[Nightly] Enhance XPU test workflows

.github/actions/diskspace-clean/action.yml

@@ -0,0 +1,23 @@
+name: Cleans up diskspace
+
+description: Cleans up diskspace
+
+runs:
+  using: composite
+  steps:
+    - name: Cleans up diskspace
+      shell: bash
+      run: |
+        set -xe -o pipefail
+        # Clean workspace
+        rm -rf ${{ github.workspace }}/* || sudo rm -rf ${{ github.workspace }}/*
+        # Clean cache
+        rm -rf /tmp/ || sudo rm -rf /tmp/
+        mkdir -m 777 /tmp || sudo mkdir -m 777 /tmp
+        rm -rf ~/.torch || sudo rm -rf ~/.torch
+        rm -rf ~/.triton || sudo rm -rf ~/.triton
+        rm -rf ~/.cache || sudo rm -rf ~/.cache
+        rm -rf ~/.conda || sudo rm -rf ~/.conda
+        # Clean docker
+        docker stop $(docker ps -aq) || true
+        docker system prune -af

.github/actions/inductor-xpu-e2e-test/action.yml

@@ -52,14 +52,6 @@ runs:
       run: |
         source activate e2e_ci
         if [[ ${{ inputs.suite }} == *"torchbench"* ]]; then
-          if [ "${{ inputs.pytorch }}" != "nightly_wheel" ]; then
-            cd ../ && rm -rf audio && git clone --single-branch -b main https://github.com/pytorch/audio.git
-            cd audio && git checkout $TORCHAUDIO_COMMIT_ID
-            python setup.py bdist_wheel && pip uninstall torchaudio -y && pip install dist/*.whl
-            cd ../ && rm -rf vision && git clone --single-branch -b main https://github.com/pytorch/vision.git
-            cd vision && git checkout $TORCHVISION_COMMIT_ID
-            python setup.py bdist_wheel && pip uninstall torchvision -y && pip install dist/*.whl
-          fi
           cd ../ && python -c "import torch, torchvision, torchaudio"
           rm -rf benchmark && git clone https://github.com/pytorch/benchmark.git
           cd benchmark && git checkout $TORCHBENCH_COMMIT_ID
@@ -80,11 +72,6 @@ runs:
           pip install --force-reinstall git+https://github.com/huggingface/transformers@${TRANSFORMERS_VERSION}
         fi
         if [[ ${{ inputs.suite }} == *"timm_models"* ]]; then
-          if [ "${{ inputs.pytorch }}" != "nightly_wheel" ]; then
-            cd ../ && rm -rf vision && git clone --single-branch -b main https://github.com/pytorch/vision.git
-            cd vision && git checkout $TORCHVISION_COMMIT_ID
-            python setup.py bdist_wheel && pip uninstall torchvision -y && pip install dist/*.whl
-          fi
           # install timm without dependencies
           pip install --no-deps git+https://github.com/huggingface/pytorch-image-models@$TIMM_COMMIT_ID
           # install timm dependencies without torch and torchvision

.github/actions/testenv-setup/action.yml

@@ -0,0 +1,57 @@
+name: Cleans up diskspace
+
+description: Cleans up diskspace
+
+inputs:
+  suite:
+    required: true
+    type: string
+    default: 'huggingface'
+    description: Dynamo benchmarks test suite. huggingface,timm_models,torchbench. Delimiter is comma
+  env_prepare:
+    required: false
+    description: If set to any value, will prepare suite test env
+  dt:
+    required: true
+    type: string
+    default: 'float32'
+    description: Data precision of the test.float32,bfloat16,float16,amp_bf16,amp_fp16. Delimiter is comma
+  mode:
+    required: true
+    type: string
+    default: 'inference'
+    description: inference,training. Delimiter is comma
+  scenario:
+    required: true
+    type: string
+    default: 'accuracy'
+    description: accuracy,performance. Delimiter is comma
+  cards:
+    required: false
+    type: string
+    default: 'all'
+    description: which cards can be used in the test
+  hf_token:
+    required: false
+    description: HUGGING_FACE_HUB_TOKEN for torchbench test
+  pytorch:
+    required: false
+    type: string
+    default: 'main'
+    description: Pytorch branch/commit
+  driver:
+    required: false
+    type: string
+    default: 'lts'
+    description: Driver lts/rolling
+
+runs:
+  using: composite
+  defaults:
+    run:
+      shell: bash -xe -o pipefail
+  steps:
+    - name: Prepare PyTorch
+      shell: bash
+      run: |
+        echo

.github/scripts/build.sh

@@ -44,7 +44,7 @@ cp -r ${WORKSPACE}/torch-xpu-ops third_party/torch-xpu-ops
 
 # Pre Build
 cd ${WORKSPACE}/pytorch
-python -m pip install requests
+python -m pip install requests cmake ninja
 python third_party/torch-xpu-ops/.github/scripts/apply_torch_pr.py
 git submodule sync && git submodule update --init --recursive
 python -m pip install -r requirements.txt
@@ -85,6 +85,17 @@ rm -rf ./tmp
 bash third_party/torch-xpu-ops/.github/scripts/rpath.sh ${WORKSPACE}/pytorch/dist/torch*.whl
 python -m pip install --force-reinstall tmp/torch*.whl
 
+# Build torchvision torchaudio
+unset PYTORCH_VERSION
+TORCHVISION_COMMIT=$(cat .github/ci_commit_pins/vision.txt)
+rm -rf xpu-vision && git clone https://github.com/pytorch/vision.git xpu-vision
+cd xpu-vision && git checkout ${TORCHVISION_COMMIT}
+python setup.py bdist_wheel && cd ..
+TORCHAUDIO_COMMIT=$(cat .github/ci_commit_pins/audio.txt)
+rm -rf xpu-audio && git clone https://github.com/pytorch/audio.git xpu-audio
+cd xpu-audio && git checkout ${TORCHAUDIO_COMMIT}
+python setup.py bdist_wheel && cd ..
+
 # Verify
 cd ${WORKSPACE}
 python ${WORKSPACE}/pytorch/torch/utils/collect_env.py
@@ -95,6 +106,8 @@ xpu_is_compiled="$(python -c 'import torch; print(torch.xpu._is_compiled())')"
 # Save wheel
 if [ "${xpu_is_compiled,,}" == "true" ];then
     cp ${WORKSPACE}/pytorch/tmp/torch*.whl ${WORKSPACE}
+    cp ${WORKSPACE}/pytorch/xpu-vision/dist/torchvision*.whl ${WORKSPACE}
+    cp ${WORKSPACE}/pytorch/xpu-audio/dist/torchaudio*.whl ${WORKSPACE}
 else
     echo "Build got failed!"
     exit 1

.github/scripts/lintrunner.sh

@@ -24,7 +24,7 @@ if ! command -v lintrunner &> /dev/null; then
 fi
 
 # Ignoring errors in one specific run
-export SHELLCHECK_OPTS="-e SC2154 -e SC2086 -e SC1091 -e SC2046"
+export SHELLCHECK_OPTS="-e SC2154 -e SC2086 -e SC1091 -e SC2046 -e SC1090"
 
 # This has already been cached in the docker image
 lintrunner init 2> /dev/null

.github/scripts/setup_test_env.sh

@@ -0,0 +1,87 @@
+#!/bin/bash
+
+set -xe -o pipefail
+export GIT_PAGER=cat
+
+# Init params
+WORKSPACE=$(realpath ${WORKSPACE:-"/tmp"})
+CONDA_ENV=${CONDA_ENV:-"xpu-test"}
+PYTHON_VERSION=${PYTHON_VERSION:-"3.10"}
+PYTORCH_REPO=${PYTORCH_REPO:-"https://github.com/pytorch/pytorch.git"}
+PYTORCH_VERSION=${PYTORCH_VERSION:-"main"}
+for var; do
+    eval "export $(echo ${var@Q} |sed "s/^'-*//g;s/=/='/")"
+done
+
+# Python env via conda
+. "$(conda info -e |awk '{if($1=="base"){printf("%s/etc/profile.d/conda.sh", $NF)}}')"
+conda create python=${PYTHON_VERSION} -y -n ${CONDA_ENV}
+conda activate ${CONDA_ENV}
+conda info -e
+which python && python -V && conda list
+python -m pip install requests pandas scipy psutil
+
+# Prepare pytorch
+if [ "${PYTORCH_VERSION}" == "release" ];then
+    python -m pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/xpu
+elif [ "${PYTORCH_VERSION}" == "test" ];then
+    python -m pip install torch torchvision torchaudio --pre --index-url https://download.pytorch.org/whl/test/xpu
+elif [ "${PYTORCH_VERSION}" == "nightly" ];then
+    python -m pip install torch torchvision torchaudio --pre --index-url https://download.pytorch.org/whl/nightly/xpu
+else
+    python -m pip install ${WORKSPACE}/torch*.whl
+fi
+TORCH_COMMIT="$(python -c 'import torch; print(torch.version.git_version)')"
+rm -rf ./pytorch
+git clone ${PYTORCH_REPO} pytorch
+cd pytorch
+git checkout ${TORCH_COMMIT}
+git remote -v && git branch && git show -s
+
+# Prepare torch-xpu-ops
+rm -rf third_party/torch-xpu-ops
+if [ "${PYTORCH_VERSION}" != "main" ];then
+    TORCH_XPU_OPS_COMMIT=$(cat third_party/xpu.txt)
+    git clone https://github.com/intel/torch-xpu-ops.git third_party/torch-xpu-ops
+    cd third_party/torch-xpu-ops
+    git checkout ${TORCH_XPU_OPS_COMMIT}
+else
+    cp -r ${WORKSPACE} third_party/torch-xpu-ops
+    cd third_party/torch-xpu-ops
+fi
+git remote -v && git branch && git show -s
+cd ../..
+if [ "${GITHUB_EVENT_NAME}" == "pull_request" ];then
+    python third_party/torch-xpu-ops/.github/scripts/apply_torch_pr.py -e https://github.com/pytorch/pytorch/pull/152940
+else
+    python third_party/torch-xpu-ops/.github/scripts/apply_torch_pr.py
+fi
+
+# Install triton
+if [ "${TRITON_VERSION}" == "pinned" ];then
+    TRITON_VERSION="$(cat .ci/docker/ci_commit_pins/triton-xpu.txt)"
+fi
+if [ -n "${TRITON_VERSION}" ];then
+    TRITON_REPO="https://github.com/intel/intel-xpu-backend-for-triton"
+    python -m pip uninstall -y pytorch-triton-xpu
+    python -m pip install "git+${TRITON_REPO}@${TRITON_VERSION}#subdirectory=python"
+fi
+
+# Install requirements
+python -m pip install -r .ci/docker/requirements-ci.txt
+python -m pip install -U pytest pytest-timeout
+
+# Collect env infos
+cd ..
+python -c "import torch; print(torch.__config__.show())"
+python -c "import torch; print(torch.__config__.parallel_info())"
+python -c "import torch; print(torch.__config__.torch.xpu.device_count())"
+python -c "import triton; print(triton.__version__)"
+python pytorch/torch/utils/collect_env.py
+
+# Clean cache
+rm -rf /tmp/ || sudo rm -rf /tmp/
+mkdir -m 777 /tmp || sudo mkdir -m 777 /tmp
+rm -rf ~/.torch || sudo rm -rf ~/.torch
+rm -rf ~/.triton || sudo rm -rf ~/.triton
+rm -rf ~/.cache || sudo rm -rf ~/.cache

.github/workflows/_linux_build.yml

@@ -28,11 +28,6 @@ on:
         type: string
         default: 'linux.idc.xpu'
         description: Runner label
-      update_lkg:
-        required: false
-        type: string
-        default: 'false'
-        description: Whether update LKG torch version to issue #1280
     outputs:
       torch_commit_id:
         description: The commit id of the torch build