1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
|
{ config
, stdenv
, lib
, fetchFromGitHub
, fetchpatch
, Foundation
, abseil-cpp
, cmake
, cpuinfo
, eigen
, flatbuffers
, gbenchmark
, glibcLocales
, gtest
, libpng
, nlohmann_json
, nsync
, pkg-config
, python3Packages
, re2
, zlib
, microsoft-gsl
, libiconv
, protobuf_21
, pythonSupport ? true
, cudaSupport ? config.cudaSupport
, cudaPackages ? {}
}@inputs:
let
version = "1.18.1";
stdenv = throw "Use effectiveStdenv instead";
effectiveStdenv = if cudaSupport then cudaPackages.backendStdenv else inputs.stdenv;
cudaArchitecturesString = cudaPackages.flags.cmakeCudaArchitecturesString;
howard-hinnant-date = fetchFromGitHub {
owner = "HowardHinnant";
repo = "date";
rev = "v3.0.1";
sha256 = "sha256-ZSjeJKAcT7mPym/4ViDvIR9nFMQEBCSUtPEuMO27Z+I=";
};
mp11 = fetchFromGitHub {
owner = "boostorg";
repo = "mp11";
rev = "boost-1.82.0";
hash = "sha256-cLPvjkf2Au+B19PJNrUkTW/VPxybi1MpPxnIl4oo4/o=";
};
safeint = fetchFromGitHub {
owner = "dcleblanc";
repo = "safeint";
rev = "ff15c6ada150a5018c5ef2172401cb4529eac9c0";
hash = "sha256-PK1ce4C0uCR4TzLFg+elZdSk5DdPCRhhwT3LvEwWnPU=";
};
pytorch_clog = effectiveStdenv.mkDerivation {
pname = "clog";
version = "3c8b153";
src = "${cpuinfo.src}/deps/clog";
nativeBuildInputs = [ cmake gbenchmark gtest ];
cmakeFlags = [
"-DUSE_SYSTEM_GOOGLEBENCHMARK=ON"
"-DUSE_SYSTEM_GOOGLETEST=ON"
"-DUSE_SYSTEM_LIBS=ON"
# 'clog' tests set 'CXX_STANDARD 11'; this conflicts with our 'gtest'.
"-DCLOG_BUILD_TESTS=OFF"
];
};
onnx = fetchFromGitHub {
owner = "onnx";
repo = "onnx";
rev = "refs/tags/v1.16.1";
hash = "sha256-I1wwfn91hdH3jORIKny0Xc73qW2P04MjkVCgcaNnQUE=";
};
cutlass = fetchFromGitHub {
owner = "NVIDIA";
repo = "cutlass";
rev = "v3.1.0";
hash = "sha256-mpaiCxiYR1WaSSkcEPTzvcREenJWklD+HRdTT5/pD54=";
};
in
effectiveStdenv.mkDerivation rec {
pname = "onnxruntime";
inherit version;
src = fetchFromGitHub {
owner = "microsoft";
repo = "onnxruntime";
rev = "refs/tags/v${version}";
hash = "sha256-+zWtbLKekGhwdBU3bm1u2F7rYejQ62epE+HcHj05/8A=";
fetchSubmodules = true;
};
patches = [
# If you stumble on these patches trying to update onnxruntime, check
# `git blame` and ping the introducers.
# Context: we want the upstream to
# - always try find_package first (FIND_PACKAGE_ARGS),
# - use MakeAvailable instead of the low-level Populate,
# - use Eigen3::Eigen as the target name (as declared by libeigen/eigen).
./0001-eigen-allow-dependency-injection.patch
# Incorporate a patch that has landed upstream which exposes new
# 'abseil-cpp' libraries & modifies the 're2' CMakeLists to fix a
# configuration error that around missing 'gmock' exports.
#
# TODO: Check if it can be dropped after 1.19.0
# https://github.com/microsoft/onnxruntime/commit/b522df0ae477e59f60acbe6c92c8a64eda96cace
./update-re2.patch
] ++ lib.optionals cudaSupport [
# We apply the referenced 1064.patch ourselves to our nix dependency.
# FIND_PACKAGE_ARGS for CUDA was added in https://github.com/microsoft/onnxruntime/commit/87744e5 so it might be possible to delete this patch after upgrading to 1.17.0
./nvcc-gsl.patch
];
nativeBuildInputs = [
cmake
pkg-config
python3Packages.python
protobuf_21
] ++ lib.optionals pythonSupport (with python3Packages; [
pip
python
pythonOutputDistHook
setuptools
wheel
]) ++ lib.optionals cudaSupport [
cudaPackages.cuda_nvcc
];
buildInputs = [
cpuinfo
eigen
glibcLocales
libpng
nlohmann_json
microsoft-gsl
pytorch_clog
zlib
] ++ lib.optionals pythonSupport (with python3Packages; [
numpy
pybind11
packaging
]) ++ lib.optionals effectiveStdenv.isDarwin [
Foundation
libiconv
] ++ lib.optionals cudaSupport (with cudaPackages; [
cuda_cccl # cub/cub.cuh
libcublas # cublas_v2.h
libcurand # curand.h
libcusparse # cusparse.h
libcufft # cufft.h
cudnn # cudnn.h
cuda_cudart
]);
nativeCheckInputs = [
gtest
] ++ lib.optionals pythonSupport (with python3Packages; [
pytest
sympy
onnx
]);
# TODO: build server, and move .so's to lib output
# Python's wheel is stored in a separate dist output
outputs = [ "out" "dev" ] ++ lib.optionals pythonSupport [ "dist" ];
enableParallelBuilding = true;
cmakeDir = "../cmake";
cmakeFlags = [
"-DABSL_ENABLE_INSTALL=ON"
"-DFETCHCONTENT_FULLY_DISCONNECTED=ON"
"-DFETCHCONTENT_QUIET=OFF"
"-DFETCHCONTENT_SOURCE_DIR_ABSEIL_CPP=${abseil-cpp.src}"
"-DFETCHCONTENT_SOURCE_DIR_DATE=${howard-hinnant-date}"
"-DFETCHCONTENT_SOURCE_DIR_FLATBUFFERS=${flatbuffers.src}"
"-DFETCHCONTENT_SOURCE_DIR_GOOGLETEST=${gtest.src}"
"-DFETCHCONTENT_SOURCE_DIR_GOOGLE_NSYNC=${nsync.src}"
"-DFETCHCONTENT_SOURCE_DIR_MP11=${mp11}"
"-DFETCHCONTENT_SOURCE_DIR_ONNX=${onnx}"
"-DFETCHCONTENT_SOURCE_DIR_RE2=${re2.src}"
"-DFETCHCONTENT_SOURCE_DIR_SAFEINT=${safeint}"
"-DFETCHCONTENT_TRY_FIND_PACKAGE_MODE=ALWAYS"
"-Donnxruntime_BUILD_SHARED_LIB=ON"
(lib.cmakeBool "onnxruntime_BUILD_UNIT_TESTS" doCheck)
"-Donnxruntime_ENABLE_LTO=ON"
"-Donnxruntime_USE_FULL_PROTOBUF=OFF"
(lib.cmakeBool "onnxruntime_USE_CUDA" cudaSupport)
(lib.cmakeBool "onnxruntime_USE_NCCL" cudaSupport)
] ++ lib.optionals pythonSupport [
"-Donnxruntime_ENABLE_PYTHON=ON"
] ++ lib.optionals cudaSupport [
(lib.cmakeFeature "FETCHCONTENT_SOURCE_DIR_CUTLASS" "${cutlass}")
(lib.cmakeFeature "onnxruntime_CUDNN_HOME" "${cudaPackages.cudnn}")
(lib.cmakeFeature "CMAKE_CUDA_ARCHITECTURES" cudaArchitecturesString)
(lib.cmakeFeature "onnxruntime_NVCC_THREADS" "1")
];
env = lib.optionalAttrs effectiveStdenv.cc.isClang {
NIX_CFLAGS_COMPILE = toString [
"-Wno-error=deprecated-declarations"
"-Wno-error=deprecated-pragma"
"-Wno-error=unused-but-set-variable"
];
};
# aarch64-linux fails cpuinfo test, because /sys/devices/system/cpu/ does not exist in the sandbox
doCheck = !(cudaSupport || effectiveStdenv.buildPlatform.system == "aarch64-linux");
requiredSystemFeatures = lib.optionals cudaSupport [ "big-parallel" ];
postPatch = ''
substituteInPlace cmake/libonnxruntime.pc.cmake.in \
--replace-fail '$'{prefix}/@CMAKE_INSTALL_ @CMAKE_INSTALL_
'' + lib.optionalString (effectiveStdenv.hostPlatform.system == "aarch64-linux") ''
# https://github.com/NixOS/nixpkgs/pull/226734#issuecomment-1663028691
rm -v onnxruntime/test/optimizer/nhwc_transformer_test.cc
'';
postBuild = lib.optionalString pythonSupport ''
${python3Packages.python.interpreter} ../setup.py bdist_wheel
'';
postInstall = ''
# perform parts of `tools/ci_build/github/linux/copy_strip_binary.sh`
install -m644 -Dt $out/include \
../include/onnxruntime/core/framework/provider_options.h \
../include/onnxruntime/core/providers/cpu/cpu_provider_factory.h \
../include/onnxruntime/core/session/onnxruntime_*.h
'';
passthru = {
inherit cudaSupport cudaPackages; # for the python module
protobuf = protobuf_21;
tests = lib.optionalAttrs pythonSupport {
python = python3Packages.onnxruntime;
};
};
meta = with lib; {
description = "Cross-platform, high performance scoring engine for ML models";
longDescription = ''
ONNX Runtime is a performance-focused complete scoring engine
for Open Neural Network Exchange (ONNX) models, with an open
extensible architecture to continually address the latest developments
in AI and Deep Learning. ONNX Runtime stays up to date with the ONNX
standard with complete implementation of all ONNX operators, and
supports all ONNX releases (1.2+) with both future and backwards
compatibility.
'';
homepage = "https://github.com/microsoft/onnxruntime";
changelog = "https://github.com/microsoft/onnxruntime/releases/tag/v${version}";
# https://github.com/microsoft/onnxruntime/blob/master/BUILD.md#architectures
platforms = platforms.unix;
license = licenses.mit;
maintainers = with maintainers; [ puffnfresh ck3d cbourjau ];
};
}
|