frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-11-16 07:24:54 +08:00
Code Issues Packages Projects Releases Wiki Activity

Update (base update)

Browse Source 
[ghstack-poisoned]
This commit is contained in:
Xuehai Pan
2024-10-22 13:37:54 +08:00
parent a2170cd153 3b186c5659
commit cecfe68c32
400 changed files with 7432 additions and 5508 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
.ci/pytorch/build.sh
View File

@ -203,7 +203,9 @@ if [[ "${BUILD_ENVIRONMENT}" == *clang* ]]; then
fi
if [[ "$BUILD_ENVIRONMENT" == *-clang*-asan* ]]; then
export USE_CUDA=0
if [[ "$BUILD_ENVIRONMENT" == *cuda* ]]; then
export USE_CUDA=1
fi
export USE_ASAN=1
export REL_WITH_DEB_INFO=1
export UBSAN_FLAGS="-fno-sanitize-recover=all"

7
.ci/pytorch/test.sh
View File

@ -196,6 +196,9 @@ install_tlparse
# ASAN test is not working
if [[ "$BUILD_ENVIRONMENT" == *asan* ]]; then
export ASAN_OPTIONS=detect_leaks=0:symbolize=1:detect_stack_use_after_return=true:strict_init_order=true:detect_odr_violation=1:detect_container_overflow=0:check_initialization_order=true:debug=true
if [[ "$BUILD_ENVIRONMENT" == *cuda* ]]; then
export ASAN_OPTIONS="${ASAN_OPTIONS}:protect_shadow_gap=0"
fi
export UBSAN_OPTIONS=print_stacktrace=1:suppressions=$PWD/ubsan.supp
export PYTORCH_TEST_WITH_ASAN=1
export PYTORCH_TEST_WITH_UBSAN=1
@ -320,6 +323,7 @@ test_inductor_distributed() {
python test/run_test.py -i distributed/test_c10d_functional_native.py --verbose
python test/run_test.py -i distributed/_tensor/test_dtensor_compile.py --verbose
python test/run_test.py -i distributed/tensor/parallel/test_micro_pipeline_tp.py --verbose
python test/run_test.py -i distributed/_composable/test_replicate_with_compiler.py --verbose
python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_comm.py --verbose
python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_training.py -k test_train_parity_multi_group --verbose
python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_training.py -k test_train_parity_with_activation_checkpointing --verbose
@ -331,11 +335,12 @@ test_inductor_distributed() {
python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_mixed_precision.py -k test_compute_dtype --verbose
python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_mixed_precision.py -k test_reduce_dtype --verbose
python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_clip_grad_norm_.py -k test_clip_grad_norm_2d --verbose
python test/run_test.py -i distributed/_composable/fsdp/test_fully_shard_compile.py --verbose
python test/run_test.py -i distributed/fsdp/test_fsdp_tp_integration.py -k test_fsdp_tp_integration --verbose
# this runs on both single-gpu and multi-gpu instance. It should be smart about skipping tests that aren't supported
# with if required # gpus aren't available
python test/run_test.py --include distributed/test_dynamo_distributed distributed/test_inductor_collectives --verbose
python test/run_test.py --include distributed/test_dynamo_distributed distributed/test_inductor_collectives distributed/test_compute_comm_reordering --verbose
assert_git_not_dirty
}

14
.github/actions/upload-test-artifacts/action.yml vendored
View File

@ -28,7 +28,7 @@ runs:
run: |
# Remove any previous test jsons if they exist
rm -f test-jsons-*.zip
zip -r "test-jsons-${FILE_SUFFIX}.zip" test -i '*.json'
zip -r "test-jsons-${FILE_SUFFIX}.zip" test/test-reports -i '*.json'
- name: Zip test reports for upload
if: runner.os != 'Windows' && !inputs.use-gha
@ -38,7 +38,7 @@ runs:
run: |
# Remove any previous test reports if they exist
rm -f test-reports-*.zip
zip -r "test-reports-${FILE_SUFFIX}.zip" test -i '*.xml' -i '*.csv'
zip -r "test-reports-${FILE_SUFFIX}.zip" test/test-reports -i '*.xml' -i '*.csv'
- name: Zip usage log for upload
if: runner.os != 'Windows' && !inputs.use-gha
@ -53,8 +53,8 @@ runs:
if [ -f 'usage_log.txt' ]; then
zip "logs-${FILE_SUFFIX}.zip" 'usage_log.txt'
fi
if ls test/**/*.log 1> /dev/null 2>&1; then
zip -r "logs-${FILE_SUFFIX}.zip" test -i '*.log'
if find "test/test-reports" -name "*.log" 2>/dev/null | grep -q .; then
zip -r "logs-${FILE_SUFFIX}.zip" test/test-reports -i '*.log'
fi
- name: Zip debugging artifacts for upload
@ -77,7 +77,7 @@ runs:
FILE_SUFFIX: ${{ inputs.file-suffix }}
run: |
# -ir => recursive include all files in pattern
7z a "test-jsons-$Env:FILE_SUFFIX.zip" -ir'!test\*.json'
7z a "test-jsons-$Env:FILE_SUFFIX.zip" -ir'!test\test-reports\*.json'
- name: Zip test reports for upload
if: runner.os == 'Windows' && !inputs.use-gha
@ -86,7 +86,7 @@ runs:
FILE_SUFFIX: ${{ inputs.file-suffix }}
run: |
# -ir => recursive include all files in pattern
7z a "test-reports-$Env:FILE_SUFFIX.zip" -ir'!test\*.xml' -ir'!test\*.csv'
7z a "test-reports-$Env:FILE_SUFFIX.zip" -ir'!test\test-reports\*.xml' -ir'!test\test-reports\*.csv'
- name: Zip usage log for upload
if: runner.os == 'Windows' && !inputs.use-gha
@ -96,7 +96,7 @@ runs:
FILE_SUFFIX: ${{ inputs.file-suffix }}
run: |
# -ir => recursive include all files in pattern
7z a "logs-$Env:FILE_SUFFIX.zip" 'usage_log.txt' -ir'!test\*.log'
7z a "logs-$Env:FILE_SUFFIX.zip" 'usage_log.txt' -ir'!test\test-reports\*.log'
# S3 upload
- name: Store Test Downloaded JSONs on S3

2
.github/scripts/generate_binary_build_matrix.py vendored
View File

@ -459,7 +459,7 @@ def generate_wheels_matrix(
".", "_"
),
"pytorch_extra_install_requirements": (
PYTORCH_EXTRA_INSTALL_REQUIREMENTS["12.1"]
PYTORCH_EXTRA_INSTALL_REQUIREMENTS["12.4"]
if os != "linux" and gpu_arch_type != "xpu"
else ""
),

2
.github/workflows/docker-builds.yml vendored
View File

@ -123,7 +123,7 @@ jobs:
IMAGE_NAME: ${{ matrix.docker-image-name }}
with:
shell: bash
timeout_minutes: 15
timeout_minutes: 30
max_attempts: 5
retry_wait_seconds: 90
command: |

8
.github/workflows/generated-linux-aarch64-binary-manywheel-nightly.yml generated vendored
View File

@ -65,7 +65,7 @@ jobs:
ALPINE_IMAGE: "arm64v8/alpine"
build_name: manywheel-py3_9-cpu-aarch64
build_environment: linux-aarch64-binary-manywheel
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
manywheel-py3_9-cpu-aarch64-test: # Testing
@ -185,7 +185,7 @@ jobs:
ALPINE_IMAGE: "arm64v8/alpine"
build_name: manywheel-py3_10-cpu-aarch64
build_environment: linux-aarch64-binary-manywheel
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
manywheel-py3_10-cpu-aarch64-test: # Testing
@ -305,7 +305,7 @@ jobs:
ALPINE_IMAGE: "arm64v8/alpine"
build_name: manywheel-py3_11-cpu-aarch64
build_environment: linux-aarch64-binary-manywheel
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
manywheel-py3_11-cpu-aarch64-test: # Testing
@ -425,7 +425,7 @@ jobs:
ALPINE_IMAGE: "arm64v8/alpine"
build_name: manywheel-py3_12-cpu-aarch64
build_environment: linux-aarch64-binary-manywheel
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
manywheel-py3_12-cpu-aarch64-test: # Testing

10
.github/workflows/generated-linux-s390x-binary-manywheel-nightly.yml generated vendored
View File

@ -64,7 +64,7 @@ jobs:
ALPINE_IMAGE: "docker.io/s390x/alpine"
build_name: manywheel-py3_9-cpu-s390x
build_environment: linux-s390x-binary-manywheel
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
manywheel-py3_9-cpu-s390x-test: # Testing
@ -133,7 +133,7 @@ jobs:
ALPINE_IMAGE: "docker.io/s390x/alpine"
build_name: manywheel-py3_10-cpu-s390x
build_environment: linux-s390x-binary-manywheel
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
manywheel-py3_10-cpu-s390x-test: # Testing
@ -202,7 +202,7 @@ jobs:
ALPINE_IMAGE: "docker.io/s390x/alpine"
build_name: manywheel-py3_11-cpu-s390x
build_environment: linux-s390x-binary-manywheel
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
manywheel-py3_11-cpu-s390x-test: # Testing
@ -271,7 +271,7 @@ jobs:
ALPINE_IMAGE: "docker.io/s390x/alpine"
build_name: manywheel-py3_12-cpu-s390x
build_environment: linux-s390x-binary-manywheel
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
manywheel-py3_12-cpu-s390x-test: # Testing
@ -340,7 +340,7 @@ jobs:
ALPINE_IMAGE: "docker.io/s390x/alpine"
build_name: manywheel-py3_13-cpu-s390x
build_environment: linux-s390x-binary-manywheel
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
secrets:
github-token: ${{ secrets.GITHUB_TOKEN }}
manywheel-py3_13-cpu-s390x-test: # Testing

10
.github/workflows/generated-macos-arm64-binary-wheel-nightly.yml generated vendored
View File

@ -46,7 +46,7 @@ jobs:
GPU_ARCH_TYPE: cpu
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.9"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
# NOTE: These environment variables are put here so that they can be applied on every job equally
# They are also here because setting them at a workflow level doesn't give us access to the
@ -162,7 +162,7 @@ jobs:
GPU_ARCH_TYPE: cpu
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.10"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
# NOTE: These environment variables are put here so that they can be applied on every job equally
# They are also here because setting them at a workflow level doesn't give us access to the
@ -278,7 +278,7 @@ jobs:
GPU_ARCH_TYPE: cpu
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.11"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
# NOTE: These environment variables are put here so that they can be applied on every job equally
# They are also here because setting them at a workflow level doesn't give us access to the
@ -394,7 +394,7 @@ jobs:
GPU_ARCH_TYPE: cpu
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.12"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
# NOTE: These environment variables are put here so that they can be applied on every job equally
# They are also here because setting them at a workflow level doesn't give us access to the
@ -510,7 +510,7 @@ jobs:
GPU_ARCH_TYPE: cpu
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.13"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
# NOTE: These environment variables are put here so that they can be applied on every job equally
# They are also here because setting them at a workflow level doesn't give us access to the

32
.github/workflows/generated-windows-binary-wheel-nightly.yml generated vendored
View File

@ -55,7 +55,7 @@ jobs:
GPU_ARCH_TYPE: cpu
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.9"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
- name: Display EC2 information
shell: bash
@ -322,7 +322,7 @@ jobs:
GPU_ARCH_TYPE: cuda
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.9"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
- name: Display EC2 information
shell: bash
@ -591,7 +591,7 @@ jobs:
GPU_ARCH_TYPE: cuda
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.9"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
- name: Display EC2 information
shell: bash
@ -860,7 +860,7 @@ jobs:
GPU_ARCH_TYPE: cuda
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.9"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
- name: Display EC2 information
shell: bash
@ -1393,7 +1393,7 @@ jobs:
GPU_ARCH_TYPE: cpu
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.10"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
- name: Display EC2 information
shell: bash
@ -1660,7 +1660,7 @@ jobs:
GPU_ARCH_TYPE: cuda
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.10"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
- name: Display EC2 information
shell: bash
@ -1929,7 +1929,7 @@ jobs:
GPU_ARCH_TYPE: cuda
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.10"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
- name: Display EC2 information
shell: bash
@ -2198,7 +2198,7 @@ jobs:
GPU_ARCH_TYPE: cuda
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.10"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
- name: Display EC2 information
shell: bash
@ -2731,7 +2731,7 @@ jobs:
GPU_ARCH_TYPE: cpu
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.11"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
- name: Display EC2 information
shell: bash
@ -2998,7 +2998,7 @@ jobs:
GPU_ARCH_TYPE: cuda
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.11"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
- name: Display EC2 information
shell: bash
@ -3267,7 +3267,7 @@ jobs:
GPU_ARCH_TYPE: cuda
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.11"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
- name: Display EC2 information
shell: bash
@ -3536,7 +3536,7 @@ jobs:
GPU_ARCH_TYPE: cuda
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.11"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
- name: Display EC2 information
shell: bash
@ -4069,7 +4069,7 @@ jobs:
GPU_ARCH_TYPE: cpu
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.12"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
- name: Display EC2 information
shell: bash
@ -4336,7 +4336,7 @@ jobs:
GPU_ARCH_TYPE: cuda
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.12"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
- name: Display EC2 information
shell: bash
@ -4605,7 +4605,7 @@ jobs:
GPU_ARCH_TYPE: cuda
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.12"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
- name: Display EC2 information
shell: bash
@ -4874,7 +4874,7 @@ jobs:
GPU_ARCH_TYPE: cuda
SKIP_ALL_TESTS: 1
DESIRED_PYTHON: "3.12"
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.1.3.1; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.0.2.54; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.2.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.4.5.107; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.1.0.106; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.1.105; platform_system == 'Linux' and platform_machine == 'x86_64'
PYTORCH_EXTRA_INSTALL_REQUIREMENTS: nvidia-cuda-nvrtc-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-runtime-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cuda-cupti-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cudnn-cu12==9.1.0.70; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cublas-cu12==12.4.5.8; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cufft-cu12==11.2.1.3; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-curand-cu12==10.3.5.147; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusolver-cu12==11.6.1.9; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparse-cu12==12.3.1.170; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-cusparselt-cu12==0.6.2; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nccl-cu12==2.21.5; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvtx-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64' | nvidia-nvjitlink-cu12==12.4.127; platform_system == 'Linux' and platform_machine == 'x86_64'
steps:
- name: Display EC2 information
shell: bash

29
.github/workflows/periodic.yml vendored
View File

@ -385,3 +385,32 @@ jobs:
build-environment: linux-focal-cuda11.8-py3.9-gcc9-experimental-split-build
docker-image: ${{ needs.linux-focal-cuda11_8-py3_9-gcc9-experimental-split-build.outputs.docker-image }}
test-matrix: ${{ needs.linux-focal-cuda11_8-py3_9-gcc9-experimental-split-build.outputs.test-matrix }}
linux-focal-cuda11_8-py3_10-gcc9-experimental-split-build:
name: linux-focal-cuda11.8-py3.10-gcc9-experimental-split-build
uses: ./.github/workflows/_linux-build.yml
needs: get-label-type
with:
runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
use_split_build: true
build-environment: linux-focal-cuda11.8-py3.10-gcc9
docker-image-name: pytorch-linux-focal-cuda11.8-cudnn9-py3-gcc9
cuda-arch-list: '7.5'
test-matrix: |
{ include: [
{ config: "distributed", shard: 1, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.g4dn.12xlarge.nvidia.gpu" },
{ config: "distributed", shard: 2, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.g4dn.12xlarge.nvidia.gpu" },
{ config: "distributed", shard: 3, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.g4dn.12xlarge.nvidia.gpu" },
]}
linux-focal-cuda11_8-py3_10-gcc9-experimental-split-build-test:
name: linux-focal-cuda11.8-py3.10-gcc9-experimental-split-build-test
uses: ./.github/workflows/_linux-test.yml
needs:
- linux-focal-cuda11_8-py3_10-gcc9-experimental-split-build
- target-determination
with:
timeout-minutes: 360
build-environment: linux-focal-cuda11.8-py3.10-gcc9-experimental-split-build
docker-image: ${{ needs.linux-focal-cuda11_8-py3_10-gcc9-experimental-split-build.outputs.docker-image }}
test-matrix: ${{ needs.linux-focal-cuda11_8-py3_10-gcc9-experimental-split-build.outputs.test-matrix }}

7
.github/workflows/pull.yml vendored
View File

@ -280,11 +280,12 @@ jobs:
runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
build-environment: linux-focal-cuda11.8-py3.10-gcc9
docker-image-name: pytorch-linux-focal-cuda11.8-cudnn9-py3-gcc9
cuda-arch-list: '7.5'
test-matrix: |
{ include: [
{ config: "distributed", shard: 1, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.8xlarge.nvidia.gpu" },
{ config: "distributed", shard: 2, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.8xlarge.nvidia.gpu" },
{ config: "distributed", shard: 3, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.8xlarge.nvidia.gpu" },
{ config: "distributed", shard: 1, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.g4dn.12xlarge.nvidia.gpu" },
{ config: "distributed", shard: 2, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.g4dn.12xlarge.nvidia.gpu" },
{ config: "distributed", shard: 3, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.g4dn.12xlarge.nvidia.gpu" },
]}
secrets: inherit

28
.github/workflows/trunk.yml vendored
View File

@ -288,31 +288,3 @@ jobs:
build-environment: linux-focal-cuda12.4-py3.10-gcc9-experimental-split-build
docker-image: ${{ needs.linux-focal-cuda12_4-py3_10-gcc9-experimental-split-build.outputs.docker-image }}
test-matrix: ${{ needs.linux-focal-cuda12_4-py3_10-gcc9-experimental-split-build.outputs.test-matrix }}
linux-focal-cuda11_8-py3_10-gcc9-experimental-split-build:
name: linux-focal-cuda11.8-py3.10-gcc9-experimental-split-build
uses: ./.github/workflows/_linux-build.yml
needs: get-label-type
with:
runner_prefix: "${{ needs.get-label-type.outputs.label-type }}"
use_split_build: true
build-environment: linux-focal-cuda11.8-py3.10-gcc9
docker-image-name: pytorch-linux-focal-cuda11.8-cudnn9-py3-gcc9
test-matrix: |
{ include: [
{ config: "distributed", shard: 1, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.8xlarge.nvidia.gpu" },
{ config: "distributed", shard: 2, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.8xlarge.nvidia.gpu" },
{ config: "distributed", shard: 3, num_shards: 3, runner: "${{ needs.get-label-type.outputs.label-type }}linux.8xlarge.nvidia.gpu" },
]}
linux-focal-cuda11_8-py3_10-gcc9-experimental-split-build-test:
name: linux-focal-cuda11.8-py3.10-gcc9-experimental-split-build-test
uses: ./.github/workflows/_linux-test.yml
needs:
- linux-focal-cuda11_8-py3_10-gcc9-experimental-split-build
- target-determination
with:
timeout-minutes: 360
build-environment: linux-focal-cuda11.8-py3.10-gcc9-experimental-split-build
docker-image: ${{ needs.linux-focal-cuda11_8-py3_10-gcc9-experimental-split-build.outputs.docker-image }}
test-matrix: ${{ needs.linux-focal-cuda11_8-py3_10-gcc9-experimental-split-build.outputs.test-matrix }}

85
.lintrunner.toml
View File

@ -241,7 +241,9 @@ exclude_patterns = [
'c10/util/*inl.h',
'c10/test/**/*.h',
'third_party/**/*',
'torch/csrc/api/**',
'torch/csrc/api/include/torch/nn/modules/common.h',
'torch/csrc/api/include/torch/linalg.h',
'torch/csrc/api/include/torch/nn/pimpl-inl.h',
'torch/csrc/autograd/generated/**',
'torch/csrc/distributed/**/*',
'torch/csrc/dynamo/eval_frame.h',
@ -1230,87 +1232,6 @@ exclude_patterns = [
'torch/fft/__init__.py',
'torch/func/__init__.py',
'torch/futures/__init__.py',
'torch/fx/__init__.py',
'torch/fx/_compatibility.py',
'torch/fx/_symbolic_trace.py',
'torch/fx/annotate.py',
'torch/fx/config.py',
'torch/fx/experimental/__init__.py',
'torch/fx/experimental/accelerator_partitioner.py',
'torch/fx/experimental/const_fold.py',
'torch/fx/experimental/debug.py',
'torch/fx/experimental/graph_gradual_typechecker.py',
'torch/fx/experimental/merge_matmul.py',
'torch/fx/experimental/meta_tracer.py',
'torch/fx/experimental/migrate_gradual_types/__init__.py',
'torch/fx/experimental/migrate_gradual_types/constraint.py',
'torch/fx/experimental/migrate_gradual_types/constraint_generator.py',
'torch/fx/experimental/migrate_gradual_types/constraint_transformation.py',
'torch/fx/experimental/migrate_gradual_types/operation.py',
'torch/fx/experimental/migrate_gradual_types/transform_to_z3.py',
'torch/fx/experimental/migrate_gradual_types/util.py',
'torch/fx/experimental/migrate_gradual_types/z3_types.py',
'torch/fx/experimental/normalize.py',
'torch/fx/experimental/optimization.py',
'torch/fx/experimental/partitioner_utils.py',
'torch/fx/experimental/refinement_types.py',
'torch/fx/experimental/rewriter.py',
'torch/fx/experimental/schema_type_annotation.py',
'torch/fx/experimental/unification/__init__.py',
'torch/fx/experimental/unification/core.py',
'torch/fx/experimental/unification/dispatch.py',
'torch/fx/experimental/unification/match.py',
'torch/fx/experimental/unification/more.py',
'torch/fx/experimental/unification/multipledispatch/__init__.py',
'torch/fx/experimental/unification/multipledispatch/conflict.py',
'torch/fx/experimental/unification/multipledispatch/core.py',
'torch/fx/experimental/unification/multipledispatch/dispatcher.py',
'torch/fx/experimental/unification/multipledispatch/utils.py',
'torch/fx/experimental/unification/multipledispatch/variadic.py',
'torch/fx/experimental/unification/unification_tools.py',
'torch/fx/experimental/unification/utils.py',
'torch/fx/experimental/unification/variable.py',
'torch/fx/experimental/unify_refinements.py',
'torch/fx/graph.py',
'torch/fx/graph_module.py',
'torch/fx/interpreter.py',
'torch/fx/node.py',
'torch/fx/operator_schemas.py',
'torch/fx/passes/__init__.py',
'torch/fx/passes/annotate_getitem_nodes.py',
'torch/fx/passes/backends/__init__.py',
'torch/fx/passes/backends/cudagraphs.py',
'torch/fx/passes/dialect/__init__.py',
'torch/fx/passes/dialect/common/__init__.py',
'torch/fx/passes/dialect/common/cse_pass.py',
'torch/fx/passes/fake_tensor_prop.py',
'torch/fx/passes/graph_drawer.py',
'torch/fx/passes/graph_manipulation.py',
'torch/fx/passes/infra/__init__.py',
'torch/fx/passes/infra/partitioner.py',
'torch/fx/passes/infra/pass_base.py',
'torch/fx/passes/infra/pass_manager.py',
'torch/fx/passes/net_min_base.py',
'torch/fx/passes/operator_support.py',
'torch/fx/passes/param_fetch.py',
'torch/fx/passes/pass_manager.py',
'torch/fx/passes/reinplace.py',
'torch/fx/passes/shape_prop.py',
'torch/fx/passes/split_module.py',
'torch/fx/passes/split_utils.py',
'torch/fx/passes/splitter_base.py',
'torch/fx/passes/tests/__init__.py',
'torch/fx/passes/tests/test_pass_manager.py',
'torch/fx/passes/tools_common.py',
'torch/fx/passes/utils/__init__.py',
'torch/fx/passes/utils/common.py',
'torch/fx/passes/utils/fuser_utils.py',
'torch/fx/passes/utils/matcher_utils.py',
'torch/fx/passes/utils/source_matcher_utils.py',
'torch/fx/proxy.py',
'torch/fx/subgraph_rewriter.py',
'torch/fx/tensor_type.py',
'torch/fx/traceback.py',
'torch/linalg/__init__.py',
'torch/monitor/__init__.py',
'torch/nested/__init__.py',

2
aten/src/ATen/CPUApplyUtils.h
View File

@ -136,7 +136,7 @@ inline bool _apply_preamble(ArrayRef<Tensor> tensors) {
checkDeviceType("CPU_tensor_apply", tensors, kCPU);
checkLayout("CPU_tensor_apply", tensors, kStrided);
if (!_all_equal_numel(tensors))
AT_ERROR(_all_equal_numel_error(tensors));
TORCH_CHECK(false, _all_equal_numel_error(tensors));
// An empty tensor has no elements
for (auto& t : tensors)
if (t.numel() == 0)

4
aten/src/ATen/CPUFixedAllocator.h
View File

@ -12,11 +12,11 @@
namespace at {
static cpu_fixed_malloc(void*, ptrdiff_t) {
AT_ERROR("attempting to resize a tensor view of an external blob");
TORCH_CHECK(false, "attempting to resize a tensor view of an external blob");
}
static cpu_fixed_realloc(void*, void*, ptrdiff_t) {
AT_ERROR("attempting to resize a tensor view of an external blob");
TORCH_CHECK(false, "attempting to resize a tensor view of an external blob");
}
static cpu_fixed_free(void* state, void* allocation) {

2
aten/src/ATen/CPUGeneratorImpl.cpp
View File

@ -189,7 +189,7 @@ void CPUGeneratorImpl::set_state(const c10::TensorImpl& new_state) {
double_normal_sample = std::optional<double>(legacy_pod->normal_y);
}
} else {
AT_ERROR("Expected either a CPUGeneratorImplStateLegacy of size ", size_legacy,
TORCH_CHECK(false, "Expected either a CPUGeneratorImplStateLegacy of size ", size_legacy,
" or a CPUGeneratorImplState of size ", size_current,
" but found the input RNG state size to be ", new_state_size);
}

30
aten/src/ATen/Context.h
View File

@ -43,19 +43,9 @@ class TORCH_API Context {
if (device_type == at::kCPU) {
return at::detail::getDefaultCPUGenerator();
} else if (device_type == at::kCUDA) {
return at::detail::getCUDAHooks().getDefaultCUDAGenerator(device.index());
} else if (device_type == at::kMPS) {
return at::detail::getMPSHooks().getDefaultMPSGenerator();
} else if (device_type == at::kXPU) {
return at::detail::getXPUHooks().getDefaultXPUGenerator(device.index());
} else if (device_type == at::kIPU) {
return at::detail::getIPUHooks().getDefaultIPUGenerator(device.index());
} else if (device_type == at::kPrivateUse1) {
return at::detail::getPrivateUse1Hooks().getDefaultGenerator(
device.index());
} else {
AT_ERROR(c10::DeviceTypeName(device_type), " device type not enabled.");
return getAcceleratorHooksInterface(device_type)
.getDefaultGenerator(device.index());
}
}
@ -77,8 +67,10 @@ class TORCH_API Context {
} else if (device_type == at::kHIP) {
return at::detail::getHIPHooks();
} else {
AT_ERROR(
c10::DeviceTypeName(device_type), " device type not an accelerator.");
TORCH_CHECK(
false,
c10::DeviceTypeName(device_type),
" device type not an accelerator.");
}
}
@ -349,18 +341,28 @@ class TORCH_API Context {
// Preserved for BC
void lazyInitCUDA() {
TORCH_WARN_DEPRECATION(
"lazyInitCUDA is deprecated. Please use lazyInitDevice(at::kCUDA) instead.")
lazyInitDevice(at::kCUDA);
}
void lazyInitHIP() {
TORCH_WARN_DEPRECATION(
"lazyInitHIP is deprecated. Please use lazyInitDevice(at::kHIP) instead.")
lazyInitDevice(at::kHIP);
}
void lazyInitXPU() {
TORCH_WARN_DEPRECATION(
"lazyInitXPU is deprecated. Please use lazyInitDevice(at::kXPU) instead.")
lazyInitDevice(at::kXPU);
}
void lazyInitMTIA() {
TORCH_WARN_DEPRECATION(
"lazyInitMTIA is deprecated. Please use lazyInitDevice(at::kMTIA) instead.")
lazyInitDevice(at::kMTIA);
}
void lazyInitPrivateUse1() {
TORCH_WARN_DEPRECATION(
"lazyInitPrivateUse1 is deprecated. Please use lazyInitDevice(at::kPrivateUse1) instead.")
lazyInitDevice(at::kPrivateUse1);
}

6
aten/src/ATen/DLConvertor.h
View File

@ -13,10 +13,12 @@ namespace at {
TORCH_API ScalarType toScalarType(const DLDataType& dtype);
TORCH_API DLManagedTensor* toDLPack(const Tensor& src);
TORCH_API Tensor fromDLPack(DLManagedTensor* src);
C10_DEPRECATED_MESSAGE("Please migrate to a non-const variant")
inline Tensor fromDLPack(const DLManagedTensor* src) {
[[deprecated("Please migrate to a non-const variant")]] inline Tensor fromDLPack(
const DLManagedTensor* src) {
return fromDLPack(const_cast<DLManagedTensor*>(src));
}
TORCH_API Tensor
fromDLPack(DLManagedTensor* src, std::function<void(void*)> deleter);
TORCH_API DLDataType getDLDataType(const Tensor& t);

24
aten/src/ATen/Dispatch.h
View File

@ -55,7 +55,8 @@ TORCH_API void record_kernel_function_dtype(std::string name);
do { \
if constexpr (!at::should_include_kernel_dtype( \
at_dispatch_name, enum_type)) { \
AT_ERROR( \
TORCH_CHECK( \
false, \
"dtype '", \
toString(enum_type), \
"' not selected for kernel tag ", \
@ -103,23 +104,23 @@ inline at::ScalarType scalar_type(at::ScalarType s) {
return s;
}
C10_DEPRECATED_MESSAGE(
[[deprecated(
"passing at::DeprecatedTypeProperties to an AT_DISPATCH macro is deprecated, "
"pass an at::ScalarType instead")
inline at::ScalarType scalar_type(const at::DeprecatedTypeProperties& t) {
"pass an at::ScalarType instead")]] inline at::ScalarType
scalar_type(const at::DeprecatedTypeProperties& t) {
return t.scalarType();
}
C10_DEPRECATED_MESSAGE(
[[deprecated(
"AT_DISPATCH_ALL_TYPES_AND_HALF is deprecated, "
"use AT_DISPATCH_ALL_TYPES_AND(at::ScalarType::Half, ...) instead")
inline void deprecated_AT_DISPATCH_ALL_TYPES_AND_HALF() {}
"use AT_DISPATCH_ALL_TYPES_AND(at::ScalarType::Half, ...) instead")]] inline void
deprecated_AT_DISPATCH_ALL_TYPES_AND_HALF() {}
C10_DEPRECATED_MESSAGE(
[[deprecated(
"AT_DISPATCH_ALL_TYPES_AND_HALF_AND_COMPLEX is deprecated, "
"use AT_DISPATCH_ALL_TYPES_AND_COMPLEX_AND(at::ScalarType::Half, ...) "
"instead")
inline void deprecated_AT_DISPATCH_ALL_TYPES_AND_HALF_AND_COMPLEX() {}
"instead")]] inline void
deprecated_AT_DISPATCH_ALL_TYPES_AND_HALF_AND_COMPLEX() {}
} // namespace detail
@ -220,7 +221,8 @@ inline void deprecated_AT_DISPATCH_ALL_TYPES_AND_HALF_AND_COMPLEX() {}
switch (_st) { \
__VA_ARGS__ \
default: \
AT_ERROR( \
TORCH_CHECK( \
false, \
'"', \
at_dispatch_name, \
"\" not implemented for '", \

2
aten/src/ATen/ExpandUtils.h
View File

@ -78,7 +78,7 @@ inline void check_defined(
const char* api_name) {
for (auto& t : tensors) {
if (!t.get().defined()) {
AT_ERROR(api_name, "(...) called with an undefined Tensor");
TORCH_CHECK(false, api_name, "(...) called with an undefined Tensor");
}
}
}

2
aten/src/ATen/FunctionalInverses.cpp
View File

@ -231,6 +231,7 @@ Tensor FunctionalInverses::slice_Tensor_inverse(const Tensor& base, const Tensor
}
}
// NOLINTNEXTLINE(performance-unnecessary-value-param)
Tensor FunctionalInverses::split_Tensor_inverse(const Tensor& base, const Tensor& mutated_view, InverseReturnMode inverse_return_mode, int64_t mutated_view_idx, c10::SymInt split_size, int64_t dim) {
// It would be nice if this logic could be re-used from autograd's split_backward(), but I don't think it can.
// For functionalization, we have only have one of the tensors from the TensorList outputed by split(), and we want to layer i
@ -452,6 +453,7 @@ Tensor FunctionalInverses::chunk_inverse(const at::Tensor & base, const at::Tens
return split_with_sizes_inverse(base, mutated_view, inverse_return_mode, mutated_view_idx, split_sizes, dim);
}
// NOLINTNEXTLINE(performance-unnecessary-value-param)
Tensor FunctionalInverses::narrow_inverse(const at::Tensor & base, const at::Tensor & mutated_view, InverseReturnMode inverse_return_mode, int dim, c10::SymInt start, c10::SymInt length) {
if (inverse_return_mode == InverseReturnMode::AlwaysView) {
// NB: assumes mutated_view is a narrowed view of base.

2
aten/src/ATen/InferSize.h
View File

@ -33,7 +33,7 @@ inline void infer_size_impl(
} else if (shape[dim] >= 0) {
newsize *= shape[dim];
} else {
AT_ERROR("invalid shape dimension ", shape[dim]);
TORCH_CHECK(false, "invalid shape dimension ", shape[dim]);
}
}

6
aten/src/ATen/OpaqueTensorImpl.h
View File

@ -45,15 +45,15 @@ struct TORCH_API OpaqueTensorImpl : public TensorImpl {
}
void set_size(int64_t dim, int64_t new_size) override {
AT_ERROR("opaque tensors do not have set_size");
TORCH_CHECK(false, "opaque tensors do not have set_size");
}
void set_stride(int64_t dim, int64_t new_stride) override {
AT_ERROR("opaque tensors do not have set_stride");
TORCH_CHECK(false, "opaque tensors do not have set_stride");
}
void set_storage_offset(int64_t storage_offset) override {
AT_ERROR("opaque tensors do not have set_storage_offset");
TORCH_CHECK(false, "opaque tensors do not have set_storage_offset");
}
#ifdef DEBUG

21
aten/src/ATen/SparseCsrTensorUtils.h
View File

@ -23,7 +23,8 @@
case kSparseBsc: \
return __VA_ARGS__(); \
default: \
AT_ERROR( \
TORCH_CHECK( \
false, \
NAME, \
" expected sparse compressed tensor layout but got ", \
the_layout); \
@ -42,7 +43,8 @@
case kSparseBsc: \
return (COLUMN_DIM_ACTION)(); \
default: \
AT_ERROR( \
TORCH_CHECK( \
false, \
NAME, \
" expected sparse compressed tensor layout but got ", \
the_layout); \
@ -61,7 +63,8 @@
case kSparseBsc: \
return (BLOCK_ACTION)(); \
default: \
AT_ERROR( \
TORCH_CHECK( \
false, \
NAME, \
" expected sparse compressed tensor layout but got ", \
the_layout); \
@ -77,7 +80,8 @@
case kSparseBsr: \
return (ROW_DIM_ACTION)(); \
default: \
AT_ERROR( \
TORCH_CHECK( \
false, \
NAME, \
" expected sparse row compressed tensor layout but got ", \
the_layout); \
@ -93,7 +97,8 @@
case kSparseBsc: \
return (COL_DIM_ACTION)(); \
default: \
AT_ERROR( \
TORCH_CHECK( \
false, \
NAME, \
" expected sparse column compressed tensor layout but got ", \
the_layout); \
@ -108,7 +113,8 @@
case kSparseCsc: \
return (ACTION)(); \
default: \
AT_ERROR( \
TORCH_CHECK( \
false, \
NAME, \
" expected sparse compressed (non-block) tensor layout but got ", \
the_layout); \
@ -123,7 +129,8 @@
case kSparseBsc: \
return (ACTION)(); \
default: \
AT_ERROR( \
TORCH_CHECK( \
false, \
NAME, \
" expected sparse compressed block tensor layout but got ", \
the_layout); \

6
aten/src/ATen/SparseTensorImpl.cpp
View File

@ -57,13 +57,13 @@ void SparseTensorImpl::release_resources() {
}
void SparseTensorImpl::set_size(int64_t dim, int64_t new_size) {
AT_ERROR("sparse tensors do not have set_size");
TORCH_CHECK(false, "sparse tensors do not have set_size");
}
void SparseTensorImpl::set_stride(int64_t dim, int64_t new_stride) {
AT_ERROR("sparse tensors do not have set_stride");
TORCH_CHECK(false, "sparse tensors do not have set_stride");
}
void SparseTensorImpl::set_storage_offset(int64_t storage_offset) {
AT_ERROR("sparse tensors do not have set_storage_offset");
TORCH_CHECK(false, "sparse tensors do not have set_storage_offset");
}
#ifdef DEBUG
bool SparseTensorImpl::has_storage() const {

4
aten/src/ATen/TensorUtils.cpp
View File

@ -155,7 +155,7 @@ void checkSameGPU(CheckedFrom c, const TensorArg& t1, const TensorArg& t2) {
}
oss << "but expected " << ((!t1->is_cpu() && !t2->is_cpu()) ? "them" : "it")
<< " to be on GPU (while checking arguments for " << c << ")";
AT_ERROR(oss.str());
TORCH_CHECK(false, oss.str());
}
TORCH_CHECK(
t1->get_device() == t2->get_device(),
@ -200,7 +200,7 @@ void checkScalarTypes(CheckedFrom c, const TensorArg& t,
}
oss << "; but got " << t->toString()
<< " instead (while checking arguments for " << c << ")";
AT_ERROR(oss.str());
TORCH_CHECK(false, oss.str());
}
}

12
aten/src/ATen/Utils.h
View File

@ -36,7 +36,8 @@ inline std::vector<TensorImpl*> checked_dense_tensor_list_unwrap(
for (const auto i : c10::irange(tensors.size())) {
const auto& expr = tensors[i];
if (expr.layout() != Layout::Strided) {
AT_ERROR(
TORCH_CHECK(
false,
"Expected dense tensor but got ",
expr.layout(),
" for sequence element ",
@ -48,7 +49,8 @@ inline std::vector<TensorImpl*> checked_dense_tensor_list_unwrap(
"'");
}
if (expr.device().type() != device_type) {
AT_ERROR(
TORCH_CHECK(
false,
"Expected object of device type ",
device_type,
" but got device type ",
@ -62,7 +64,8 @@ inline std::vector<TensorImpl*> checked_dense_tensor_list_unwrap(
"'");
}
if (expr.scalar_type() != scalar_type) {
AT_ERROR(
TORCH_CHECK(
false,
"Expected object of scalar type ",
scalar_type,
" but got scalar type ",
@ -96,7 +99,8 @@ std::array<int64_t, N> check_intlist(
return res;
}
if (list.size() != N) {
AT_ERROR(
TORCH_CHECK(
false,
"Expected a list of ",
N,
" ints but got ",

2
aten/src/ATen/autocast_mode.cpp
View File

@ -149,7 +149,7 @@ Banned functions
*******************************/
static Tensor binary_cross_entropy_banned(const Tensor &, const Tensor &, const std::optional<Tensor>&, int64_t) {
AT_ERROR("torch.nn.functional.binary_cross_entropy and torch.nn.BCELoss are unsafe to autocast.\n"
TORCH_CHECK(false, "torch.nn.functional.binary_cross_entropy and torch.nn.BCELoss are unsafe to autocast.\n"
"Many models use a sigmoid layer right before the binary cross entropy layer.\n"
"In this case, combine the two layers using torch.nn.functional.binary_cross_entropy_with_logits\n"
"or torch.nn.BCEWithLogitsLoss. binary_cross_entropy_with_logits and BCEWithLogits are\n"

53
aten/src/ATen/autocast_mode.h
View File

@ -23,36 +23,37 @@ TORCH_API bool is_autocast_cache_enabled();
TORCH_API void set_autocast_cache_enabled(bool enabled);
// deprecated CUDA-specific autocast APIs
C10_DEPRECATED_MESSAGE(
"at::autocast::is_enabled() is deprecated. Please use at::autocast::is_autocast_enabled(at::kCUDA) instead.")
TORCH_API inline bool is_enabled() {
[[deprecated(
"at::autocast::is_enabled() is deprecated. Please use at::autocast::is_autocast_enabled(at::kCUDA) instead.")]] TORCH_API inline bool
is_enabled() {
TORCH_WARN_DEPRECATION(
"at::autocast::",
__func__,
"() is deprecated. Please use at::autocast::is_autocast_enabled(at::kCUDA) instead.")
return is_autocast_enabled(at::kCUDA);
}
C10_DEPRECATED_MESSAGE(
"at::autocast::set_enabled(enabled) is deprecated. Please use at::autocast::set_autocast_enabled(at::kCUDA, enabled) instead.")
TORCH_API inline void set_enabled(bool enabled) {
[[deprecated(
"at::autocast::set_enabled(enabled) is deprecated. Please use at::autocast::set_autocast_enabled(at::kCUDA, enabled) instead.")]] TORCH_API inline void
set_enabled(bool enabled) {
TORCH_WARN_DEPRECATION(
"at::autocast::",
__func__,
"(enabled) is deprecated. Please use at::autocast::set_autocast_enabled(at::kCUDA, enabled) instead.")
set_autocast_enabled(at::kCUDA, enabled);
}
C10_DEPRECATED_MESSAGE(
"at::autocast::get_autocast_gpu_dtype() is deprecated. Please use at::autocast::get_autocast_dtype(at::kCUDA) instead.")
TORCH_API inline at::ScalarType get_autocast_gpu_dtype() {
[[deprecated(
"at::autocast::get_autocast_gpu_dtype() is deprecated. Please use at::autocast::get_autocast_dtype(at::kCUDA) instead.")]] TORCH_API inline at::
ScalarType
get_autocast_gpu_dtype() {
TORCH_WARN_DEPRECATION(
"at::autocast::",
__func__,
"() is deprecated. Please use at::autocast::get_autocast_dtype(at::kCUDA) instead.")
return get_autocast_dtype(at::kCUDA);
}
C10_DEPRECATED_MESSAGE(
"at::autocast::set_autocast_gpu_dtype(dtype) is deprecated. Please use at::autocast::set_autocast_dtype(at::kCUDA, dtype) instead.")
TORCH_API inline void set_autocast_gpu_dtype(at::ScalarType dtype) {
[[deprecated(
"at::autocast::set_autocast_gpu_dtype(dtype) is deprecated. Please use at::autocast::set_autocast_dtype(at::kCUDA, dtype) instead.")]] TORCH_API inline void
set_autocast_gpu_dtype(at::ScalarType dtype) {
TORCH_WARN_DEPRECATION(
"at::autocast::",
__func__,
@ -61,11 +62,10 @@ TORCH_API inline void set_autocast_gpu_dtype(at::ScalarType dtype) {
}
#define DECLARE_DEPRECATED_AUTOCAST_APIS(name, device_type) \
C10_DEPRECATED_MESSAGE( \
[[deprecated( \
"at::autocast::is_" #name \
"_enabled() is deprecated. Please use at::autocast::is_autocast_enabled(" #device_type \
") instead.") \
TORCH_API inline bool is_##name##_enabled() { \
") instead.")]] TORCH_API inline bool is_##name##_enabled() { \
TORCH_WARN_DEPRECATION( \
"at::autocast::", \
__func__, \
@ -74,11 +74,11 @@ TORCH_API inline void set_autocast_gpu_dtype(at::ScalarType dtype) {
return is_autocast_enabled(device_type); \
} \
\
C10_DEPRECATED_MESSAGE( \
[[deprecated( \
"at::autocast::set_" #name \
"_enabled(enabled) is deprecated. Please use at::autocast::set_autocast_enabled(" #device_type \
", enabled) instead.") \
TORCH_API inline void set_##name##_enabled(bool enabled) { \
", enabled) instead.")]] TORCH_API inline void \
set_##name##_enabled(bool enabled) { \
TORCH_WARN_DEPRECATION( \
"at::autocast::", \
__func__, \
@ -87,11 +87,11 @@ TORCH_API inline void set_autocast_gpu_dtype(at::ScalarType dtype) {
set_autocast_enabled(device_type, enabled); \
} \
\
C10_DEPRECATED_MESSAGE( \
[[deprecated( \
"at::autocast::get_autocast_" #name \
"_dtype() is deprecated. Please use at::autocast::get_autocast_dtype(" #device_type \
") instead.") \
TORCH_API inline at::ScalarType get_autocast_##name##_dtype() { \
") instead.")]] TORCH_API inline at::ScalarType \
get_autocast_##name##_dtype() { \
TORCH_WARN_DEPRECATION( \
"at::autocast::", \
__func__, \
@ -100,11 +100,11 @@ TORCH_API inline void set_autocast_gpu_dtype(at::ScalarType dtype) {
return get_autocast_dtype(device_type); \
} \
\
C10_DEPRECATED_MESSAGE( \
[[deprecated( \
"at::autocast::set_autocast_" #name \
"_dtype(dtype) is deprecated. Please use at::autocast::set_autocast_dtype(" #device_type \
", dtype) instead.") \
TORCH_API inline void set_autocast_##name##_dtype(at::ScalarType dtype) { \
", dtype) instead.")]] TORCH_API inline void \
set_autocast_##name##_dtype(at::ScalarType dtype) { \
TORCH_WARN_DEPRECATION( \
"at::autocast::", \
__func__, \
@ -211,7 +211,7 @@ inline at::ScalarType prioritize(
const Tensor& nextArg,
c10::DeviceType device_type = c10::DeviceType::CUDA) {
if (current == at::kDouble) {
AT_ERROR("promote type is double in at::autocast::prioritize");
TORCH_CHECK(false, "promote type is double in at::autocast::prioritize");
return current;
}
at::ScalarType lower_precision_fp =
@ -225,7 +225,8 @@ inline at::ScalarType prioritize(
} else if (current == lower_precision_fp && next == lower_precision_fp) {
return lower_precision_fp;
} else {
AT_ERROR("Unexpected floating ScalarType in at::autocast::prioritize");
TORCH_CHECK(
false, "Unexpected floating ScalarType in at::autocast::prioritize");
return current;
}
} else {

18
aten/src/ATen/core/DistributionsHelper.h
View File

@ -95,11 +95,9 @@ struct uniform_int_distribution {
template <typename T>
struct uniform_real_distribution {
C10_HOST_DEVICE inline uniform_real_distribution(T from, T to) {
C10_HOST_DEVICE inline uniform_real_distribution(T from, T to) : from_(from), to_(to) {
TORCH_CHECK_IF_NOT_ON_CUDA(from <= to);
TORCH_CHECK_IF_NOT_ON_CUDA(to - from <= std::numeric_limits<T>::max());
from_ = from;
to_ = to;
}
template <typename RNG>
@ -186,10 +184,8 @@ DISTRIBUTION_HELPER_GENERATE_NEXT_NORMAL_METHODS(float);
template <typename T>
struct normal_distribution {
C10_HOST_DEVICE inline normal_distribution(T mean_in, T stdv_in) {
C10_HOST_DEVICE inline normal_distribution(T mean_in, T stdv_in) : mean(mean_in), stdv(stdv_in) {
TORCH_CHECK_IF_NOT_ON_CUDA(stdv_in >= 0, "stdv_in must be positive: ", stdv_in);
mean = mean_in;
stdv = stdv_in;
}
template <typename RNG>
@ -236,9 +232,8 @@ template <> struct DiscreteDistributionType<double> { using type = double; };
template <typename T>
struct bernoulli_distribution {
C10_HOST_DEVICE inline bernoulli_distribution(T p_in) {
C10_HOST_DEVICE inline bernoulli_distribution(T p_in) : p(p_in) {
TORCH_CHECK_IF_NOT_ON_CUDA(p_in >= 0 && p_in <= 1);
p = p_in;
}
template <typename RNG>
@ -257,9 +252,8 @@ struct bernoulli_distribution {
template <typename T>
struct geometric_distribution {
C10_HOST_DEVICE inline geometric_distribution(T p_in) {
C10_HOST_DEVICE inline geometric_distribution(T p_in) : p(p_in) {
TORCH_CHECK_IF_NOT_ON_CUDA(p_in > 0 && p_in < 1);
p = p_in;
}
template <typename RNG>
@ -317,10 +311,8 @@ struct cauchy_distribution {
template <typename T>
struct lognormal_distribution {
C10_HOST_DEVICE inline lognormal_distribution(T mean_in, T stdv_in) {
C10_HOST_DEVICE inline lognormal_distribution(T mean_in, T stdv_in) : mean(mean_in), stdv(stdv_in) {
TORCH_CHECK_IF_NOT_ON_CUDA(stdv_in > 0);
mean = mean_in;
stdv = stdv_in;
}
template<typename RNG>

3
aten/src/ATen/core/TensorAccessor.h
View File

@ -263,9 +263,8 @@ public:
// Can't put this directly into the macro function args because of commas
#define AT_X GenericPackedTensorAccessor<T, N, PtrTraits, index_t>
// Old name for `GenericPackedTensorAccessor`
template <typename T, size_t N, template <typename U> class PtrTraits = DefaultPtrTraits, typename index_t = int64_t>
C10_DEFINE_DEPRECATED_USING(PackedTensorAccessor, AT_X)
using PackedTensorAccessor [[deprecated("Old name for `GenericPackedTensorAccessor`")]] = AT_X;
#undef AT_X

2
aten/src/ATen/core/enum_type.h
View File

@ -28,7 +28,7 @@ struct TORCH_API EnumType : public NamedType {
std::move(enum_names_values),
std::move(cu)));
default:
AT_ERROR(
TORCH_CHECK(false,
"Cannot create Enum with value type '",
value->str(),
"', only int, float and string are supported");

2
aten/src/ATen/core/function_schema_inl.h
View File

@ -55,7 +55,7 @@ inline void FunctionSchema::checkAndNormalizeInputs(
inputs.push_back(*argument.default_value());
continue;
}
AT_ERROR(
TORCH_CHECK(false,
name(),
"() is missing value for argument '",
argument.name(),

8
aten/src/ATen/core/ivalue.cpp
View File

@ -756,7 +756,7 @@ IValueComparator getLessThanComparator(const IValue& v) {
torch::jit::Function* lt_func =
checkObjectSortSchema(v.type()->expect<ClassType>(), why_not);
if (!lt_func) {
AT_ERROR(why_not.str());
TORCH_CHECK(false, why_not.str());
}
return [lt_func](const IValue& a, const IValue& b) {
@ -772,7 +772,7 @@ IValueComparator getLessThanComparator(const IValue& v) {
};
}
AT_ERROR("IValues of type: ", v.tagKind(), " are not comparable");
TORCH_CHECK(false, "IValues of type: ", v.tagKind(), " are not comparable");
}
IValueComparator getGreaterThanComparator(const IValue& v) {
@ -967,7 +967,7 @@ IValue IValue::deepcopy(
copy = *this;
} break;
default: {
AT_ERROR("Can't deepcopy IValue with tag: ", tagKind());
TORCH_CHECK(false, "Can't deepcopy IValue with tag: ", tagKind());
}
}
// NB: this doesn't work if an object contains itself, and it may
@ -1050,7 +1050,7 @@ c10::intrusive_ptr<ivalue::Object> ivalue::Object::deepcopy(
}
err << ". Please define serialization methods via def_pickle() for "
"this class.";
AT_ERROR(err.str());
TORCH_CHECK(false, err.str());
}
object->setSlot(i, slots_[i].deepcopy(memo, device));
}

11
aten/src/ATen/core/ivalue.h
View File

@ -809,12 +809,9 @@ struct TORCH_API IValue final {
IValue(c10::Dict<Key, Value> v);
template <class Key, class Value>
/// \cond
/// DOXYGEN_CANNOT_HANDLE_CONSTRUCTORS_WITH_MACROS_SO_EXCLUDE_THIS_LINE_FROM_DOXYGEN
C10_DEPRECATED_MESSAGE(
"IValues based on std::unordered_map<K, V> are slow and deprecated. Please use c10::Dict<K, V> instead.")
/// \endcond
IValue(std::unordered_map<Key, Value> v);
[[deprecated(
"IValues based on std::unordered_map<K, V> are slow and deprecated. Please use c10::Dict<K, V> instead.")]]
IValue(std::unordered_map<Key, Value> v);
template <class T, enable_if_ivalue_constructible<T> = nullptr>
IValue(std::optional<T> v);
@ -1163,7 +1160,7 @@ struct TORCH_API IValue final {
// this value different (e.g. using NaN boxing), and this would make it more
// costly to determine the tag for all types vs just determining if something
// is a particular type. Instead we want clients to use the `isX` methods when
// possible. If for perf. reasons you really, absolutely, must have a jump
// possible. If for performance reasons you really, absolutely, must have a jump
// table, then we can revisit this.
enum class Tag : uint32_t {
#define DEFINE_TAG(x) x,

21
aten/src/ATen/core/ivalue_inl.h
View File

@ -863,6 +863,19 @@ struct C10_EXPORT ivalue::Future final : c10::intrusive_ptr_target {
Future& operator=(const Future&) = delete;
Future& operator=(Future&&) = delete;
// Destructor
// Explicitly destroy events under device guard, otherwise it can lead to
// extra context being created on device 0. Reason: python garbage collector
// calls this destructor, but python GC does not have a device context, so a
// "default" one (usually on device 0) could be created when we go down the
// line of event destroy.
~Future() override {
while (!events_.empty()) {
c10::OptionalDeviceGuard deviceGuard(events_.back().device());
events_.pop_back();
}
}
struct TORCH_API FutureError final : public std::exception {
explicit FutureError(std::string&& error_msg_)
: error_msg(std::move(error_msg_)) {}
@ -1758,8 +1771,8 @@ struct _fake_type {};
template <class Elem>
// TODO this is deprecated but we don't throw a warning because a lot of ops in
// native_functions.yaml still return std::vector.
// C10_DEPRECATED_MESSAGE("IValues based on std::vector<T> are potentially slow
// and deprecated. Please use torch::List<T> instead.")
// [[deprecated("IValues based on std::vector<T> are potentially slow
// and deprecated. Please use torch::List<T> instead.")]]
std::vector<Elem> generic_to(IValue ivalue, _fake_type<std::vector<Elem>>) {
// We need to do a deep copy of the vector because there might be other
// references to this same IValue that also use the list. We can't just
@ -1895,8 +1908,8 @@ c10::Dict<Key, Value> generic_to(
}
template <typename K, typename V>
C10_DEPRECATED_MESSAGE(
"IValues based on std::unordered_map are slow and deprecated. Please use c10::Dict<K, V> instead.")
[[deprecated(
"IValues based on std::unordered_map are slow and deprecated. Please use c10::Dict<K, V> instead.")]]
std::unordered_map<K, V> generic_to(
IValue ivalue,
_fake_type<std::unordered_map<K, V>>) {

2
aten/src/ATen/core/jit_type.h
View File

@ -938,7 +938,7 @@ struct TORCH_API DictType : public SharedType {
case TypeKind::DeviceObjType:
return DictTypePtr(new DictType(std::move(key), std::move(value)));
default:
AT_ERROR(
TORCH_CHECK(false,
"Cannot create dict for key type '",
key->str(),
"', only int, float, complex, Tensor, device and string keys are supported");

2
aten/src/ATen/core/jit_type_base.h
View File

@ -585,7 +585,7 @@ struct TORCH_API Type {
virtual TypePtr createWithContained(
// NOLINTNEXTLINE(performance-unnecessary-value-param)
std::vector<TypePtr> /*contained_types*/) const {
AT_ERROR(
TORCH_CHECK(false,
"type with contained types did not overload createWithContained: ",
str());
}

2
aten/src/ATen/core/op_registration/op_registration.h
View File

@ -562,7 +562,7 @@ public:
}
template<class Lambda>
C10_DEPRECATED_MESSAGE("Registering operator kernels with stateful lambdas (i.e. lambdas with a capture) has non-obvious behavior. This is deprecated. Please use a lambda without a capture or a functor class instead.")
[[deprecated("Registering operator kernels with stateful lambdas (i.e. lambdas with a capture) has non-obvious behavior. This is deprecated. Please use a lambda without a capture or a functor class instead.")]]
// enable_if: only enable it if Lambda is actually a functor but not a stateless lambda
std::enable_if_t<guts::is_functor<Lambda>::value && !guts::is_stateless_lambda<std::decay_t<Lambda>>::value, RegisterOperators&&>
op(const std::string& schemaOrName, Lambda&& lambda, Options&& options = RegisterOperators::options()) && {

2
aten/src/ATen/core/stack.h
View File

@ -21,7 +21,7 @@ class Operation {
public:
template <typename F,
std::enable_if_t<accepts<F, Stack*>::value, int> = 0>
C10_DEPRECATED_MESSAGE("Please use void(Stack&) to register operator instead.")
[[deprecated("Please use void(Stack&) to register operator instead.")]]
// NOLINTNEXTLINE(cppcoreguidelines-missing-std-forward)
Operation(F&& raw): op_([raw = std::forward<F>(raw)](Stack& stack) {
raw(&stack);

2
aten/src/ATen/core/type.cpp
View File

@ -629,7 +629,7 @@ MatchTypeReturn matchTypeVariables(
}
}
AT_ERROR("Unhandled free variable container: ", formal->repr_str());
TORCH_CHECK(false, "Unhandled free variable container: ", formal->repr_str());
}
// change return types like List[List[t]] into List[List[int]]

6
aten/src/ATen/cuda/CUDABlas.cpp
View File

@ -34,7 +34,7 @@ static rocblas_operation hipOperationToRocOperation(hipblasOperation_t op)
case HIPBLAS_OP_C:
return rocblas_operation_conjugate_transpose;
}
AT_ERROR("HIPBLAS_STATUS_INVALID_ENUM");
TORCH_CHECK(false, "HIPBLAS_STATUS_INVALID_ENUM");
}
static hipblasStatus_t rocBLASStatusToHIPStatus(rocblas_status error)
{
@ -57,7 +57,7 @@ static hipblasStatus_t rocBLASStatusToHIPStatus(rocblas_status error)
case rocblas_status_internal_error:
return HIPBLAS_STATUS_INTERNAL_ERROR;
}
AT_ERROR("HIPBLAS_STATUS_INVALID_ENUM");
TORCH_CHECK(false, "HIPBLAS_STATUS_INVALID_ENUM");
}
// hipblas does not have hipblasSetMathMode
#define hipblasSetMathMode(handle, flags) HIPBLAS_STATUS_SUCCESS
@ -116,7 +116,7 @@ static cublasOperation_t _cublasOpFromChar(char op) {
case 'C':
return CUBLAS_OP_C;
}
AT_ERROR(
TORCH_CHECK(false,
"_cublasOpFromChar input should be 't', 'n' or 'c' but got `", op, "`");
}

10
aten/src/ATen/cuda/Exceptions.h
View File

@ -165,9 +165,9 @@ constexpr const char* _cusolver_backend_suggestion = \
[[maybe_unused]] CUresult get_error_str_err = \
at::globalContext().getNVRTC().cuGetErrorString(__err, &err_str); \
if (get_error_str_err != CUDA_SUCCESS) { \
AT_ERROR("CUDA driver error: unknown error"); \
TORCH_CHECK(false, "CUDA driver error: unknown error"); \
} else { \
AT_ERROR("CUDA driver error: ", err_str); \
TORCH_CHECK(false, "CUDA driver error: ", err_str); \
} \
} \
} while (0)
@ -178,7 +178,7 @@ constexpr const char* _cusolver_backend_suggestion = \
do { \
CUresult __err = EXPR; \
if (__err != CUDA_SUCCESS) { \
AT_ERROR("CUDA driver error: ", static_cast<int>(__err)); \
TORCH_CHECK(false, "CUDA driver error: ", static_cast<int>(__err)); \
} \
} while (0)
@ -198,9 +198,9 @@ constexpr const char* _cusolver_backend_suggestion = \
nvrtcResult __err = EXPR; \
if (__err != NVRTC_SUCCESS) { \
if (static_cast<int>(__err) != 7) { \
AT_ERROR("CUDA NVRTC error: ", at::globalContext().getNVRTC().nvrtcGetErrorString(__err)); \
TORCH_CHECK(false, "CUDA NVRTC error: ", at::globalContext().getNVRTC().nvrtcGetErrorString(__err)); \
} else { \
AT_ERROR("CUDA NVRTC error: NVRTC_ERROR_BUILTIN_OPERATION_FAILURE"); \
TORCH_CHECK(false, "CUDA NVRTC error: NVRTC_ERROR_BUILTIN_OPERATION_FAILURE"); \
} \
} \
} while (0)

6
aten/src/ATen/cuda/detail/CUDAHooks.cpp
View File

@ -103,7 +103,7 @@ void CUDAHooks::init() const {
#endif
}
const Generator& CUDAHooks::getDefaultCUDAGenerator(DeviceIndex device_index) const {
const Generator& CUDAHooks::getDefaultGenerator(DeviceIndex device_index) const {
return at::cuda::detail::getDefaultCUDAGenerator(device_index);
}
@ -300,7 +300,7 @@ long CUDAHooks::versionCuDNN() const {
#if AT_CUDNN_ENABLED()
return CUDNN_VERSION;
#else
AT_ERROR("Cannot query CuDNN version if ATen_cuda is not built with CuDNN");
TORCH_CHECK(false, "Cannot query CuDNN version if ATen_cuda is not built with CuDNN");
#endif
}
@ -408,7 +408,7 @@ double CUDAHooks::batchnormMinEpsilonCuDNN() const {
#if AT_CUDNN_ENABLED()
return CUDNN_BN_MIN_EPSILON;
#else
AT_ERROR(
TORCH_CHECK(false,
"Cannot query CUDNN_BN_MIN_EPSILON if ATen_cuda is not built with CuDNN");
#endif
}

3
aten/src/ATen/cuda/detail/CUDAHooks.h
View File

@ -22,7 +22,8 @@ struct CUDAHooks : public at::CUDAHooksInterface {
void init() const override;
Device getDeviceFromPtr(void* data) const override;
bool isPinnedPtr(const void* data) const override;
const Generator& getDefaultCUDAGenerator(DeviceIndex device_index = -1) const override;
const Generator& getDefaultGenerator(
DeviceIndex device_index = -1) const override;
bool hasCUDA() const override;
bool hasMAGMA() const override;
bool hasCuDNN() const override;

4
aten/src/ATen/cuda/tunable/GemmHipblaslt.h
View File

@ -310,7 +310,7 @@ static hipblasOperation_t _hipblasOpFromChar(char op) {
case 'C':
return HIPBLAS_OP_C;
}
AT_ERROR(
TORCH_CHECK(false,
"_hipblasOpFromChar input should be 't', 'n' or 'c' but got `", op, "`");
}
@ -323,7 +323,7 @@ static char _charFromhipblasOp(hipblasOperation_t op) {
case HIPBLAS_OP_C:
return 'C';
}
AT_ERROR(
TORCH_CHECK(false,
"_charFromhipblasOp input should be HIPBLAS_OP_N/T/C but got `", op, "`");
}

2
aten/src/ATen/cuda/tunable/GemmRocblas.h
View File

@ -130,7 +130,7 @@ static rocblas_operation _rocblasOpFromChar(char op) {
case 'C':
return rocblas_operation_conjugate_transpose;
}
AT_ERROR(
TORCH_CHECK(false,
"_rocblasOpFromChar input should be 't', 'n' or 'c' but got `", op, "`");
}

20
aten/src/ATen/cuda/tunable/TunableGemm.h
View File

@ -197,15 +197,15 @@ class GemmTunableOp : public TunableOp<GemmParams<T>, StreamTimer> {
this->RegisterOp(std::string("Default"), std::make_unique<DefaultGemmOp<T>>());
#ifdef USE_ROCM
static const char *env_rocblas = std::getenv("PYTORCH_TUNABLEOP_ROCBLAS_ENABLED");
if (env_rocblas == nullptr || strcmp(env_rocblas, "1") == 0) {
static const auto env_rocblas = c10::utils::check_env("PYTORCH_TUNABLEOP_ROCBLAS_ENABLED");
if (!env_rocblas.has_value() || env_rocblas.value()) {
for (auto&& [name, op] : GetRocBlasGemmTypeStringAndOps<T>()) {
this->RegisterOp(std::move(name), std::move(op));
}
}
static const char *env_hipblaslt = std::getenv("PYTORCH_TUNABLEOP_HIPBLASLT_ENABLED");
if (env_hipblaslt == nullptr || strcmp(env_hipblaslt, "1") == 0) {
static const auto env_hipblaslt = c10::utils::check_env("PYTORCH_TUNABLEOP_HIPBLASLT_ENABLED");
if (!env_hipblaslt.has_value() || env_hipblaslt.value()) {
// disallow tuning of hipblaslt with c10::complex
if constexpr (
!std::is_same_v<T, c10::complex<float>> &&
@ -230,8 +230,8 @@ class GemmAndBiasTunableOp : public TunableOp<GemmAndBiasParams<T>, StreamTimer>
this->RegisterOp(std::string("Default"), std::make_unique<DefaultGemmAndBiasOp<T>>());
#ifdef USE_ROCM
static const char *env_hipblaslt = std::getenv("PYTORCH_TUNABLEOP_HIPBLASLT_ENABLED");
if (env_hipblaslt == nullptr || strcmp(env_hipblaslt, "1") == 0) {
static const auto env_hipblaslt = c10::utils::check_env("PYTORCH_TUNABLEOP_HIPBLASLT_ENABLED");
if (!env_hipblaslt.has_value() || env_hipblaslt.value()) {
// disallow tuning of hipblaslt with c10::complex
if constexpr (
!std::is_same_v<T, c10::complex<float>> &&
@ -256,15 +256,15 @@ class GemmStridedBatchedTunableOp : public TunableOp<GemmStridedBatchedParams<T>
this->RegisterOp(std::string("Default"), std::make_unique<DefaultGemmStridedBatchedOp<T>>());
#ifdef USE_ROCM
static const char *env_rocblas = std::getenv("PYTORCH_TUNABLEOP_ROCBLAS_ENABLED");
if (env_rocblas == nullptr || strcmp(env_rocblas, "1") == 0) {
static const auto env_rocblas = c10::utils::check_env("PYTORCH_TUNABLEOP_ROCBLAS_ENABLED");
if (!env_rocblas.has_value() || env_rocblas.value()) {
for (auto&& [name, op] : GetRocBlasGemmStridedBatchedTypeStringAndOps<T>()) {
this->RegisterOp(std::move(name), std::move(op));
}
}
static const char *env_hipblaslt = std::getenv("PYTORCH_TUNABLEOP_HIPBLASLT_ENABLED");
if (env_hipblaslt == nullptr || strcmp(env_hipblaslt, "1") == 0) {
static const auto env_hipblaslt = c10::utils::check_env("PYTORCH_TUNABLEOP_HIPBLASLT_ENABLED");
if (!env_hipblaslt.has_value() || env_hipblaslt.value()) {
// disallow tuning of hipblaslt with c10::complex
if constexpr (
!std::is_same_v<T, c10::complex<float>> &&

2
aten/src/ATen/cudnn/AutocastRNN.cpp
View File

@ -113,7 +113,7 @@ _cudnn_rnn_cast_reflatten(const Tensor & input,
batch_sizes,
dropout_state);
#else // AT_CUDNN_ENABLED()
AT_ERROR("autocast::_cudnn_rnn_cast_reflatten: ATen not compiled with cuDNN support");
TORCH_CHECK(false, "autocast::_cudnn_rnn_cast_reflatten: ATen not compiled with cuDNN support");
return {Tensor{}, Tensor{}, Tensor{}, Tensor{}, Tensor{}}; // never reached, placates the compiler
#endif // AT_CUDNN_ENABLED()
}

17
aten/src/ATen/detail/AcceleratorHooksInterface.h
View File

@ -1,9 +1,13 @@
#pragma once
#include <ATen/core/Generator.h>
#include <c10/core/Allocator.h>
#include <c10/core/Device.h>
#include <c10/core/Stream.h>
#include <c10/core/Allocator.h>
C10_DIAGNOSTIC_PUSH_AND_IGNORED_IF_DEFINED("-Wunused-parameter")
namespace at {
// AcceleratorHooksInterface is a shared interface provided by all
@ -58,7 +62,18 @@ struct TORCH_API AcceleratorHooksInterface {
virtual Device getDeviceFromPtr(void* data) const {
TORCH_CHECK(false, "Backend doesn't support getDeviceFromPtr()");
}
virtual const Generator& getDefaultGenerator(
C10_UNUSED DeviceIndex device_index = -1) const {
TORCH_CHECK(false, "Backend doesn`t support getDefaultGenerator()");
}
virtual Generator getNewGenerator(
C10_UNUSED DeviceIndex device_index = -1) const {
TORCH_CHECK(false, "Backend doesn`t support getNewGenerator()");
}
};
} // namespace at
C10_DIAGNOSTIC_POP()

13
aten/src/ATen/detail/CUDAHooksInterface.h
View File

@ -6,16 +6,13 @@
#include <ATen/detail/AcceleratorHooksInterface.h>
// Forward-declares at::Generator and at::cuda::NVRTC
// NB: Class must live in `at` due to limitations of Registry.h.
namespace at {
struct Generator;
// Forward-declares at::cuda::NVRTC
namespace cuda {
struct NVRTC;
} // namespace cuda
} // namespace at
// NB: Class must live in `at` due to limitations of Registry.h.
namespace at {
#ifdef _MSC_VER
constexpr const char* CUDA_HELP =
@ -69,8 +66,8 @@ struct TORCH_API CUDAHooksInterface : AcceleratorHooksInterface {
TORCH_CHECK(false, "Cannot initialize CUDA without ATen_cuda library. ", CUDA_HELP);
}
virtual const Generator& getDefaultCUDAGenerator(
[[maybe_unused]] DeviceIndex device_index = -1) const {
const Generator& getDefaultGenerator(
[[maybe_unused]] DeviceIndex device_index = -1) const override {
TORCH_CHECK(
false,
"Cannot get default CUDA generator without ATen_cuda library. ",

19
aten/src/ATen/detail/HIPHooksInterface.h
View File

@ -1,19 +1,13 @@
#pragma once
#include <c10/core/Allocator.h>
#include <c10/core/GeneratorImpl.h>
#include <c10/util/Exception.h>
#include <c10/util/Registry.h>
#include <ATen/detail/AcceleratorHooksInterface.h>
#include <memory>
namespace at {
class Context;
}
// NB: Class must live in `at` due to limitations of Registry.h.
namespace at {
@ -30,8 +24,9 @@ struct TORCH_API HIPHooksInterface : AcceleratorHooksInterface {
TORCH_CHECK(false, "Cannot initialize HIP without ATen_hip library.");
}
virtual std::unique_ptr<c10::GeneratorImpl> initHIPGenerator(Context*) const {
AT_ERROR("Cannot initialize HIP generator without ATen_hip library.");
const Generator& getDefaultGenerator(
C10_UNUSED DeviceIndex device_index = -1) const override {
TORCH_CHECK(false, "Cannot initialize HIP without ATen_hip library.");
}
virtual bool hasHIP() const {
@ -47,11 +42,7 @@ struct TORCH_API HIPHooksInterface : AcceleratorHooksInterface {
}
Allocator* getPinnedMemoryAllocator() const override {
AT_ERROR("Pinned memory requires HIP.");
}
virtual void registerHIPTypes(Context*) const {
AT_ERROR("Cannot registerHIPTypes() without ATen_hip library.");
TORCH_CHECK(false, "Pinned memory requires HIP.");
}
virtual int getNumGPUs() const {
@ -59,7 +50,7 @@ struct TORCH_API HIPHooksInterface : AcceleratorHooksInterface {
}
bool hasPrimaryContext(DeviceIndex device_index) const override {
AT_ERROR("Cannot check primary context without ATen_hip library.");
TORCH_CHECK(false, "Cannot check primary context without ATen_hip library.");
}
};

17
aten/src/ATen/detail/IPUHooksInterface.h
View File

@ -1,6 +1,5 @@
#pragma once
#include <ATen/core/Generator.h>
#include <ATen/detail/AcceleratorHooksInterface.h>
#include <c10/core/Allocator.h>
@ -9,7 +8,7 @@
namespace at {
struct TORCH_API IPUHooksInterface: AcceleratorHooksInterface {
struct TORCH_API IPUHooksInterface : AcceleratorHooksInterface {
~IPUHooksInterface() override = default;
void init() const override {
@ -21,16 +20,14 @@ struct TORCH_API IPUHooksInterface: AcceleratorHooksInterface {
return false;
}
virtual const Generator& getDefaultIPUGenerator(
DeviceIndex device_index [[maybe_unused]] = -1) const {
AT_ERROR(
"Cannot get the default IPU generator: the IPU backend is not "
"available.");
const Generator& getDefaultGenerator(
C10_UNUSED DeviceIndex device_index = -1) const override {
TORCH_CHECK(false, "Cannot initialize IPU without ATen_ipu library.");
}
virtual Generator newIPUGenerator(DeviceIndex device_index [[maybe_unused]] = -1) const {
AT_ERROR(
"Cannot create a new IPU generator: the IPU backend is not available.");
Generator getNewGenerator(
DeviceIndex device_index [[maybe_unused]] = -1) const override {
TORCH_CHECK(false, "Cannot initialize IPU without ATen_ipu library.");
}
};

7
aten/src/ATen/detail/MPSHooksInterface.h
View File

@ -2,9 +2,9 @@
#pragma once
#include <c10/core/Allocator.h>
#include <ATen/core/Generator.h>
#include <ATen/detail/AcceleratorHooksInterface.h>
#include <c10/core/Allocator.h>
#include <c10/util/Exception.h>
#include <c10/util/Registry.h>
@ -31,7 +31,8 @@ struct TORCH_API MPSHooksInterface : AcceleratorHooksInterface {
virtual bool isOnMacOSorNewer(unsigned major = 13, unsigned minor = 0) const {
FAIL_MPSHOOKS_FUNC(__func__);
}
virtual const Generator& getDefaultMPSGenerator() const {
const Generator& getDefaultGenerator(
C10_UNUSED DeviceIndex device_index = -1) const override {
FAIL_MPSHOOKS_FUNC(__func__);
}
virtual Allocator* getMPSDeviceAllocator() const {

14
aten/src/ATen/detail/PrivateUse1HooksInterface.h
View File

@ -1,18 +1,20 @@
#pragma once
#include <ATen/core/Generator.h>
#include <ATen/detail/AcceleratorHooksInterface.h>
#include <c10/core/Allocator.h>
#include <c10/core/Device.h>
#include <c10/core/Storage.h>
#include <c10/util/Exception.h>
C10_DIAGNOSTIC_PUSH_AND_IGNORED_IF_DEFINED("-Wunused-parameter")
namespace at {
struct TORCH_API PrivateUse1HooksInterface : AcceleratorHooksInterface {
~PrivateUse1HooksInterface() override = default;
virtual const at::Generator& getDefaultGenerator(
c10::DeviceIndex device_index) const {
const at::Generator& getDefaultGenerator(
c10::DeviceIndex device_index) const override {
TORCH_CHECK_NOT_IMPLEMENTED(
false,
"You should register `PrivateUse1HooksInterface` for PrivateUse1 before call `getDefaultGenerator`.");
@ -24,17 +26,17 @@ struct TORCH_API PrivateUse1HooksInterface : AcceleratorHooksInterface {
"You should register `PrivateUse1HooksInterface` for PrivateUse1 before call `getDeviceFromPtr`.");
}
virtual bool isPinnedPtr(const void* data) const override {
bool isPinnedPtr(const void* data) const override {
return false;
}
virtual Allocator* getPinnedMemoryAllocator() const override {
Allocator* getPinnedMemoryAllocator() const override {
TORCH_CHECK(
false,
"You should register `PrivateUse1HooksInterface` for PrivateUse1 before call `getPinnedMemoryAllocator`.");
}
virtual bool hasPrimaryContext(DeviceIndex device_index) const override {
bool hasPrimaryContext(DeviceIndex device_index) const override {
TORCH_CHECK_NOT_IMPLEMENTED(
false,
"You should register `PrivateUse1HooksInterface` for PrivateUse1 before call `hasPrimaryContext`.");

15
aten/src/ATen/detail/XPUHooksInterface.h
View File

@ -4,7 +4,6 @@
#include <c10/util/Exception.h>
#include <c10/util/Registry.h>
#include <ATen/core/Generator.h>
#include <ATen/detail/AcceleratorHooksInterface.h>
C10_DIAGNOSTIC_PUSH_AND_IGNORED_IF_DEFINED("-Wunused-parameter")
@ -32,17 +31,17 @@ struct TORCH_API XPUHooksInterface : AcceleratorHooksInterface{
TORCH_CHECK(false, "Cannot get XPU global device index without ATen_xpu library.");
}
virtual Generator getXPUGenerator(
[[maybe_unused]] DeviceIndex device_index = -1) const {
TORCH_CHECK(false, "Cannot get XPU generator without ATen_xpu library.");
}
virtual const Generator& getDefaultXPUGenerator(
[[maybe_unused]] DeviceIndex device_index = -1) const {
const Generator& getDefaultGenerator(
[[maybe_unused]] DeviceIndex device_index = -1) const override {
TORCH_CHECK(
false, "Cannot get default XPU generator without ATen_xpu library.");
}
Generator getNewGenerator(
[[maybe_unused]] DeviceIndex device_index = -1) const override {
TORCH_CHECK(false, "Cannot get XPU generator without ATen_xpu library.");
}
virtual DeviceIndex getNumGPUs() const {
return 0;
}

2
aten/src/ATen/dlpack.h
View File

@ -32,7 +32,9 @@
#define DLPACK_DLL
#endif
// NOLINTNEXTLINE(modernize-deprecated-headers)
#include <stdint.h>
// NOLINTNEXTLINE(modernize-deprecated-headers)
#include <stddef.h>
#ifdef __cplusplus

2
aten/src/ATen/functorch/BatchRulesModules.cpp
View File

@ -224,7 +224,7 @@ static Tensor one_hot_decomposition_hack(const Tensor &self, int64_t num_classes
// but shape inference is not possible.
if (self.sym_numel() == 0) {
if (num_classes <= 0) {
AT_ERROR("Can not infer total number of classes from empty tensor.");
TORCH_CHECK(false, "Can not infer total number of classes from empty tensor.");
} else {
shape.emplace_back(num_classes);
return at::empty_symint(shape, self.options());

14
aten/src/ATen/functorch/BatchRulesReduceOps.cpp
View File

@ -103,7 +103,7 @@ template<
// optional cannot be used in a template, otherwise we would use it here.
int maybe_keepdim_arg_pos
>
void boxed_reduction_batch_rule(const c10::OperatorHandle& op, torch::jit::Stack* stack) {
static void boxed_reduction_batch_rule(const c10::OperatorHandle& op, torch::jit::Stack* stack) {
const auto& schema = op.schema();
const auto num_returns = schema.returns().size();
const auto num_arguments = schema.arguments().size();
@ -357,21 +357,21 @@ static std::tuple<Tensor, std::optional<int64_t>> searchsorted_batch_rule(
// B<...>D, B<...>V -> no change
if (buckets_bdim.has_value() && self_bdim.has_value()) {
auto self_ = moveBatchDimToFront(self, self_bdim);
auto result = at::searchsorted(buckets, self_, out_int32, right, std::move(side), sorter_);
auto result = at::searchsorted(buckets, self_, out_int32, right, side, sorter_);
return std::make_tuple(std::move(result), 0);
}
// B<...>D, <...>V -> B<...>D, B<...>V
if (buckets_bdim.has_value() && !self_bdim.has_value()) {
auto self_ = moveBatchDimToFront(self, self_bdim);
self_ = ensure_has_bdim(self_, self_bdim.has_value(), buckets.size(0));
auto result = at::searchsorted(buckets, self_, out_int32, right, std::move(side), sorter_);
auto result = at::searchsorted(buckets, self_, out_int32, right, side, sorter_);
return std::make_tuple(std::move(result), 0);
}
// <...>D, B<...>V -> <...>D, <...>(BV)
if (!buckets_bdim.has_value() && self_bdim.has_value()) {
auto bdim_size = self.size(*self_bdim);
auto self_ = reshape_dim_into(*self_bdim, -1, self);
auto result = at::searchsorted(buckets, self_, out_int32, right, std::move(side), sorter_);
auto result = at::searchsorted(buckets, self_, out_int32, right, side, sorter_);
result = reshape_dim_outof(-1, bdim_size, result);
return std::make_tuple(result, result.dim() - 2);
}
@ -382,7 +382,7 @@ static std::tuple<Tensor, std::optional<int64_t>> searchsorted_batch_rule(
if (buckets_bdim.has_value() && self_bdim.has_value()) {
auto self_ = moveBatchDimToFront(self, self_bdim);
auto self_view_ = self_logical_rank == 0 ? self_.unsqueeze(-1) : self_.flatten(1);
auto result = at::searchsorted(buckets, self_view_, out_int32, right, std::move(side), sorter_);
auto result = at::searchsorted(buckets, self_view_, out_int32, right, side, sorter_);
result = self_logical_rank == 0 ? result.squeeze(-1) : result.view(self_.sizes());
return std::make_tuple(std::move(result), 0);
}
@ -391,13 +391,13 @@ static std::tuple<Tensor, std::optional<int64_t>> searchsorted_batch_rule(
auto bdim_size = buckets.size(*buckets_bdim);
auto self_ = ensure_has_bdim(self, false, bdim_size);
auto self_view_ = self_logical_rank == 0 ? self_.unsqueeze(-1) : self_.flatten(1);
auto result = at::searchsorted(buckets, self_view_, out_int32, right, std::move(side), sorter_);
auto result = at::searchsorted(buckets, self_view_, out_int32, right, side, sorter_);
result = self_logical_rank == 0 ? result.squeeze(-1) : result.view(self_.sizes());
return std::make_tuple(std::move(result), 0);
}
// D, B* -> no change
if (!buckets_bdim.has_value() && self_bdim.has_value()) {
auto result = at::searchsorted(buckets, self, out_int32, right, std::move(side), sorter_);
auto result = at::searchsorted(buckets, self, out_int32, right, side, sorter_);
return std::make_tuple(std::move(result), self_bdim);
}
TORCH_INTERNAL_ASSERT(false);

2
aten/src/ATen/metal/Context.cpp
View File

@ -16,7 +16,7 @@ at::Tensor& metal_copy_(at::Tensor& self, const at::Tensor& src) {
if (p) {
return p->metal_copy_(self, src);
}
AT_ERROR("Metal backend was not linked to the build");
TORCH_CHECK(false, "Metal backend was not linked to the build");
}
} // namespace at::metal

2
aten/src/ATen/miopen/AutocastRNN.cpp
View File

@ -46,7 +46,7 @@ miopen_rnn(const Tensor & input_r,
fn_dropout_state_opt);
#else
AT_ERROR("autocast::miopen_rnn: ATen not compiled with ROCm enabled");
TORCH_CHECK(false, "autocast::miopen_rnn: ATen not compiled with ROCm enabled");
return {Tensor{}, Tensor{}, Tensor{}, Tensor{}, Tensor{}}; // placate the compiler
#endif

3
aten/src/ATen/mps/MPSHooks.h
View File

@ -19,7 +19,8 @@ struct MPSHooks : public at::MPSHooksInterface {
bool isOnMacOSorNewer(unsigned major, unsigned minor) const override;
// MPSGeneratorImpl interface
const Generator& getDefaultMPSGenerator() const override;
const Generator& getDefaultGenerator(
DeviceIndex device_index = -1) const override;
// MPSStream interface
void deviceSynchronize() const override;

2
aten/src/ATen/mps/MPSHooks.mm
View File

@ -59,7 +59,7 @@ Allocator* MPSHooks::getMPSDeviceAllocator() const {
return at::mps::GetMPSAllocator();
}
const Generator& MPSHooks::getDefaultMPSGenerator() const {
const Generator& MPSHooks::getDefaultGenerator([[maybe_unused]] DeviceIndex device_index) const {
return at::mps::detail::getDefaultMPSGenerator();
}

14
aten/src/ATen/mps/MPSProfiler.mm
View File

@ -189,7 +189,7 @@ void MPSProfiler::initialize() {
currentSigint.sa_flags = SA_RESTART;
sigfillset(&currentSigint.sa_mask);
if (sigaction(SIGINT, &currentSigint, &previousSigint) == -1) {
AT_ERROR("Cannot install SIGINT handler for MPSProfiler.");
TORCH_CHECK(false, "Cannot install SIGINT handler for MPSProfiler.");
}
}
}
@ -207,7 +207,7 @@ void MPSProfiler::StartTrace(const std::string& mode, bool waitUntilCompleted) {
} else if (token == "event") {
m_profile_options |= ProfileOptions::ALL_SIGNPOST_EVENTS;
} else {
AT_ERROR("Invalid Signpost trace mode: ", token);
TORCH_CHECK(false, "Invalid Signpost trace mode: ", token);
}
}
}
@ -654,7 +654,7 @@ bool MPSProfiler::isProfileInfoLoggingEnabled(BaseInfo::Type infoType, bool isEx
isInfoLoggingEnabled = (m_log_options & LogOptions::CPU_FALLBACK_INFO);
break;
default:
AT_ERROR("invalid profiling info type");
TORCH_CHECK(false, "invalid profiling info type");
}
if (!isInfoLoggingEnabled) {
return false;
@ -685,7 +685,7 @@ void MPSProfiler::emitSignpostEvent(SignpostTypes signpost_type,
os_signpost_event_emit(m_os_log_events, signpost_id, kEvtSignpostCPUFallbacksStr, "%s", msg);
break;
default:
AT_ERROR("unknown SignpostType in MPS profiler");
TORCH_CHECK(false, "unknown SignpostType in MPS profiler");
}
}
@ -709,7 +709,7 @@ void MPSProfiler::beginSignpostInterval(SignpostTypes signpost_type,
os_signpost_interval_begin(m_os_log_intervals, signpost_id, kIntSignpostCPUFallbacksStr, "%s", msg);
break;
default:
AT_ERROR("unknown SignpostType in MPS profiler");
TORCH_CHECK(false, "unknown SignpostType in MPS profiler");
}
}
@ -728,7 +728,7 @@ void MPSProfiler::endSignpostInterval(SignpostTypes signpost_type, os_signpost_i
os_signpost_interval_end(m_os_log_intervals, signpost_id, kIntSignpostCPUFallbacksStr);
break;
default:
AT_ERROR("unknown SignpostType in MPS profiler");
TORCH_CHECK(false, "unknown SignpostType in MPS profiler");
}
}
@ -750,7 +750,7 @@ MPSProfiler::SignpostTypes MPSProfiler::getSignpostType(BaseInfo::Type infoType)
case BaseInfo::Type::CPU_FALLBACK:
return SignpostTypes::CPU_FALLBACK;
default:
AT_ERROR("invalid profiling info type");
TORCH_CHECK(false, "invalid profiling info type");
}
}

2
aten/src/ATen/native/BatchLinearAlgebra.cpp
View File

@ -1624,7 +1624,7 @@ Tensor inverse(const Tensor& A) {
template<typename scalar_t>
static void apply_cholesky_solve(Tensor& b, Tensor& A, bool upper, Tensor& infos) {
#if !AT_BUILD_WITH_LAPACK()
AT_ERROR("cholesky_solve: LAPACK library not found in compilation");
TORCH_CHECK(false, "cholesky_solve: LAPACK library not found in compilation");
#else
char uplo = upper ? 'U' : 'L';

2
aten/src/ATen/native/ConvUtils.h
View File

@ -168,7 +168,7 @@ static void check_args(CheckedFrom c, IntArrayRef args, size_t expected_size, co
ss << arg_name << " should be greater than zero but got (";
std::copy(args.begin(), args.end() - 1, std::ostream_iterator<int>(ss,", "));
ss << args.back() << ")" << " (while checking arguments for " << c << ")";
AT_ERROR(ss.str());
TORCH_CHECK(false, ss.str());
}
}

4
aten/src/ATen/native/Convolution.cpp
View File

@ -719,7 +719,7 @@ static void check_shape_forward(const at::Tensor& input,
separator = " x ";
}
AT_ERROR("Calculated padded input size per channel: (", input_ss.str(), "). "
TORCH_CHECK(false, "Calculated padded input size per channel: (", input_ss.str(), "). "
"Kernel size: (", kernel_ss.str(), "). Kernel size can't be greater than actual input size");
}
} else { // transposed
@ -1304,7 +1304,7 @@ ConvBackend _select_conv_backend(
}
// Error out if no suitable backend was found.
AT_ERROR("unsupported ConvNd parameters");
TORCH_CHECK(false, "unsupported ConvNd parameters");
}
// Selects a backend for convolution based on the inputs and params.

2
aten/src/ATen/native/DispatchStub.cpp
View File

@ -262,7 +262,7 @@ void* DispatchStubImpl::get_call_ptr(
false, "DispatchStub: missing kernel for ", device_type);
return nullptr;
case ErrorType::DeviceNotSupported:
AT_ERROR("DispatchStub: unsupported device type", device_type);
TORCH_CHECK(false, "DispatchStub: unsupported device type", device_type);
}
}

2
aten/src/ATen/native/Embedding.cpp
View File

@ -81,7 +81,7 @@ Tensor embedding_sparse_backward(
// TODO: implement scale_grad_by_freq
if (scale_grad_by_freq) {
AT_ERROR(
TORCH_CHECK(false,
"embedding_backward: scale_grad_by_freq not supported with sparse gradients");
}

2
aten/src/ATen/native/Fill.cpp
View File

@ -104,7 +104,7 @@ Tensor& fill_diagonal_(Tensor& self, const Scalar& fill_value, bool wrap) {
int64_t dim1 = height;
for (const auto i : c10::irange(1, nDims)) {
if (self.size(i) != dim1) {
AT_ERROR("all dimensions of input must be of equal length");
TORCH_CHECK(false, "all dimensions of input must be of equal length");
}
}
}

2
aten/src/ATen/native/LinearAlgebraUtils.h
View File

@ -269,7 +269,7 @@ inline double _get_epsilon(const ScalarType& sc_type) {
case at::ScalarType::Double:
return std::numeric_limits<double>::epsilon();
default:
AT_ERROR("This function doesn't handle types other than float and double");
TORCH_CHECK(false, "This function doesn't handle types other than float and double");
}
}

2
aten/src/ATen/native/MaxUnpooling.cpp
View File

@ -136,7 +136,7 @@ static void max_unpooling3d_shape_check(
if (gradOutput.defined()) {
if (oT != gradOutput.size(dimt) || oH != gradOutput.size(dimh) ||
oW != gradOutput.size(dimw)) {
AT_ERROR(
TORCH_CHECK(false,
"Inconsistent gradOutput size. oT= ",
oT,
", oH= ",

4
aten/src/ATen/native/NaiveConvolutionTranspose2d.cpp
View File

@ -85,7 +85,7 @@ static inline void slow_conv_transpose2d_shape_check(
check_dim_size(bias, 1, 0, weight.size(1));
}
} else if (!weight_nullable) {
AT_ERROR("weight tensor is expected to be non-nullable");
TORCH_CHECK(false, "weight tensor is expected to be non-nullable");
}
int ndim = input.dim();
@ -112,7 +112,7 @@ static inline void slow_conv_transpose2d_shape_check(
(dilation_width * (kernel_width - 1) + 1) + output_padding_width;
if (output_width < 1 || output_height < 1) {
AT_ERROR(
TORCH_CHECK(false,
"Given input size per channel: (",
input_height,
" x ",

4
aten/src/ATen/native/NaiveConvolutionTranspose3d.cpp
View File

@ -107,7 +107,7 @@ static inline void slow_conv_transpose3d_shape_check(
check_dim_size(bias, 1, 0, weight.size(1));
}
} else if (!weight_nullable) {
AT_ERROR("weight tensor is expected to be non-nullable");
TORCH_CHECK(false, "weight tensor is expected to be non-nullable");
}
int ndim = input.dim();
@ -142,7 +142,7 @@ static inline void slow_conv_transpose3d_shape_check(
output_padding_width;
if (output_depth < 1 || output_width < 1 || output_height < 1) {
AT_ERROR(
TORCH_CHECK(false,
"Given input size per channel: (",
input_depth,
" x ",

4
aten/src/ATen/native/Normalization.cpp
View File

@ -573,12 +573,12 @@ std::tuple<Tensor, Tensor, Tensor, Tensor, int64_t> _batch_norm_impl_index(
if (running_mean.defined()) {
check_dims_match_num_input_features("running_mean", num_features, running_mean.sym_numel());
} else if (!training) {
AT_ERROR("running_mean must be defined in evaluation mode");
TORCH_CHECK(false, "running_mean must be defined in evaluation mode");
}
if (running_var.defined()) {
check_dims_match_num_input_features("running_var", num_features, running_var.sym_numel());
} else if (!training) {
AT_ERROR("running_var must be defined in evaluation mode");
TORCH_CHECK(false, "running_var must be defined in evaluation mode");
}
if (weight.defined()) {
check_dims_match_num_input_features("weight", num_features, weight.sym_numel());

2
aten/src/ATen/native/Onehot.cpp
View File

@ -34,7 +34,7 @@ Tensor one_hot(const Tensor &self, int64_t num_classes) {
// but shape inference is not possible.
if (self.numel() == 0) {
if (num_classes <= 0) {
AT_ERROR("Can not infer total number of classes from empty tensor.");
TORCH_CHECK(false, "Can not infer total number of classes from empty tensor.");
} else {
shape.push_back(num_classes);
return at::empty(shape, self.options());

2
aten/src/ATen/native/PackedSequence.cpp
View File

@ -51,7 +51,7 @@ std::tuple<Tensor, Tensor> _pack_padded_sequence(const Tensor& _input, const Ten
// NB: enforce_sorted is implemented at a Python level, but the sortedness
// check lives here. If enforce_sorted=False then this error should never
// get called.
AT_ERROR("`lengths` array must be sorted in decreasing order when "
TORCH_CHECK(false, "`lengths` array must be sorted in decreasing order when "
"`enforce_sorted` is True. You can pass `enforce_sorted=False` "
"to pack_padded_sequence and/or pack_sequence to sidestep this "
"requirement if you do not need ONNX exportability.");

2
aten/src/ATen/native/Repeat.cpp
View File

@ -83,7 +83,7 @@ Tensor repeat_interleave_symint(
repeats.sym_size(0), " and input.size(", dim.value(), ") = ", input.sym_size(dim.value())
);
} else {
AT_ERROR("repeats must be 0-dim or 1-dim tensor");
TORCH_CHECK(false, "repeats must be 0-dim or 1-dim tensor");
}
auto ret = input.index_select(

30
aten/src/ATen/native/SpectralOps.cpp
View File

@ -881,12 +881,12 @@ Tensor stft(const Tensor& self, const int64_t n_fft, const std::optional<int64_t
if (!at::isFloatingType(self.scalar_type()) && !at::isComplexType(self.scalar_type())) {
std::ostringstream ss;
REPR(ss) << ": expected a tensor of floating point or complex values";
AT_ERROR(ss.str());
TORCH_CHECK(false, ss.str());
}
if (self.dim() > 2 || self.dim() < 1) {
std::ostringstream ss;
REPR(ss) << ": expected a 1D or 2D tensor";
AT_ERROR(ss.str());
TORCH_CHECK(false, ss.str());
}
Tensor input = self;
if (self.dim() == 1) {
@ -911,24 +911,24 @@ Tensor stft(const Tensor& self, const int64_t n_fft, const std::optional<int64_t
std::ostringstream ss;
REPR(ss) << ": expected 0 < n_fft < " << len
<< ", but got n_fft=" << win_length;
AT_ERROR(ss.str());
TORCH_CHECK(false, ss.str());
}
if (hop_length <= 0) {
std::ostringstream ss;
REPR(ss) << ": expected hop_length > 0, but got hop_length=" << hop_length;
AT_ERROR(ss.str());
TORCH_CHECK(false, ss.str());
}
if (win_length <= 0 || win_length > n_fft) {
std::ostringstream ss;
REPR(ss) << ": expected 0 < win_length <= n_fft, but got win_length="
<< win_length;
AT_ERROR(ss.str());
TORCH_CHECK(false, ss.str());
}
if (window.defined() && (window.dim() != 1 || window.size(0) != win_length)) {
std::ostringstream ss;
REPR(ss) << ": expected a 1D window tensor of size equal to win_length="
<< win_length << ", but got window with size " << window.sizes();
AT_ERROR(ss.str());
TORCH_CHECK(false, ss.str());
}
#undef REPR
auto window_ = window;
@ -1063,17 +1063,17 @@ Tensor istft(const Tensor& self, const int64_t n_fft, const std::optional<int64_
if (input.numel() == 0) {
std::ostringstream ss;
REPR(ss) << ": input tensor cannot be empty.";
AT_ERROR(ss.str());
TORCH_CHECK(false, ss.str());
}
if (input_dim != 3 && input_dim != 4) {
std::ostringstream ss;
REPR(ss) << ": expected a tensor with 3 or 4 dimensions, but got " << input_dim;
AT_ERROR(ss.str());
TORCH_CHECK(false, ss.str());
}
if (input.size(-1) != 2) {
std::ostringstream ss;
REPR(ss) << ": expected the last dimension to be 2 (corresponding to real and imaginary parts), but got " << self.size(-1);
AT_ERROR(ss.str());
TORCH_CHECK(false, ss.str());
}
const bool onesided = onesidedOpt.value_or(fft_size != n_fft);
@ -1081,32 +1081,32 @@ Tensor istft(const Tensor& self, const int64_t n_fft, const std::optional<int64_
if (n_fft / 2 + 1 != fft_size) {
std::ostringstream ss;
REPR(ss) << ": expected the frequency dimension (3rd to the last) of the input tensor to match n_fft / 2 + 1 when onesided=True, but got " << fft_size;
AT_ERROR(ss.str());
TORCH_CHECK(false, ss.str());
}
} else {
if (n_fft != fft_size) {
std::ostringstream ss;
REPR(ss) << ": expected the frequency dimension (3rd to the last) of the input tensor to match n_fft when onesided=False, but got " << fft_size;
AT_ERROR(ss.str());
TORCH_CHECK(false, ss.str());
}
}
if (!(0 < hop_length && hop_length <= win_length)) {
std::ostringstream ss;
REPR(ss) << ": expected 0 < hop_length <= win_length";
AT_ERROR(ss.str());
TORCH_CHECK(false, ss.str());
}
if (!(0 < win_length && win_length <= n_fft)) {
std::ostringstream ss;
REPR(ss) << ": expected 0 < win_length <= n_fft";
AT_ERROR(ss.str());
TORCH_CHECK(false, ss.str());
}
if (window.defined()) {
if (window.dim() != 1 || window.size(0) != win_length) {
std::ostringstream ss;
REPR(ss) << ": Invalid window shape. window has to be 1D and length of `win_length`";
AT_ERROR(ss.str());
TORCH_CHECK(false, ss.str());
}
}
@ -1175,7 +1175,7 @@ Tensor istft(const Tensor& self, const int64_t n_fft, const std::optional<int64_
if (at::is_scalar_tensor_true(window_envelop_lowest)) {
std::ostringstream ss;
REPR(ss) << "window overlap add min: " << window_envelop_lowest;
AT_ERROR(ss.str());
TORCH_CHECK(false, ss.str());
}
y = (y / window_envelop); // size: (channel, expected_output_signal_len)

2
aten/src/ATen/native/SpectralOpsUtils.h
View File

@ -63,7 +63,7 @@ inline int64_t infer_ft_complex_to_real_onesided_size(int64_t complex_size,
std::ostringstream ss;
ss << "expected real signal size " << expected_size << " is incompatible "
<< "with onesided complex frequency size " << complex_size;
AT_ERROR(ss.str());
TORCH_CHECK(false, ss.str());
}
}

8
aten/src/ATen/native/SummaryOps.cpp
View File

@ -26,19 +26,19 @@ Tensor _bincount_cpu_template(
const Tensor& weights,
int64_t minlength) {
if (minlength < 0) {
AT_ERROR("minlength should be >= 0");
TORCH_CHECK(false, "minlength should be >= 0");
}
if (self.dim() == 1 && self.numel() == 0) {
return at::zeros({minlength}, kLong);
}
if (self.dim() != 1 || *self.min().data_ptr<input_t>() < 0) {
AT_ERROR("bincount only supports 1-d non-negative integral inputs.");
TORCH_CHECK(false, "bincount only supports 1-d non-negative integral inputs.");
}
// Ensure max_val < 2 ^ 63 - 1 (9223372036854775807)
auto max_val = *self.max().data_ptr<input_t>();
if (max_val >= std::numeric_limits<int64_t>::max()) {
AT_ERROR(
TORCH_CHECK(false,
"maximum value of input overflowed, it should be < ",
std::numeric_limits<int64_t>::max(),
" but got ",
@ -48,7 +48,7 @@ Tensor _bincount_cpu_template(
bool has_weights = weights.defined();
if (has_weights && (weights.dim() != 1 || weights.size(0) != self.size(0))) {
AT_ERROR("weights should be 1-d and have the same length as input");
TORCH_CHECK(false, "weights should be 1-d and have the same length as input");
}
Tensor output;

58
aten/src/ATen/native/TensorConversions.cpp
View File

@ -588,7 +588,7 @@ Tensor to_dense_backward(const Tensor& grad, const Tensor& input_, std::optional
case kMkldnn:
return grad.to_mkldnn(input_.scalar_type());
default:
AT_ERROR("to_dense_backward: Unsupported input layout: ", input_layout);
TORCH_CHECK(false, "to_dense_backward: Unsupported input layout: ", input_layout);
return Tensor{};
}
}
@ -928,23 +928,23 @@ void _to_sparse_check_arguments(const std::string& funcname, const Tensor& self,
auto layout_from_valid = layout_from == kStrided || layout_from == kSparse || at::sparse_csr::is_sparse_compressed(layout_from);
if (!layout_from_valid) {
AT_ERROR(funcname, ": unexpected source layout ", layout_from);
TORCH_CHECK(false, funcname, ": unexpected source layout ", layout_from);
}
if (layout_from == kStrided) {
if (sparse_dim == 0 && self.dim() > 0) {
AT_ERROR(funcname, ": sparse_dim argument must be in >0 when self.dim()>0");
TORCH_CHECK(false, funcname, ": sparse_dim argument must be in >0 when self.dim()>0");
}
if (sparse_dim < 0 || sparse_dim > self.dim()) {
AT_ERROR(funcname, ": sparse_dim argument must be in [0,", self.dim(), "] range, but ", sparse_dim, " is given");
TORCH_CHECK(false, funcname, ": sparse_dim argument must be in [0,", self.dim(), "] range, but ", sparse_dim, " is given");
}
} else if (layout_from == kSparse) {
if (sparse_dim != self.sparse_dim()) {
AT_ERROR(funcname, ": conversion from ", layout_from, " to ", kSparse, " with sparse_dim argument !=self.sparse_dim() is not supported");
TORCH_CHECK(false, funcname, ": conversion from ", layout_from, " to ", kSparse, " with sparse_dim argument !=self.sparse_dim() is not supported");
}
} else if (at::sparse_csr::is_sparse_compressed(layout_from)) {
if (sparse_dim != 2) {
AT_ERROR(funcname, ": conversion from ", layout_from, " to ", kSparse, " with sparse_dim argument !=2 is not supported");
TORCH_CHECK(false, funcname, ": conversion from ", layout_from, " to ", kSparse, " with sparse_dim argument !=2 is not supported");
}
}
}
@ -956,40 +956,40 @@ void _to_sparse_check_arguments(const std::string& funcname, const Tensor& self,
auto layout_from_valid = layout_from == kStrided || layout_from == kSparse || at::sparse_csr::is_sparse_compressed(layout_from);
if (!layout_from_valid) {
AT_ERROR(funcname, ": unexpected source layout ", layout_from);
TORCH_CHECK(false, funcname, ": unexpected source layout ", layout_from);
}
auto layout_to_valid = layout_to == kStrided || layout_to == kSparse || at::sparse_csr::is_sparse_compressed(layout_to);
if (!layout_to_valid) {
AT_ERROR(funcname, ": unexpected source layout ", layout_from);
TORCH_CHECK(false, funcname, ": unexpected source layout ", layout_from);
}
if (layout_from == kSparse && layout_to != kSparse) {
if (self.sparse_dim() != 2) {
AT_ERROR(funcname, ": conversion from ", layout_from, " to ", layout_to, " for input tensors with sparse_dim()!=2 is not supported");
TORCH_CHECK(false, funcname, ": conversion from ", layout_from, " to ", layout_to, " for input tensors with sparse_dim()!=2 is not supported");
}
}
if ((layout_from == kSparseCsr || layout_from == kSparseCsc) &&
(layout_to == kSparseBsr || layout_to == kSparseBsc)) {
if (sparse_csr::numBatchDimensions(self) > 0) {
AT_ERROR(funcname, ": conversion from ", layout_from, " to ", layout_to, " for batched inputs is not supported");
TORCH_CHECK(false, funcname, ": conversion from ", layout_from, " to ", layout_to, " for batched inputs is not supported");
}
}
if (blocksize.has_value()) {
if (blocksize.value().size() != 2) {
AT_ERROR(funcname, ": blocksize needs to be a tuple of size 2, but got ", blocksize.value().size());
TORCH_CHECK(false, funcname, ": blocksize needs to be a tuple of size 2, but got ", blocksize.value().size());
}
auto blocksize_to = *blocksize;
if (blocksize_to[0] <= 0 || blocksize_to[1] <= 0) {
AT_ERROR(funcname, ": blocksize needs to be positive, but got ", blocksize_to);
TORCH_CHECK(false, funcname, ": blocksize needs to be positive, but got ", blocksize_to);
}
if (layout_to == kSparseBsr || layout_to == kSparseBsc) {
if (layout_from == kSparseBsr || layout_from == kSparseBsc) {
auto blocksize_from = at::sparse_csr::getBlockSize(self);
if (!(blocksize_to == blocksize_from)) {
AT_ERROR(funcname, ": conversion from ", layout_from, " to ", layout_to, " with blocksize changed from ", blocksize_from, " to ", blocksize_to, " is not supported");
TORCH_CHECK(false, funcname, ": conversion from ", layout_from, " to ", layout_to, " with blocksize changed from ", blocksize_from, " to ", blocksize_to, " is not supported");
}
} else {
auto dense_dim = (layout_from == kStrided) ? dense_dim_opt.value_or(0) : self.dense_dim();
@ -997,35 +997,35 @@ void _to_sparse_check_arguments(const std::string& funcname, const Tensor& self,
auto sparse_col_dim = -(dense_dim + 1);
if ((self.size(sparse_row_dim) % blocksize_to[0] != 0) ||
(self.size(sparse_col_dim) % blocksize_to[1] != 0)) {
AT_ERROR(funcname, ": tensor sparse size (", self.size(sparse_row_dim), ",", self.size(sparse_row_dim), ") must be divisible by given blocksize (", blocksize_to[0], ",", blocksize_to[1], ")");
TORCH_CHECK(false, funcname, ": tensor sparse size (", self.size(sparse_row_dim), ",", self.size(sparse_row_dim), ") must be divisible by given blocksize (", blocksize_to[0], ",", blocksize_to[1], ")");
}
}
} else {
AT_ERROR(funcname, ": conversion from ", layout_from, " to ", layout_to, " with blocksize argument given is not supported");
TORCH_CHECK(false, funcname, ": conversion from ", layout_from, " to ", layout_to, " with blocksize argument given is not supported");
}
} else {
if ((layout_to == kSparseBsr || layout_to == kSparseBsc) &&
!(layout_from == kSparseBsr && layout_from == kSparseBsc)) {
AT_ERROR(funcname, ": conversion from ", layout_from, " to ", layout_to, " without blocksize argument given is not supported");
TORCH_CHECK(false, funcname, ": conversion from ", layout_from, " to ", layout_to, " without blocksize argument given is not supported");
}
}
if (dense_dim_opt.has_value()) {
if (layout_from != kStrided) {
AT_ERROR(funcname, ": conversion from ", layout_from, " to ", layout_to, " with dense_dim argument given is not supported");
TORCH_CHECK(false, funcname, ": conversion from ", layout_from, " to ", layout_to, " with dense_dim argument given is not supported");
}
auto dense_dim = *dense_dim_opt;
if (layout_to == kSparse) {
if (dense_dim == self.dim() && self.dim() > 0) {
AT_ERROR(funcname, ": dense_dim argument must be !=self.dim() when self.dim()>0");
TORCH_CHECK(false, funcname, ": dense_dim argument must be !=self.dim() when self.dim()>0");
}
if (dense_dim < 0 || dense_dim > self.dim()) {
AT_ERROR(funcname, ": dense_dim argument must be in [0,", self.dim(), "] range, but ", dense_dim, " is given");
TORCH_CHECK(false, funcname, ": dense_dim argument must be in [0,", self.dim(), "] range, but ", dense_dim, " is given");
}
} else {
if (dense_dim < 0 || dense_dim > self.dim() - 2) {
AT_ERROR(funcname, ": dense_dim argument must be in [0,", self.dim() - 2, "] range, but ", dense_dim, " is given");
TORCH_CHECK(false, funcname, ": dense_dim argument must be in [0,", self.dim() - 2, "] range, but ", dense_dim, " is given");
}
}
}
@ -1129,7 +1129,7 @@ Tensor dense_to_sparse_with_mask(const Tensor& self, const Tensor& mask, std::op
break;
}
AT_ERROR("dense_to_sparse_with_mask: ", self.layout(), " to ", layout_to, " conversion not supported");
TORCH_CHECK(false, "dense_to_sparse_with_mask: ", self.layout(), " to ", layout_to, " conversion not supported");
return Tensor{};
}
@ -1181,7 +1181,7 @@ Tensor dense_to_sparse(const Tensor& self, std::optional<c10::Layout> layout, Op
break;
}
AT_ERROR("dense_to_sparse: ", self.layout(), " to ", layout_to, " conversion not supported");
TORCH_CHECK(false, "dense_to_sparse: ", self.layout(), " to ", layout_to, " conversion not supported");
return Tensor{};
}
@ -1440,7 +1440,7 @@ Tensor sparse_compressed_to_sparse_csr(const Tensor& self, std::optional<int64_t
return sparse_compressed_to_flipped(self, std::nullopt, "to_sparse_csr");
}
AT_ERROR("sparse_compressed_to_sparse_csr: expected SparseCsr or SparseCsc layout but got ", self.layout());
TORCH_CHECK(false, "sparse_compressed_to_sparse_csr: expected SparseCsr or SparseCsc layout but got ", self.layout());
return Tensor{};
}
@ -1453,7 +1453,7 @@ Tensor sparse_compressed_to_sparse_csc(const Tensor& self, std::optional<int64_t
return sparse_compressed_to_flipped(self, std::nullopt, "to_sparse_csc");
}
AT_ERROR("sparse_compressed_to_sparse_csc: expected SparseCsr or SparseCsc layout but got ", self.layout());
TORCH_CHECK(false, "sparse_compressed_to_sparse_csc: expected SparseCsr or SparseCsc layout but got ", self.layout());
return Tensor{};
}
@ -1828,7 +1828,7 @@ Tensor sparse_compressed_to_sparse_bsr(const Tensor& self, IntArrayRef blocksize
return self.to_sparse_csr(dense_dim_opt).to_sparse_bsr(blocksize);
}
AT_ERROR("sparse_compressed_to_sparse_bsr: expected SparseCsr, SparseCsc, SparseBsr or SparseBsc layout but got ", self.layout());
TORCH_CHECK(false, "sparse_compressed_to_sparse_bsr: expected SparseCsr, SparseCsc, SparseBsr or SparseBsc layout but got ", self.layout());
return Tensor{};
}
@ -1850,14 +1850,14 @@ Tensor sparse_compressed_to_sparse_bsc(const Tensor& self, IntArrayRef blocksize
return self.to_sparse_csc(dense_dim_opt).to_sparse_bsc(blocksize);
}
AT_ERROR("sparse_compressed_to_sparse_bsc: expected SparseCsr, SparseCsc, SparseBsr or SparseBsc layout but got ", self.layout());
TORCH_CHECK(false, "sparse_compressed_to_sparse_bsc: expected SparseCsr, SparseCsc, SparseBsr or SparseBsc layout but got ", self.layout());
return Tensor{};
}
Tensor sparse_coo_to_sparse(const Tensor& self, const int64_t sparse_dim) {
_to_sparse_check_arguments("sparse_coo_to_sparse", self, sparse_dim);
AT_ERROR("sparse_coo_to_sparse: ", self.layout(), " to ", kSparse, " conversion not supported");
TORCH_CHECK(false, "sparse_coo_to_sparse: ", self.layout(), " to ", kSparse, " conversion not supported");
return Tensor{};
}
@ -1927,7 +1927,7 @@ Tensor sparse_compressed_to_sparse(const Tensor& self, std::optional<c10::Layout
break;
}
AT_ERROR("sparse_compressed_to_sparse: ", self.layout(), " to ", layout_to, " conversion not supported");
TORCH_CHECK(false, "sparse_compressed_to_sparse: ", self.layout(), " to ", layout_to, " conversion not supported");
return Tensor{};
}
@ -1951,7 +1951,7 @@ Tensor sparse_coo_to_sparse(const Tensor& self, std::optional<c10::Layout> layou
break;
}
AT_ERROR("sparse_coo_to_sparse: ", self.layout(), " to ", layout_to, " conversion not supported");
TORCH_CHECK(false, "sparse_coo_to_sparse: ", self.layout(), " to ", layout_to, " conversion not supported");
return Tensor{};
}

2
aten/src/ATen/native/TensorProperties.cpp
View File

@ -101,7 +101,7 @@ bool cudnn_is_acceptable(const Tensor& self) {
Tensor & detach_(Tensor & self) {
// this just exists to give us a hook in VariableType and an entry in Declarations.yaml
//AT_ERROR("detach_ is not implemented for Tensor");
//TORCH_CHECK(false, "detach_ is not implemented for Tensor");
return self;
}

10
aten/src/ATen/native/cpu/MaxUnpoolKernel.cpp
View File

@ -83,11 +83,11 @@ void cpu_max_unpool(
if (optional_error_index) {
if constexpr (is_3d) {
AT_ERROR("Found an invalid max index: ", optional_error_index.value(),
TORCH_CHECK(false, "Found an invalid max index: ", optional_error_index.value(),
" (output volumes are of size ", output_depth,
"x", output_height, "x", output_width);
} else {
AT_ERROR("Found an invalid max index: ", optional_error_index.value(),
TORCH_CHECK(false, "Found an invalid max index: ", optional_error_index.value(),
" (output volumes are of size ", output_height,
"x", output_width);
}
@ -151,7 +151,7 @@ void cpu_max_unpool_channels_last(
});
if (optional_error_index) {
AT_ERROR("Found an invalid max index: ", optional_error_index.value(),
TORCH_CHECK(false, "Found an invalid max index: ", optional_error_index.value(),
" (output volumes are of size ", output_height,
"x", output_width, ")");
}
@ -223,12 +223,12 @@ void cpu_max_unpool_backward(
if (optional_error_index) {
if (is_3d) {
AT_ERROR("invalid max index ", optional_error_index.value(),
TORCH_CHECK(false, "invalid max index ", optional_error_index.value(),
", odepth= ", output_depth,
", owidth= ", output_width,
", oheight= ", output_height);
} else {
AT_ERROR("invalid max index ", optional_error_index.value(),
TORCH_CHECK(false, "invalid max index ", optional_error_index.value(),
", owidth= ", output_width,
", oheight= ", output_height);
}

17
aten/src/ATen/native/cuda/Blas.cpp
View File

@ -180,19 +180,10 @@ cuda::blas::GEMMAndBiasActivationEpilogue activation_to_gemm_and_blas_arg(Activa
static bool getDisableAddmmCudaLt() {
static const char* env_value = std::getenv("DISABLE_ADDMM_CUDA_LT");
// When DISABLE_ADDMM_CUDA_LT is unset the default is TRUE on
// AMD architectures otherwise it is FALSE.
#ifdef USE_ROCM
if (env_value != nullptr && strcmp(env_value, "0") == 0) {
return false;
}
return true;
#else
if (env_value != nullptr && strcmp(env_value, "1") == 0) {
return true;
}
return false;
#endif
}
#ifdef USE_ROCM
@ -325,14 +316,6 @@ Tensor& addmm_out_cuda_impl(Tensor& result, const Tensor& self, const Tensor& ma
}
self__sizes = self_->sizes();
} else {
#if defined(USE_ROCM)
useLtInterface = !disable_addmm_cuda_lt &&
result.dim() == 2 && result.is_contiguous() &&
isSupportedHipLtROCmArch(self.device().index()) &&
(scalar_type == at::ScalarType::Float ||
scalar_type == at::ScalarType::Half ||
scalar_type == at::ScalarType::BFloat16);
#endif
self_ = c10::MaybeOwned<Tensor>::borrowed(self);
self__sizes = self_->sizes();
TORCH_CHECK(result.dim() == 2, "tensors must be 2-D");

2
aten/src/ATen/native/cuda/CuFFTUtils.h
View File

@ -66,7 +66,7 @@ static inline void CUFFT_CHECK(cufftResult error)
if (error != CUFFT_SUCCESS) {
std::ostringstream ss;
ss << "cuFFT error: " << _cudaGetErrorEnum(error);
AT_ERROR(ss.str());
TORCH_CHECK(false, ss.str());
}
}

2
aten/src/ATen/native/cuda/EmbeddingBag.cu
View File

@ -462,7 +462,7 @@ Tensor _embedding_bag_dense_backward_cuda(const Tensor &grad_, const Tensor &ind
padding_idx);
default:
AT_ERROR(
TORCH_CHECK(false,
"Unknown mode for embedding_bag_backward_cuda ", mode);
}
}

14
aten/src/ATen/native/cuda/FractionalMaxPool2d.cu
View File

@ -45,10 +45,10 @@ __device__ inline int get_interval(accscalar_t sample,
template <typename scalar_t>
__global__ void fractional_max_pool2d_out_cuda_frame(
PackedTensorAccessor<scalar_t, 4> output,
PackedTensorAccessor<int64_t, 4> indices,
PackedTensorAccessor<const scalar_t, 4> input,
PackedTensorAccessor<const scalar_t, 3> samples,
GenericPackedTensorAccessor<scalar_t, 4> output,
GenericPackedTensorAccessor<int64_t, 4> indices,
GenericPackedTensorAccessor<const scalar_t, 4> input,
GenericPackedTensorAccessor<const scalar_t, 3> samples,
int poolSizeH, int poolSizeW) {
using accscalar_t = at::acc_type<scalar_t, /*is_cuda=*/true>;
@ -102,9 +102,9 @@ __global__ void fractional_max_pool2d_out_cuda_frame(
template <typename scalar_t>
__global__ void fractional_max_pool2d_backward_out_cuda_frame(
PackedTensorAccessor<scalar_t, 4> gradInput,
PackedTensorAccessor<const scalar_t, 4> gradOutput,
PackedTensorAccessor<const int64_t, 4> indices) {
GenericPackedTensorAccessor<scalar_t, 4> gradInput,
GenericPackedTensorAccessor<const scalar_t, 4> gradOutput,
GenericPackedTensorAccessor<const int64_t, 4> indices) {
// Output (h, w) point that this thread is responsible for
int ourOutputPoint = threadIdx.x + blockIdx.x * blockDim.x;
int plane = blockIdx.y;

1
aten/src/ATen/native/cuda/KernelUtils.cuh
View File

@ -8,6 +8,7 @@
// ROCm 6.3 is planned to have these functions, but until then here they are.
#if defined(USE_ROCM) && ROCM_VERSION >= 60201
#include <hip/hip_fp16.h>
#include <hip/hip_bf16.h>
__device__ inline __hip_bfloat162 preview_unsafeAtomicAdd(__hip_bfloat162* address, __hip_bfloat162 value) {
#if (defined(__gfx940__) || defined(__gfx941__) || defined(__gfx942__)) && \

4
aten/src/ATen/native/cuda/MaxUnpooling.cu
View File

@ -267,7 +267,7 @@ static void max_unpooling3d_shape_check(
if (gradOutput.defined()) {
if (oT != gradOutput.size(dimt) || oH != gradOutput.size(dimh) ||
oW != gradOutput.size(dimw)) {
AT_ERROR(
TORCH_CHECK(false,
"Inconsistent gradOutput size. oT= ",
oT,
", oH= ",
@ -447,7 +447,7 @@ at::Tensor& max_unpooling2d_backward_out_cuda(const Tensor& grad_output_,
nInputRows = self.size(dimh);
if (oheight != grad_output.size(dimh) || owidth != grad_output.size(dimw)) {
AT_ERROR(
TORCH_CHECK(false,
"Inconsistent gradOutput size. output height: ",
oheight,
", output width= ",

6
aten/src/ATen/native/cuda/MixedDtypesLinear.cu
View File

@ -164,7 +164,7 @@ mixed_dtypes_linear_dispatch_bias_activation(
ElementInputB,
fastertransformer::EpilogueOpNoBias>(input, weight, scale, bias);
}
AT_ERROR("mixed_dtypes_linear_dispatch_bias_activation: Activation \"",
TORCH_CHECK(false, "mixed_dtypes_linear_dispatch_bias_activation: Activation \"",
activation, "\" is not supported");
return Tensor{};
}
@ -185,7 +185,7 @@ mixed_dtypes_linear_dispatch_bias_activation(
ElementInputB,
fastertransformer::EpilogueOpBiasSilu>(input, weight, scale, bias);
}
AT_ERROR("mixed_dtypes_linear_dispatch_bias_activation: Activation \"",
TORCH_CHECK(false, "mixed_dtypes_linear_dispatch_bias_activation: Activation \"",
activation, "\" is not supported");
return Tensor{};
}
@ -198,7 +198,7 @@ _mixed_dtypes_linear(const Tensor& input, const Tensor& weight,
const std::optional<Tensor>& bias_opt,
const std::optional<c10::string_view> activation_opt) {
#if defined(USE_ROCM) || defined(_MSC_VER) || (defined(CUDA_VERSION) && CUDA_VERSION < 11080)
AT_ERROR("_mixed_dtypes_linear: not compiled for this platform");
TORCH_CHECK(false, "_mixed_dtypes_linear: not compiled for this platform");
return Tensor{};
#else
const auto bias = bias_opt.has_value() ? *bias_opt : Tensor{};

4
aten/src/ATen/native/cuda/NaiveConvolutionTranspose2d.cu
View File

@ -88,7 +88,7 @@ static inline void slow_conv_transpose2d_shape_check(
check_dim_size(bias, 1, 0, weight.size(1));
}
} else if (!weight_nullable) {
AT_ERROR("weight tensor is expected to be non-nullable");
TORCH_CHECK(false, "weight tensor is expected to be non-nullable");
}
int ndim = input.dim();
@ -115,7 +115,7 @@ static inline void slow_conv_transpose2d_shape_check(
(dilation_width * (kernel_width - 1) + 1) + output_padding_width;
if (output_width < 1 || output_height < 1) {
AT_ERROR(
TORCH_CHECK(false,
"Given input size per channel: (",
input_height,
" x ",

4
aten/src/ATen/native/cuda/NaiveConvolutionTranspose3d.cu
View File

@ -106,7 +106,7 @@ static inline void slow_conv_transpose3d_shape_check(
check_dim_size(bias, 1, 0, weight.size(1));
}
} else if (!weight_nullable) {
AT_ERROR("weight tensor is expected to be non-nullable");
TORCH_CHECK(false, "weight tensor is expected to be non-nullable");
}
int ndim = input.dim();
@ -140,7 +140,7 @@ static inline void slow_conv_transpose3d_shape_check(
(dilation_width * (kernel_width - 1) + 1) + output_padding_width;
if (output_depth < 1 || output_width < 1 || output_height < 1) {
AT_ERROR(
TORCH_CHECK(false,
"Given input size per channel: (",
input_depth,
" x ",

4
aten/src/ATen/native/cuda/Sorting.cu
View File

@ -184,7 +184,7 @@ struct KthValueLauncher {
int64_t slice_size) {
dim3 grid;
if (!getGridFromTiles(num_slices, grid)) {
AT_ERROR("slices are too many");
TORCH_CHECK(false, "slices are too many");
}
dim3 block(std::min(
@ -221,7 +221,7 @@ struct MedianLauncher {
int64_t slice_size) {
dim3 grid;
if (!getGridFromTiles(num_slices, grid)) {
AT_ERROR("slices are too many");
TORCH_CHECK(false, "slices are too many");
}
dim3 block(std::min(

4
aten/src/ATen/native/cuda/SparseMM.cu
View File

@ -12,10 +12,10 @@ namespace at::native {
// sparse, sparse, sparse, dense, real, real -> sparse
Tensor& _sspaddmm_out_only_sparse_cuda(const Tensor& self,
const Tensor& mat1, const Tensor& mat2, const Scalar& beta, const Scalar& alpha, Tensor& result) {
AT_ERROR("tensor.sspaddmm(...) can only be called on sparse tensors");
TORCH_CHECK(false, "tensor.sspaddmm(...) can only be called on sparse tensors");
}
Tensor& _sspaddmm_out_cuda(const Tensor& self,
const Tensor& mat1, const Tensor& mat2, const Scalar& beta, const Scalar& alpha, Tensor& result) {
AT_ERROR("NYI: CUDA sspaddmm is not implemented");
TORCH_CHECK(false, "NYI: CUDA sspaddmm is not implemented");
}
} // namespace at::native

10
aten/src/ATen/native/cuda/SummaryOps.cu
View File

@ -251,7 +251,7 @@ Tensor _bincount_cuda_template(
const Tensor& weights,
int64_t minlength) {
if (minlength < 0) {
AT_ERROR("minlength should be >= 0");
TORCH_CHECK(false, "minlength should be >= 0");
}
if (self.dim() == 1 && self.numel() == 0) {
return at::zeros(
@ -264,12 +264,12 @@ Tensor _bincount_cuda_template(
if (self.dim() != 1 ||
(!std::is_same_v<input_t, uint8_t> &&
*self.min().cpu().const_data_ptr<input_t>() < 0)) {
AT_ERROR("bincount only supports 1-d non-negative integral inputs.");
TORCH_CHECK(false, "bincount only supports 1-d non-negative integral inputs.");
}
bool has_weights = weights.defined();
if (has_weights && (weights.dim() != 1 || weights.size(0) != self.size(0))) {
AT_ERROR("weights should be 1-d and have the same length as input");
TORCH_CHECK(false, "weights should be 1-d and have the same length as input");
}
const int64_t nbins =
@ -312,7 +312,7 @@ Tensor _histc_cuda_template(
at::acc_type<input_t, /*is_cuda=*/true> min,
at::acc_type<input_t, /*is_cuda=*/true> max) {
if (nbins <= 0) {
AT_ERROR("bins must be > 0");
TORCH_CHECK(false, "bins must be > 0");
}
Tensor output = at::zeros(
{nbins},
@ -387,7 +387,7 @@ Tensor _histc_cuda(
const Scalar& min,
const Scalar& max) {
if (self.scalar_type() == ScalarType::Half) {
AT_ERROR("HalfTensor is not supported");
TORCH_CHECK(false, "HalfTensor is not supported");
}
// See Note [Writing Nondeterministic Operations]
// Nondeterministic because of atomicAdd usage

4
aten/src/ATen/native/cuda/UpSampleBilinear2d.cu
View File

@ -37,8 +37,8 @@ __global__ void upsample_bilinear2d_out_frame(
const accscalar_t rheight,
const accscalar_t rwidth,
const bool align_corners,
const PackedTensorAccessor<const scalar_t, 4> idata,
PackedTensorAccessor<scalar_t, 4> odata) {
const GenericPackedTensorAccessor<const scalar_t, 4> idata,
GenericPackedTensorAccessor<scalar_t, 4> odata) {
int index = threadIdx.x + blockIdx.x * blockDim.x;
const int batchsize = idata.size(0);

8
aten/src/ATen/native/cuda/linalg/BatchLinearAlgebra.cpp
View File

@ -1158,7 +1158,7 @@ REGISTER_CUDA_DISPATCH(ldl_solve_stub, &ldl_solve_kernel)
template <typename scalar_t>
static void apply_cholesky_solve(Tensor& b, Tensor& A, bool upper, int64_t& info) {
#if !AT_MAGMA_ENABLED()
AT_ERROR("cholesky_solve: MAGMA library not found in "
TORCH_CHECK(false, "cholesky_solve: MAGMA library not found in "
"compilation. Please rebuild with MAGMA.");
#else
magma_uplo_t uplo = upper ? MagmaUpper : MagmaLower;
@ -1476,7 +1476,7 @@ template <typename scalar_t>
static void apply_lu_factor_looped_magma(const Tensor& input, const Tensor& pivots, const Tensor& infos, bool compute_pivots) {
#if !AT_MAGMA_ENABLED()
// This should never be thrown if the calling functions are correct.
AT_ERROR("linalg.lu_factor: PyTorch was not compiled with MAGMA support.");
TORCH_CHECK(false, "linalg.lu_factor: PyTorch was not compiled with MAGMA support.");
#else
// magmaLu and magmaLuNoPiv require infos and pivots tensor to be on CPU
// the data is later copied back to the appropriate output tensor
@ -1677,7 +1677,7 @@ REGISTER_CUDA_DISPATCH(lu_factor_stub, &lu_factor);
template <typename scalar_t>
static void apply_triangular_solve_batched_magma(const Tensor& A, const Tensor& b, bool left, bool upper, TransposeType transpose, bool unitriangular) {
#if !AT_MAGMA_ENABLED()
AT_ERROR("triangular_solve: MAGMA library not found in "
TORCH_CHECK(false, "triangular_solve: MAGMA library not found in "
"compilation. Please rebuild with MAGMA.");
#else
magma_uplo_t uplo = upper ? MagmaUpper : MagmaLower;
@ -2106,7 +2106,7 @@ static void apply_svd_magma(const Tensor& A,
const Tensor& Vh,
const Tensor& info) {
#if !AT_MAGMA_ENABLED()
AT_ERROR("linalg.svd: MAGMA library not found in "
TORCH_CHECK(false, "linalg.svd: MAGMA library not found in "
"compilation. Please rebuild with MAGMA.");
#else
using value_t = typename c10::scalar_value_type<scalar_t>::type;

2
aten/src/ATen/native/cuda/linalg/MagmaUtils.h
View File

@ -59,7 +59,7 @@ struct MAGMAQueue {
static inline magma_int_t magma_int_cast(int64_t value, const char* varname) {
auto result = static_cast<magma_int_t>(value);
if (static_cast<int64_t>(result) != value) {
AT_ERROR("magma: The value of ", varname, "(", (long long)value,
TORCH_CHECK(false, "magma: The value of ", varname, "(", (long long)value,
") is too large to fit into a magma_int_t (", sizeof(magma_int_t), " bytes)");
}
return result;

6
aten/src/ATen/native/cudnn/AffineGridGenerator.cpp
View File

@ -25,7 +25,8 @@ Tensor cudnn_affine_grid_generator_forward(
int64_t C,
int64_t H,
int64_t W) {
AT_ERROR(
TORCH_CHECK(
false,
"cudnn_affine_grid_generator_forward: ATen not compiled with cuDNN support");
}
@ -35,7 +36,8 @@ Tensor cudnn_affine_grid_generator_backward(
int64_t C,
int64_t H,
int64_t W) {
AT_ERROR(
TORCH_CHECK(
false,
"cudnn_affine_grid_generator_backward: ATen not compiled with cuDNN support");
}

Some files were not shown because too many files have changed in this diff Show More