mirror of
https://github.com/saymrwulf/onnxruntime.git
synced 2026-05-16 21:00:14 +00:00
72186bbb71
### Description * Build cuda nhwc ops by default. * Deprecate `--enable_cuda_nhwc_ops` in build.py and add `--disable_cuda_nhwc_ops` option Note that it requires cuDNN 9.x. If you build with cuDNN 8, NHWC ops will be disabled automatically. ### Motivation and Context In general, NHWC is faster than NCHW for convolution in Nvidia GPUs with Tensor Cores, and this could improve performance for vision models. This is the first step to prefer NHWC for CUDA in 1.21 release. Next step is to do some tests on popular vision models. If it help in most models and devices, set `prefer_nhwc=1` as default cuda provider option.
49 lines
1.5 KiB
Bash
Executable file
49 lines
1.5 KiB
Bash
Executable file
#!/bin/bash
|
|
set -ex
|
|
#Every cuda container has this $CUDA_VERSION env var set.
|
|
SHORT_CUDA_VERSION=$(echo $CUDA_VERSION | sed 's/\([[:digit:]]\+\.[[:digit:]]\+\)\.[[:digit:]]\+/\1/')
|
|
|
|
BUILD_ARGS=('--config'
|
|
'Release'
|
|
'--update'
|
|
'--build'
|
|
'--skip_submodule_sync'
|
|
'--build_shared_lib'
|
|
'--parallel'
|
|
'--use_binskim_compliant_compile_flags'
|
|
'--build_wheel'
|
|
'--enable_onnx_tests'
|
|
'--use_cuda'
|
|
"--cuda_version=$SHORT_CUDA_VERSION"
|
|
"--cuda_home=/usr/local/cuda-$SHORT_CUDA_VERSION"
|
|
"--cudnn_home=/usr/local/cuda-$SHORT_CUDA_VERSION"
|
|
"--enable_cuda_profiling"
|
|
"--enable_pybind"
|
|
"--build_java"
|
|
"--cmake_extra_defines"
|
|
"CMAKE_CUDA_ARCHITECTURES=75"
|
|
"onnxruntime_BUILD_UNIT_TESTS=ON"
|
|
"onnxruntime_ENABLE_CUDA_EP_INTERNAL_TESTS=ON")
|
|
if [ -x "$(command -v ninja)" ]; then
|
|
BUILD_ARGS+=('--cmake_generator' 'Ninja')
|
|
fi
|
|
|
|
if [ -d /build ]; then
|
|
BUILD_ARGS+=('--build_dir' '/build')
|
|
else
|
|
BUILD_ARGS+=('--build_dir' 'build')
|
|
fi
|
|
|
|
if [ -x "$(command -v ccache)" ]; then
|
|
ccache -s;
|
|
#BUILD_ARGS+=("--use_cache")
|
|
fi
|
|
if [ -f /opt/python/cp312-cp312/bin/python3 ]; then
|
|
/opt/python/cp312-cp312/bin/python3 tools/ci_build/build.py "${BUILD_ARGS[@]}"
|
|
else
|
|
python3 tools/ci_build/build.py "${BUILD_ARGS[@]}"
|
|
fi
|
|
if [ -x "$(command -v ccache)" ]; then
|
|
ccache -sv
|
|
ccache -z
|
|
fi
|