Skip to content

Add upsample_bilinear2d channels last kernel implementation #3865

Add upsample_bilinear2d channels last kernel implementation

Add upsample_bilinear2d channels last kernel implementation #3865

Workflow file for this run

name: pull
on:
pull_request:
types:
- opened
- synchronize
- reopened
- converted_to_draft
- ready_for_review
branches:
- main
- release/*
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}
cancel-in-progress: true
permissions: read-all
jobs:
preci-ut:
# Don't run on forked repos and draft PRs
if: ${{ (github.repository_owner == 'intel') && (github.event.pull_request.draft == false) }}
uses: ./.github/workflows/_linux_ut.yml
with:
ut: op_regression,op_regression_dev1,op_extended,op_ut
runner: linux.idc.xpu
preci-ut-abi-0:
# Don't run on forked repos and draft PRs
if: ${{ (github.repository_owner == 'intel') && (github.event.pull_request.draft == false) }}
uses: ./.github/workflows/_linux_ut.yml
with:
abi: 0
ut: op_extended
runner: linux.idc.xpu
Inductor-XPU-E2E-CI-Tests:
runs-on: pvc_e2e
# Don't run on forked repos and draft PRs
if: ${{ (github.repository_owner == 'intel') && (github.event.pull_request.draft == false) }}
timeout-minutes: 900
steps:
- name: Checkout torch-xpu-ops
uses: actions/checkout@v4
- name: Prepare Conda ENV
run: |
which conda && conda clean -ay
conda remove --all -y -n e2e_ci || rm -rf $(dirname ${CONDA_EXE})/../envs/e2e_ci
conda create -n e2e_ci python=3.10 cmake ninja -y
source activate e2e_ci
pip install mkl-static mkl-include
pip install pandas scipy tqdm
- name: Prepare Stock Pytorch
run: |
pwd
cd ../ && rm -rf pytorch
source activate e2e_ci
git clone -b main https://github.com/pytorch/pytorch pytorch
cd pytorch
# apply PRs for stock pytorch
pip install requests
# https://github.com/mengfei25/pytorch/pull/18 internal use only for subset model list
python ../torch-xpu-ops/.github/scripts/apply_torch_pr.py -e https://github.com/mengfei25/pytorch/pull/18
git status && git show -s
git submodule sync && git submodule update --init --recursive
rm -rf third_party/torch-xpu-ops && cp -r ../torch-xpu-ops third_party/
# Workaround for torch-xpu-ops ci test
sed -i "s/checkout --quiet \${TORCH_XPU_OPS_COMMIT}/log -n 1/g" caffe2/CMakeLists.txt
- name: Triton Installation
run: |
source activate e2e_ci
cd ../pytorch
TRITON_REPO="https://github.com/intel/intel-xpu-backend-for-triton"
TRITON_PINNED_COMMIT=$(cat .ci/docker/ci_commit_pins/triton-xpu.txt)
echo ${TRITON_REPO}@${TRITON_PINNED_COMMIT}
pip install --force-reinstall "git+${TRITON_REPO}@${TRITON_PINNED_COMMIT}#subdirectory=python"
- name: Build Pytorch XPU
run: |
source activate e2e_ci
source .github/scripts/env.sh
cd ../pytorch
pip install -r requirements.txt
export CMAKE_PREFIX_PATH=${CMAKE_PREFIX_PATH}:${CONDA_PREFIX:-"$(dirname $(which conda))/../"}
python setup.py bdist_wheel
pip install --force-reinstall dist/*.whl
- name: Identify pinned versions
run: |
cd ../pytorch
echo "TRITON_COMMIT_ID=$(<.ci/docker/ci_commit_pins/triton-xpu.txt)" >> "${GITHUB_ENV}"
echo "TORCHVISION_COMMIT_ID=$(<.github/ci_commit_pins/vision.txt)" >> "${GITHUB_ENV}"
echo "TORCHBENCH_COMMIT_ID=$(<third_party/torch-xpu-ops/.github/ci_commit_pins/torchbench.txt)" >> "${GITHUB_ENV}"
echo "TORCHAUDIO_COMMIT_ID=$(<.github/ci_commit_pins/audio.txt)" >> "${GITHUB_ENV}"
echo "TRANSFORMERS_VERSION=$(<.ci/docker/ci_commit_pins/huggingface.txt)" >> "${GITHUB_ENV}"
echo "TIMM_COMMIT_ID=$(<.ci/docker/ci_commit_pins/timm.txt)" >> "${GITHUB_ENV}"
- name: Torch Config
run: |
echo "$GITHUB_ENV"
rm -rf ../pytorch/inductor_log
rm -rf /tmp/torchinductor_*
cd ..
source activate e2e_ci
python -c "import triton; print(triton.__version__)"
python pytorch/torch/utils/collect_env.py
- name: Huggingface BF16 Training Accuracy Test
uses: ./.github/actions/inductor-xpu-e2e-test
with:
suite: huggingface
dt: bfloat16
mode: training
scenario: accuracy
env_prepare: true
hf_token: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
- name: Huggingface FP16 Training Accuracy Test
uses: ./.github/actions/inductor-xpu-e2e-test
with:
suite: huggingface
dt: float16
mode: training
scenario: accuracy
hf_token: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
- name: Timm_models BF16 Training Accuracy Test
uses: ./.github/actions/inductor-xpu-e2e-test
with:
suite: timm_models
dt: bfloat16
mode: training
scenario: accuracy
env_prepare: true
hf_token: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
- name: Torchbench BF16 Training Accuracy Test
uses: ./.github/actions/inductor-xpu-e2e-test
with:
suite: torchbench
dt: bfloat16
mode: training
scenario: accuracy
env_prepare: true
hf_token: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
- name: Summarize archieve files
if: ${{ ! cancelled() }}
run: |
rm -rf ${{ github.workspace }}/upload_files
cp -r ${{ github.workspace }}/../pytorch/inductor_log ${{ github.workspace }}/upload_files
failed_case=$(grep "Real failed: models: *[1-9]" ${{ github.workspace }}/upload_files/summary_accuracy.log |wc -l || true)
if [ ${failed_case} -ne 0 ];then
grep -E "Real failed: models: [1-9]|Summary for" ${{ github.workspace }}/summary_accuracy.log
exit 1
fi
- name: Upload Inductor XPU E2E Data
if: ${{ ! cancelled() }}
uses: actions/upload-artifact@v4
with:
name: Inductor-XPU-E2E-Data-${{ github.event.pull_request.number || github.sha }}
path: ${{ github.workspace }}/upload_files