Skip to content

Commit f412f85

Browse files
BoyuanFengBoyuanFeng
committed
Merge branch 'main' into bf/NopKernelSchedulerNode-device
2 parents f94f3e4 + 51b7528 commit f412f85

File tree

624 files changed

+21462
-12289
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

624 files changed

+21462
-12289
lines changed

‎.ci/docker/ci_commit_pins/triton-xpu.txt‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1 +1 @@
1-
91b14bf5593cf58a8541f3e6b9125600a867d4ef
1+
e98b6fcb8df5b44eb0d0addb6767c573d37ba024

‎.ci/docker/common/install_conda.sh‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -65,7 +65,7 @@ if [ -n "$ANACONDA_PYTHON_VERSION" ]; then
6565

6666
# Install PyTorch conda deps, as per https://github.com/pytorch/pytorch README
6767
if [[ $(uname -m) == "aarch64" ]]; then
68-
conda_install "openblas==0.3.25=*openmp*"
68+
conda_install "openblas==0.3.28=*openmp*"
6969
else
7070
conda_install "mkl=2021.4.0 mkl-include=2021.4.0"
7171
fi

‎.ci/docker/common/install_cuda.sh‎

Lines changed: 13 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -38,6 +38,17 @@ function install_cusparselt_062 {
3838
rm -rf tmp_cusparselt
3939
}
4040

41+
function install_cusparselt_063 {
42+
# cuSparseLt license: https://docs.nvidia.com/cuda/cusparselt/license.html
43+
mkdir tmp_cusparselt && pushd tmp_cusparselt
44+
wget -q https://developer.download.nvidia.com/compute/cusparselt/redist/libcusparse_lt/linux-x86_64/libcusparse_lt-linux-x86_64-0.6.3.2-archive.tar.xz
45+
tar xf libcusparse_lt-linux-x86_64-0.6.3.2-archive.tar.xz
46+
cp -a libcusparse_lt-linux-x86_64-0.6.3.2-archive/include/* /usr/local/cuda/include/
47+
cp -a libcusparse_lt-linux-x86_64-0.6.3.2-archive/lib/* /usr/local/cuda/lib64/
48+
popd
49+
rm -rf tmp_cusparselt
50+
}
51+
4152
function install_118 {
4253
CUDNN_VERSION=9.1.0.70
4354
echo "Installing CUDA 11.8 and cuDNN ${CUDNN_VERSION} and NCCL ${NCCL_VERSION} and cuSparseLt-0.4.0"
@@ -140,7 +151,7 @@ function install_124 {
140151
}
141152

142153
function install_126 {
143-
echo "Installing CUDA 12.6.3 and cuDNN ${CUDNN_VERSION} and NCCL ${NCCL_VERSION} and cuSparseLt-0.6.2"
154+
echo "Installing CUDA 12.6.3 and cuDNN ${CUDNN_VERSION} and NCCL ${NCCL_VERSION} and cuSparseLt-0.6.3"
144155
rm -rf /usr/local/cuda-12.6 /usr/local/cuda
145156
# install CUDA 12.6.3 in the same container
146157
wget -q https://developer.download.nvidia.com/compute/cuda/12.6.3/local_installers/cuda_12.6.3_560.35.05_linux.run
@@ -167,7 +178,7 @@ function install_126 {
167178
cd ..
168179
rm -rf nccl
169180

170-
install_cusparselt_062
181+
install_cusparselt_063
171182

172183
ldconfig
173184
}

‎.ci/docker/common/install_cuda_aarch64.sh‎

Lines changed: 18 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -17,6 +17,17 @@ function install_cusparselt_062 {
1717
rm -rf tmp_cusparselt
1818
}
1919

20+
function install_cusparselt_063 {
21+
# cuSparseLt license: https://docs.nvidia.com/cuda/cusparselt/license.html
22+
mkdir tmp_cusparselt && pushd tmp_cusparselt
23+
wget -q https://developer.download.nvidia.com/compute/cusparselt/redist/libcusparse_lt/linux-x86_64/libcusparse_lt-linux-x86_64-0.6.3.2-archive.tar.xz
24+
tar xf libcusparse_lt-linux-x86_64-0.6.3.2-archive.tar.xz
25+
cp -a libcusparse_lt-linux-x86_64-0.6.3.2-archive/include/* /usr/local/cuda/include/
26+
cp -a libcusparse_lt-linux-x86_64-0.6.3.2-archive/lib/* /usr/local/cuda/lib64/
27+
popd
28+
rm -rf tmp_cusparselt
29+
}
30+
2031
function install_124 {
2132
CUDNN_VERSION=9.1.0.70
2233
echo "Installing CUDA 12.4.1 and cuDNN ${CUDNN_VERSION} and NCCL ${NCCL_VERSION} and cuSparseLt-0.6.2"
@@ -83,13 +94,13 @@ function prune_124 {
8394
}
8495

8596
function install_126 {
86-
echo "Installing CUDA 12.6.2 and cuDNN ${CUDNN_VERSION} and NCCL ${NCCL_VERSION} and cuSparseLt-0.6.2"
97+
echo "Installing CUDA 12.6.3 and cuDNN ${CUDNN_VERSION} and NCCL ${NCCL_VERSION} and cuSparseLt-0.6.3"
8798
rm -rf /usr/local/cuda-12.6 /usr/local/cuda
88-
# install CUDA 12.6.2 in the same container
89-
wget -q https://developer.download.nvidia.com/compute/cuda/12.6.2/local_installers/cuda_12.6.2_560.35.03_linux_sbsa.run
90-
chmod +x cuda_12.6.2_560.35.03_linux_sbsa.run
91-
./cuda_12.6.2_560.35.03_linux_sbsa.run --toolkit --silent
92-
rm -f cuda_12.6.2_560.35.03_linux_sbsa.run
99+
# install CUDA 12.6.3 in the same container
100+
wget -q https://developer.download.nvidia.com/compute/cuda/12.6.3/local_installers/cuda_12.6.3_560.35.05_linux_sbsa.run
101+
chmod +x cuda_12.6.3_560.35.05_linux_sbsa.run
102+
./cuda_12.6.3_560.35.05_linux_sbsa.run --toolkit --silent
103+
rm -f cuda_12.6.3_560.35.05_linux_sbsa.run
93104
rm -f /usr/local/cuda && ln -s /usr/local/cuda-12.6 /usr/local/cuda
94105

95106
# cuDNN license: https://developer.nvidia.com/cudnn/license_agreement
@@ -110,7 +121,7 @@ function install_126 {
110121
cd ..
111122
rm -rf nccl
112123

113-
install_cusparselt_062
124+
install_cusparselt_063
114125

115126
ldconfig
116127
}

‎.ci/docker/common/install_openblas.sh‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -4,7 +4,7 @@
44
set -ex
55

66
cd /
7-
git clone https://github.com/OpenMathLib/OpenBLAS.git -b v0.3.25 --depth 1 --shallow-submodules
7+
git clone https://github.com/OpenMathLib/OpenBLAS.git -b v0.3.28 --depth 1 --shallow-submodules
88

99

1010
OPENBLAS_BUILD_FLAGS="

‎.ci/docker/manywheel/Dockerfile_2_28‎

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -145,6 +145,8 @@ FROM cpu_final as rocm_final
145145
ARG ROCM_VERSION=6.0
146146
ARG PYTORCH_ROCM_ARCH
147147
ENV PYTORCH_ROCM_ARCH ${PYTORCH_ROCM_ARCH}
148+
ARG DEVTOOLSET_VERSION=11
149+
ENV LDFLAGS="-Wl,-rpath=/opt/rh/gcc-toolset-${DEVTOOLSET_VERSION}/root/usr/lib64 -Wl,-rpath=/opt/rh/gcc-toolset-${DEVTOOLSET_VERSION}/root/usr/lib"
148150
# Somewhere in ROCm stack, we still use non-existing /opt/rocm/hip path,
149151
# below workaround helps avoid error
150152
ENV ROCM_PATH /opt/rocm

‎.ci/docker/manywheel/build.sh‎

Lines changed: 3 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -91,8 +91,10 @@ case ${GPU_ARCH_TYPE} in
9191
TARGET=rocm_final
9292
DOCKER_TAG=rocm${GPU_ARCH_VERSION}
9393
GPU_IMAGE=rocm/dev-centos-7:${GPU_ARCH_VERSION}-complete
94+
DEVTOOLSET_VERSION="9"
9495
if [ ${GPU_ARCH_TYPE} == "rocm-manylinux_2_28" ]; then
9596
MANY_LINUX_VERSION="2_28"
97+
DEVTOOLSET_VERSION="11"
9698
GPU_IMAGE=rocm/dev-almalinux-8:${GPU_ARCH_VERSION}-complete
9799
fi
98100
PYTORCH_ROCM_ARCH="gfx900;gfx906;gfx908;gfx90a;gfx942;gfx1030;gfx1100"
@@ -103,7 +105,7 @@ case ${GPU_ARCH_TYPE} in
103105
echo "ERROR: rocm regex failed"
104106
exit 1
105107
fi
106-
DOCKER_GPU_BUILD_ARG="--build-arg ROCM_VERSION=${GPU_ARCH_VERSION} --build-arg PYTORCH_ROCM_ARCH=${PYTORCH_ROCM_ARCH} --build-arg DEVTOOLSET_VERSION=9"
108+
DOCKER_GPU_BUILD_ARG="--build-arg ROCM_VERSION=${GPU_ARCH_VERSION} --build-arg PYTORCH_ROCM_ARCH=${PYTORCH_ROCM_ARCH} --build-arg DEVTOOLSET_VERSION=${DEVTOOLSET_VERSION}"
107109
;;
108110
xpu)
109111
TARGET=xpu_final

‎.ci/docker/requirements-ci.txt‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -90,7 +90,7 @@ librosa>=0.6.2 ; python_version < "3.11"
9090
#Pinned versions:
9191
#test that import:
9292

93-
mypy==1.11.2
93+
mypy==1.13.0
9494
# Pin MyPy version because new errors are likely to appear with each release
9595
#Description: linter
9696
#Pinned versions: 1.10.0

‎.ci/manywheel/build.sh‎

Lines changed: 4 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -15,9 +15,12 @@ case "${GPU_ARCH_TYPE:-BLANK}" in
1515
rocm)
1616
bash "${SCRIPTPATH}/build_rocm.sh"
1717
;;
18-
cpu | cpu-cxx11-abi | cpu-s390x | xpu)
18+
cpu | cpu-cxx11-abi | cpu-s390x)
1919
bash "${SCRIPTPATH}/build_cpu.sh"
2020
;;
21+
xpu)
22+
bash "${SCRIPTPATH}/build_xpu.sh"
23+
;;
2124
*)
2225
echo "Un-recognized GPU_ARCH_TYPE '${GPU_ARCH_TYPE}', exiting..."
2326
exit 1

‎.ci/manywheel/build_cpu.sh‎

Lines changed: 4 additions & 33 deletions
Original file line numberDiff line numberDiff line change
@@ -2,8 +2,6 @@
22

33
set -ex
44

5-
GPU_ARCH_TYPE=${GPU_ARCH_TYPE:-cpu}
6-
75
export TH_BINARY_BUILD=1
86
export USE_CUDA=0
97

@@ -17,22 +15,13 @@ if [[ -z "$EXTRA_CAFFE2_CMAKE_FLAGS" ]]; then
1715
EXTRA_CAFFE2_CMAKE_FLAGS=()
1816
fi
1917

20-
DIR_SUFFIX=cpu
21-
if [[ "$GPU_ARCH_TYPE" == "xpu" ]]; then
22-
DIR_SUFFIX=xpu
23-
# Refer https://www.intel.com/content/www/us/en/developer/articles/tool/pytorch-prerequisites-for-intel-gpus.html
24-
source /opt/intel/oneapi/compiler/latest/env/vars.sh
25-
source /opt/intel/oneapi/pti/latest/env/vars.sh
26-
export USE_STATIC_MKL=1
27-
fi
28-
29-
WHEELHOUSE_DIR="wheelhouse$DIR_SUFFIX"
30-
LIBTORCH_HOUSE_DIR="libtorch_house$DIR_SUFFIX"
18+
WHEELHOUSE_DIR="wheelhousecpu"
19+
LIBTORCH_HOUSE_DIR="libtorch_housecpu"
3120
if [[ -z "$PYTORCH_FINAL_PACKAGE_DIR" ]]; then
3221
if [[ -z "$BUILD_PYTHONLESS" ]]; then
33-
PYTORCH_FINAL_PACKAGE_DIR="/remote/wheelhouse$DIR_SUFFIX"
22+
PYTORCH_FINAL_PACKAGE_DIR="/remote/wheelhousecpu"
3423
else
35-
PYTORCH_FINAL_PACKAGE_DIR="/remote/libtorch_house$DIR_SUFFIX"
24+
PYTORCH_FINAL_PACKAGE_DIR="/remote/libtorch_housecpu"
3625
fi
3726
fi
3827
mkdir -p "$PYTORCH_FINAL_PACKAGE_DIR" || true
@@ -60,24 +49,6 @@ DEPS_SONAME=(
6049
"libgomp.so.1"
6150
)
6251

63-
if [[ "$GPU_ARCH_TYPE" == "xpu" ]]; then
64-
echo "Bundling with xpu support package libs."
65-
DEPS_LIST+=(
66-
"/opt/intel/oneapi/compiler/latest/lib/libOpenCL.so.1"
67-
"/opt/intel/oneapi/compiler/latest/lib/libsvml.so"
68-
"/opt/intel/oneapi/compiler/latest/lib/libirng.so"
69-
"/opt/intel/oneapi/compiler/latest/lib/libimf.so"
70-
"/opt/intel/oneapi/compiler/latest/lib/libintlc.so.5"
71-
)
72-
DEPS_SONAME+=(
73-
"libOpenCL.so.1"
74-
"libsvml.so"
75-
"libirng.so"
76-
"libimf.so"
77-
"libintlc.so.5"
78-
)
79-
fi
80-
8152
rm -rf /usr/local/cuda*
8253

8354
SOURCE_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null && pwd )"

0 commit comments

Comments
 (0)