent (dirname where)))
1534 (mkdir-p parent)
1535 (with-directory-excursion parent
1536 (when (file-exists? name*)
1537 (delete-file-recursively name*))
1538 (copy-recursively what name*)
1539 (map make-file-writable
1540 (find-files name* ".*"))))))))
1541 (list "boringssl"
1542 "cub"
1543 "double-conversion"
1544 "farmhash"
1545 "fft2d"
1546 "highwayhash"
1547 "nsync"
1548 "re2"))
1549
1550 (rename-file "../build/cub/src/cub/cub-1.8.0/"
1551 "../build/cub/src/cub/cub/")
1552 #t))
1553 (add-after 'unpack 'fix-python-build
1554 (lambda* (#:key inputs outputs #:allow-other-keys)
1555 (mkdir-p "protobuf-src")
1556 (invoke "tar" "xf" (assoc-ref inputs "protobuf:src")
1557 "-C" "protobuf-src" "--strip-components=1")
1558 (mkdir-p "eigen-src")
1559 (invoke "tar" "xf" (assoc-ref inputs "eigen:src")
1560 "-C" "eigen-src" "--strip-components=1")
1561
1562 (substitute* "tensorflow/contrib/cmake/tf_python.cmake"
1563 ;; Ensure that all Python dependencies can be found at build time.
1564 (("PYTHONPATH=\\$\\{CMAKE_CURRENT_BINARY_DIR\\}/tf_python" m)
1565 (string-append m ":" (getenv "PYTHONPATH")))
1566 ;; Take protobuf source files from our source package.
1567 (("\\$\\{CMAKE_CURRENT_BINARY_DIR\\}/protobuf/src/protobuf/src/google")
1568 (string-append (getcwd) "/protobuf-src/src/google")))
1569
1570 (substitute* '("tensorflow/contrib/cmake/tf_shared_lib.cmake"
1571 "tensorflow/contrib/cmake/tf_python.cmake")
1572 ;; Take Eigen source files from our source package.
1573 (("\\$\\{CMAKE_CURRENT_BINARY_DIR\\}/eigen/src/eigen/")
1574 (string-append (getcwd) "/eigen-src/"))
1575 ;; Take Eigen headers from our own package.
1576 (("\\$\\{CMAKE_CURRENT_BINARY_DIR\\}/external/eigen_archive")
1577 (string-append (assoc-ref inputs "eigen") "/include/eigen3")))
1578
1579 ;; Correct the RUNPATH of ops libraries generated for Python.
1580 ;; TODO: this doesn't work :(
1581 ;; /gnu/store/...-tensorflow-1.9.0/lib/python3.7/site-packages/tensorflow/contrib/seq2seq/python/ops/lib_beam_search_ops.so:
1582 ;; warning: RUNPATH contains bogus entries: ("/tmp/guix-build-tensorflow-1.9.0.drv-0/source/tensorflow/contrib/build")
1583 ;; /gnu/store/...-tensorflow-1.9.0/lib/python3.7/site-packages/tensorflow/contrib/seq2seq/python/ops/lib_beam_search_ops.so:
1584 ;; error: depends on 'libpywrap_tensorflow_internal.so', which
1585 ;; cannot be found in RUNPATH ...
1586 (substitute* "tensorflow/contrib/cmake/tf_cc_ops.cmake"
1587 (("set_target_properties.*")
1588 (string-append "set_target_properties(${_AT_TARGET} PROPERTIES \
1589 COMPILE_FLAGS ${target_compile_flags} \
1590 INSTALL_RPATH_USE_LINK_PATH TRUE \
1591 INSTALL_RPATH " (assoc-ref outputs "out") "/lib)\n")))
1592 #t))
1593 (add-after 'build 'build-pip-package
1594 (lambda* (#:key outputs #:allow-other-keys)
1595 (setenv "LDFLAGS"
1596 (string-append "-Wl,-rpath="
1597 (assoc-ref outputs "out") "/lib"))
1598 (invoke "make" "tf_python_build_pip_package")
1599 #t))
1600 (add-after 'build-pip-package 'install-python
1601 (lambda* (#:key outputs #:allow-other-keys)
1602 (let ((out (assoc-ref outputs "out"))
1603 (wheel (car (find-files "../build/tf_python/dist/" "\\.whl$"))))
1604 (invoke "python" "-m" "pip" "install" wheel
1605 (string-append "--prefix=" out))
1606
1607 ;; XXX: broken RUNPATH, see fix-python-build phase.
1608 (delete-file
1609 (string-append
1610 out "/lib/python3.7/site-packages/tensorflow/contrib/"
1611 "seq2seq/python/ops/lib_beam_search_ops.so"))
1612 #t))))))
1613 (native-inputs
1614 `(("pkg-config" ,pkg-config)
1615 ("protobuf:native" ,protobuf-3.6) ; protoc
1616 ("protobuf:src" ,(package-source protobuf-3.6))
1617 ("eigen:src" ,(package-source eigen-for-tensorflow))
1618 ;; install_pip_packages.sh wants setuptools 39.1.0 specifically.
1619 ("python-setuptools" ,python-setuptools-for-tensorflow)
1620
1621 ;; The commit hashes and URLs for third-party source code are taken
1622 ;; from "tensorflow/workspace.bzl".
1623 ("boringssl-src"
1624 ,(let ((commit "ee7aa02")
1625 (revision "1"))
1626 (origin
1627 (method git-fetch)
1628 (uri (git-reference
1629 (url "https://boringssl.googlesource.com/boringssl")
1630 (commit commit)))
1631 (file-name (string-append "boringssl-0-" revision
1632 (string-take commit 7)
1633 "-checkout"))
1634 (sha256
1635 (base32
1636 "1jf693q0nw0adsic6cgmbdx6g7wr4rj4vxa8j1hpn792fqhd8wgw")))))
1637 ("cub-src"
1638 ,(let ((version "1.8.0"))
1639 (origin
1640 (method url-fetch)
1641 (uri (string-append "https://mirror.bazel.build/github.com/NVlabs/"
1642 "cub/archive/" version ".zip"))
1643 (file-name (string-append "cub-" version ".zip"))
1644 (sha256
1645 (base32
1646 "1hsqikqridb90dkxkjr2918dcry6pfh46ccnwrzawl56aamhdykb")))))
1647 ("double-conversion-src"
1648 ,(let ((commit "5664746")
1649 (revision "1"))
1650 (origin
1651 (method git-fetch)
1652 (uri (git-reference
1653 (url "https://github.com/google/double-conversion.git")
1654 (commit commit)))
1655 (file-name
1656 (git-file-name "double-conversion"
1657 (string-append "0-" revision "."
1658 (string-take commit 7))))
1659 (sha256
1660 (base32
1661 "1h5lppqqxcvdg5jq42i5msgwx20ryij3apvmndflngrgdpc04gn1")))))
1662 ("farmhash-src"
1663 ,(let ((commit "816a4ae622e964763ca0862d9dbd19324a1eaf45"))
1664 (origin
1665 (method url-fetch)
1666 (uri (string-append
1667 "https://mirror.bazel.build/github.com/google/farmhash/archive/"
1668 commit ".tar.gz"))
1669 (file-name (string-append "farmhash-0-" (string-take commit 7)
1670 ".tar.gz"))
1671 (sha256
1672 (base32
1673 "185b2xdxl4d4cnsnv6abg8s22gxvx8673jq2yaq85bz4cdy58q35")))))
1674 ;; The license notice on the home page at
1675 ;; http://www.kurims.kyoto-u.ac.jp/~ooura/fft.html says:
1676 ;; Copyright Takuya OOURA, 1996-2001
1677 ;;
1678 ;; You may use, copy, modify and distribute this code for any purpose
1679 ;; (include commercial use) and without fee. Please refer to this
1680 ;; package when you modify this code.
1681 ;;
1682 ;; We take the identical tarball from the Bazel mirror, because the URL
1683 ;; at the home page is not versioned and might change.
1684 ("fft2d-src"
1685 ,(origin
1686 (method url-fetch)
1687 (uri "https://mirror.bazel.build/www.kurims.kyoto-u.ac.jp/~ooura/fft.tgz")
1688 (file-name "fft2d.tar.gz")
1689 (sha256
1690 (base32
1691 "15jjkfvhqvl2c0753d2di8hz0pyzn598g74wqy79awdrf1y67fsj"))))
1692 ("highwayhash-src"
1693 ,(let ((commit "be5edafc2e1a455768e260ccd68ae7317b6690ee")
1694 (revision "1"))
1695 (origin
1696 (method git-fetch)
1697 (uri (git-reference
1698 (url "https://github.com/google/highwayhash.git")
1699 (commit commit)))
1700 (file-name (string-append "highwayhash-0-" revision
1701 (string-take commit 7)
1702 "-checkout"))
1703 (sha256
1704 (base32
1705 "154jwf98cyy54hldr94pgjn85zynly3abpnc1avmb8a18lzwjyb6")))))
1706 ("nsync-src"
1707 ,(let ((version "0559ce013feac8db639ee1bf776aca0325d28777")
1708 (revision "1"))
1709 (origin
1710 (method url-fetch)
1711 (uri (string-append "https://mirror.bazel.build/"
1712 "github.com/google/nsync/archive/"
1713 version ".tar.gz"))
1714 (file-name (string-append "nsync-0." revision
1715 "-" (string-take version 7)
1716 ".tar.gz"))
1717 (sha256
1718 (base32
1719 "0qdkyqym34x739mmzv97ah5r7ph462v5xkxqxvidmcfqbi64b132")))))
1720 ("re2-src"
1721 ,(let ((commit "e7efc48")
1722 (revision "1"))
1723 (origin
1724 (method git-fetch)
1725 (uri (git-reference
1726 (url "https://github.com/google/re2")
1727 (commit commit)))
1728 (file-name (string-append "re2-0-" revision
1729 (string-take commit 7)
1730 "-checkout"))
1731 (sha256
1732 (base32
1733 "161g9841rjfsy5pn52fcis0s9hdr7rxvb06pad38j5rppfihvign")))))
1734 ("googletest" ,googletest)
1735 ("swig" ,swig)
1736 ("unzip" ,unzip)))
1737 (propagated-inputs
1738 `(("python-absl-py" ,python-absl-py)
1739 ("python-astor" ,python-astor)
1740 ("python-gast" ,python-gast)
1741 ("python-grpcio" ,python-grpcio)
1742 ("python-numpy" ,python-numpy)
1743 ("python-protobuf" ,python-protobuf-3.6)
1744 ("python-six" ,python-six)
1745 ("python-termcolo" ,python-termcolor)
1746 ("python-wheel" ,python-wheel)))
1747 (inputs
1748 `(("c-ares" ,c-ares)
1749 ("eigen" ,eigen-for-tensorflow)
1750 ("gemmlowp" ,gemmlowp-for-tensorflow)
1751 ("lmdb" ,lmdb)
1752 ("libjpeg" ,libjpeg)
1753 ("libpng" ,libpng)
1754 ("giflib" ,giflib)
1755 ("grpc" ,grpc)
1756 ("jsoncpp" ,jsoncpp-for-tensorflow)
1757 ("snappy" ,snappy)
1758 ("sqlite" ,sqlite)
1759 ("protobuf" ,protobuf-3.6)
1760 ("python" ,python-wrapper)
1761 ("zlib" ,zlib)))
1762 (home-page "https://tensorflow.org")
1763 (synopsis "Machine learning framework")
1764 (description
1765 "TensorFlow is a flexible platform for building and training machine
1766 learning models. It provides a library for high performance numerical
1767 computation and includes high level Python APIs, including both a sequential
1768 API for beginners that allows users to build models quickly by plugging
1769 together building blocks and a subclassing API with an imperative style for
1770 advanced research.")
1771 (license license:asl2.0)))
1772
1773 (define-public python-iml
1774 (package
1775 (name "python-iml")
1776 (version "0.6.2")
1777 (source
1778 (origin
1779 (method url-fetch)
1780 (uri (pypi-uri "iml" version))
1781 (sha256
1782 (base32
1783 "1k8szlpm19rcwcxdny9qdm3gmaqq8akb4xlvrzyz8c2d679aak6l"))))
1784 (build-system python-build-system)
1785 (propagated-inputs
1786 `(("ipython" ,python-ipython)
1787 ("nose" ,python-nose)
1788 ("numpy" ,python-numpy)
1789 ("pandas" ,python-pandas)
1790 ("scipy" ,python-scipy)))
1791 (home-page "http://github.com/interpretable-ml/iml")
1792 (synopsis "Interpretable Machine Learning (iML) package")
1793 (description "Interpretable ML (iML) is a set of data type objects,
1794 visualizations, and interfaces that can be used by any method designed to
1795 explain the predictions of machine learning models (or really the output of
1796 any function). It currently contains the interface and IO code from the Shap
1797 project, and it will potentially also do the same for the Lime project.")
1798 (license license:expat)))
1799
1800 (define-public python-keras-applications
1801 (package
1802 (name "python-keras-applications")
1803 (version "1.0.8")
1804 (source
1805 (origin
1806 (method url-fetch)
1807 (uri (pypi-uri "Keras_Applications" version))
1808 (sha256
1809 (base32
1810 "1rcz31ca4axa6kzhjx4lwqxbg4wvlljkj8qj9a7p9sfd5fhzjyam"))))
1811 (build-system python-build-system)
1812 ;; The tests require Keras, but this package is needed to build Keras.
1813 (arguments '(#:tests? #f))
1814 (propagated-inputs
1815 `(("python-h5py" ,python-h5py)
1816 ("python-numpy" ,python-numpy)))
1817 (native-inputs
1818 `(("python-pytest" ,python-pytest)
1819 ("python-pytest-cov" ,python-pytest-cov)
1820 ("python-pytest-pep8" ,python-pytest-pep8)
1821 ("python-pytest-xdist" ,python-pytest-xdist)))
1822 (home-page "https://github.com/keras-team/keras-applications")
1823 (synopsis "Reference implementations of popular deep learning models")
1824 (description
1825 "This package provides reference implementations of popular deep learning
1826 models for use with the Keras deep learning framework.")
1827 (license license:expat)))
1828
1829 (define-public python-keras-preprocessing
1830 (package
1831 (name "python-keras-preprocessing")
1832 (version "1.1.0")
1833 (source
1834 (origin
1835 (method url-fetch)
1836 (uri (pypi-uri "Keras_Preprocessing" version))
1837 (sha256
1838 (base32
1839 "1r98nm4k1svsqjyaqkfk23i31bl1kcfcyp7094yyj3c43phfp3as"))))
1840 (build-system python-build-system)
1841 (propagated-inputs
1842 `(("python-numpy" ,python-numpy)
1843 ("python-six" ,python-six)))
1844 (native-inputs
1845 `(("python-pandas" ,python-pandas)
1846 ("python-pillow" ,python-pillow)
1847 ("python-pytest" ,python-pytest)
1848 ("python-pytest-cov" ,python-pytest-cov)
1849 ("python-pytest-xdist" ,python-pytest-xdist)
1850 ("tensorflow" ,tensorflow)))
1851 (home-page "https://github.com/keras-team/keras-preprocessing/")
1852 (synopsis "Data preprocessing and augmentation for deep learning models")
1853 (description
1854 "Keras Preprocessing is the data preprocessing and data augmentation
1855 module of the Keras deep learning library. It provides utilities for working
1856 with image data, text data, and sequence data.")
1857 (license license:expat)))
1858
1859 (define-public python-keras
1860 (package
1861 (name "python-keras")
1862 (version "2.2.4")
1863 (source
1864 (origin
1865 (method url-fetch)
1866 (uri (pypi-uri "Keras" version))
1867 (patches (search-patches "python-keras-integration-test.patch"))
1868 (sha256
1869 (base32
1870 "1j8bsqzh49vjdxy6l1k4iwax5vpjzniynyd041xjavdzvfii1dlh"))))
1871 (build-system python-build-system)
1872 (arguments
1873 `(#:phases
1874 (modify-phases %standard-phases
1875 (add-after 'unpack 'remove-tests-for-unavailable-features
1876 (lambda _
1877 (delete-file "keras/backend/theano_backend.py")
1878 (delete-file "keras/backend/cntk_backend.py")
1879 (delete-file "tests/keras/backend/backend_test.py")
1880
1881 ;; FIXME: This doesn't work because Tensorflow is missing the
1882 ;; coder ops library.
1883 (delete-file "tests/keras/test_callbacks.py")
1884 #t))
1885 (replace 'check
1886 (lambda _
1887 ;; These tests attempt to download data files from the internet.
1888 (delete-file "tests/integration_tests/test_datasets.py")
1889 (delete-file "tests/integration_tests/imagenet_utils_test.py")
1890
1891 (setenv "PYTHONPATH"
1892 (string-append (getcwd) "/build/lib:"
1893 (getenv "PYTHONPATH")))
1894 (invoke "py.test" "-v"
1895 "-p" "no:cacheprovider"
1896 "--ignore" "keras/utils"))))))
1897 (propagated-inputs
1898 `(("python-h5py" ,python-h5py)
1899 ("python-keras-applications" ,python-keras-applications)
1900 ("python-keras-preprocessing" ,python-keras-preprocessing)
1901 ("python-numpy" ,python-numpy)
1902 ("python-pydot" ,python-pydot)
1903 ("python-pyyaml" ,python-pyyaml)
1904 ("python-scipy" ,python-scipy)
1905 ("python-six" ,python-six)
1906 ("tensorflow" ,tensorflow)
1907 ("graphviz" ,graphviz)))
1908 (native-inputs
1909 `(("python-pandas" ,python-pandas)
1910 ("python-pytest" ,python-pytest)
1911 ("python-pytest-cov" ,python-pytest-cov)
1912 ("python-pytest-pep8" ,python-pytest-pep8)
1913 ("python-pytest-timeout" ,python-pytest-timeout)
1914 ("python-pytest-xdist" ,python-pytest-xdist)
1915 ("python-sphinx" ,python-sphinx)
1916 ("python-requests" ,python-requests)))
1917 (home-page "https://github.com/keras-team/keras")
1918 (synopsis "High-level deep learning framework")
1919 (description "Keras is a high-level neural networks API, written in Python
1920 and capable of running on top of TensorFlow. It was developed with a focus on
1921 enabling fast experimentation. Use Keras if you need a deep learning library
1922 that:
1923
1924 @itemize
1925 @item Allows for easy and fast prototyping (through user friendliness,
1926 modularity, and extensibility).
1927 @item Supports both convolutional networks and recurrent networks, as well as
1928 combinations of the two.
1929 @item Runs seamlessly on CPU and GPU.
1930 @end itemize\n")
1931 (license license:expat)))
1932
1933 (define-public sbcl-cl-libsvm-format
1934 (let ((commit "3300f84fd8d9f5beafc114f543f9d83417c742fb")
1935 (revision "0"))
1936 (package
1937 (name "sbcl-cl-libsvm-format")
1938 (version (git-version "0.1.0" revision commit))
1939 (source
1940 (origin
1941 (method git-fetch)
1942 (uri (git-reference
1943 (url "https://github.com/masatoi/cl-libsvm-format.git")
1944 (commit commit)))
1945 (file-name (git-file-name name version))
1946 (sha256
1947 (base32
1948 "0284aj84xszhkhlivaigf9qj855fxad3mzmv3zfr0qzb5k0nzwrg"))))
1949 (build-system asdf-build-system/sbcl)
1950 (native-inputs
1951 `(("prove" ,sbcl-prove)
1952 ("prove-asdf" ,sbcl-prove-asdf)))
1953 (inputs
1954 `(("alexandria" ,sbcl-alexandria)))
1955 (synopsis "LibSVM data format reader for Common Lisp")
1956 (description
1957 "This Common Lisp library provides a fast reader for data in LibSVM
1958 format.")
1959 (home-page "https://github.com/masatoi/cl-libsvm-format")
1960 (license license:expat))))
1961
1962 (define-public cl-libsvm-format
1963 (sbcl-package->cl-source-package sbcl-cl-libsvm-format))
1964
1965 (define-public ecl-cl-libsvm-format
1966 (sbcl-package->ecl-package sbcl-cl-libsvm-format))
1967
1968 (define-public sbcl-cl-online-learning
1969 (let ((commit "fc7a34f4f161cd1c7dd747d2ed8f698947781423")
1970 (revision "0"))
1971 (package
1972 (name "sbcl-cl-online-learning")
1973 (version (git-version "0.5" revision commit))
1974 (source
1975 (origin
1976 (method git-fetch)
1977 (uri (git-reference
1978 (url "https://github.com/masatoi/cl-online-learning.git")
1979 (commit commit)))
1980 (file-name (git-file-name name version))
1981 (sha256
1982 (base32
1983 "14x95rlg80ay5hv645ki57pqvy12v28hz4k1w0f6bsfi2rmpxchq"))))
1984 (build-system asdf-build-system/sbcl)
1985 (native-inputs
1986 `(("prove" ,sbcl-prove)
1987 ("prove-asdf" ,sbcl-prove-asdf)))
1988 (inputs
1989 `(("cl-libsvm-format" ,sbcl-cl-libsvm-format)
1990 ("cl-store" ,sbcl-cl-store)))
1991 (arguments
1992 `(;; FIXME: Tests pass but then the check phase crashes
1993 #:tests? #f))
1994 (synopsis "Online Machine Learning for Common Lisp")
1995 (description
1996 "This library contains a collection of machine learning algorithms for
1997 online linear classification written in Common Lisp.")
1998 (home-page "https://github.com/masatoi/cl-online-learning")
1999 (license license:expat))))
2000
2001 (define-public cl-online-learning
2002 (sbcl-package->cl-source-package sbcl-cl-online-learning))
2003
2004 (define-public ecl-cl-online-learning
2005 (sbcl-package->ecl-package sbcl-cl-online-learning))
2006
2007 (define-public sbcl-cl-random-forest
2008 (let ((commit "85fbdd4596d40e824f70f1b7cf239cf544e49d51")
2009 (revision "0"))
2010 (package
2011 (name "sbcl-cl-random-forest")
2012 (version (git-version "0.1" revision commit))
2013 (source
2014 (origin
2015 (method git-fetch)
2016 (uri (git-reference
2017 (url "https://github.com/masatoi/cl-random-forest.git")
2018 (commit commit)))
2019 (file-name (git-file-name name version))
2020 (sha256
2021 (base32
2022 "097xv60i1ndz68sg9p4pc7c5gvyp9i1xgw966b4wwfq3x6hbz421"))))
2023 (build-system asdf-build-system/sbcl)
2024 (native-inputs
2025 `(("prove" ,sbcl-prove)
2026 ("prove-asdf" ,sbcl-prove-asdf)
2027 ("trivial-garbage" ,sbcl-trivial-garbage)))
2028 (inputs
2029 `(("alexandria" ,sbcl-alexandria)
2030 ("cl-libsvm-format" ,sbcl-cl-libsvm-format)
2031 ("cl-online-learning" ,sbcl-cl-online-learning)
2032 ("lparallel" ,sbcl-lparallel)))
2033 (arguments
2034 `(;; The tests download data from the Internet
2035 #:tests? #f
2036 #:phases
2037 (modify-phases %standard-phases
2038 (add-after 'unpack 'add-sb-cltl2-dependency
2039 (lambda _
2040 ;; sb-cltl2 is required by lparallel when using sbcl, but it is
2041 ;; not loaded automatically.
2042 (substitute* "cl-random-forest.asd"
2043 (("\\(in-package :cl-user\\)")
2044 "(in-package :cl-user) #+sbcl (require :sb-cltl2)"))
2045 #t)))))
2046 (synopsis "Random Forest and Global Refinement for Common Lisp")
2047 (description
2048 "CL-random-forest is an implementation of Random Forest for multiclass
2049 classification and univariate regression written in Common Lisp. It also
2050 includes an implementation of Global Refinement of Random Forest.")
2051 (home-page "https://github.com/masatoi/cl-random-forest")
2052 (license license:expat))))
2053
2054 (define-public cl-random-forest
2055 (sbcl-package->cl-source-package sbcl-cl-random-forest))
2056
2057 (define-public ecl-cl-random-forest
2058 (sbcl-package->ecl-package sbcl-cl-random-forest))
2059
2060 (define-public gloo
2061 (let ((version "0.0.0") ; no proper version tag
2062 (commit "ca528e32fea9ca8f2b16053cff17160290fc84ce")
2063 (revision "0"))
2064 (package
2065 (name "gloo")
2066 (version (git-version version revision commit))
2067 (source
2068 (origin
2069 (method git-fetch)
2070 (uri (git-reference
2071 (url "https://github.com/facebookincubator/gloo.git")
2072 (commit commit)))
2073 (file-name (git-file-name name version))
2074 (sha256
2075 (base32
2076 "1q9f80zy75f6njrzrqkmhc0g3qxs4gskr7ns2jdqanxa2ww7a99w"))))
2077 (build-system cmake-build-system)
2078 (native-inputs
2079 `(("googletest" ,googletest)))
2080 (arguments
2081 `(#:configure-flags '("-DBUILD_TEST=1")
2082 #:phases
2083 (modify-phases %standard-phases
2084 (replace 'check
2085 (lambda _
2086 (invoke "make" "gloo_test")
2087 #t)))))
2088 (synopsis "Collective communications library")
2089 (description
2090 "Gloo is a collective communications library. It comes with a
2091 number of collective algorithms useful for machine learning applications.
2092 These include a barrier, broadcast, and allreduce.")
2093 (home-page "https://github.com/facebookincubator/gloo")
2094 (license license:bsd-3))))
2095
2096 (define-public python-umap-learn
2097 (package
2098 (name "python-umap-learn")
2099 (version "0.3.10")
2100 (source
2101 (origin
2102 (method url-fetch)
2103 (uri (pypi-uri "umap-learn" version))
2104 (sha256
2105 (base32
2106 "02ada2yy6km6zgk2836kg1c97yrcpalvan34p8c57446finnpki1"))))
2107 (build-system python-build-system)
2108 (native-inputs
2109 `(("python-joblib" ,python-joblib)
2110 ("python-nose" ,python-nose)))
2111 (propagated-inputs
2112 `(("python-numba" ,python-numba)
2113 ("python-numpy" ,python-numpy)
2114 ("python-scikit-learn" ,python-scikit-learn)
2115 ("python-scipy" ,python-scipy)))
2116 (home-page "https://github.com/lmcinnes/umap")
2117 (synopsis
2118 "Uniform Manifold Approximation and Projection")
2119 (description
2120 "Uniform Manifold Approximation and Projection is a dimension reduction
2121 technique that can be used for visualisation similarly to t-SNE, but also for
2122 general non-linear dimension reduction.")
2123 (license license:bsd-3)))