.github/workflows/pkgci_test_torch.yml

# Copyright 2025 The IREE Authors
#
# Licensed under the Apache License v2.0 with LLVM Exceptions.
# See https://llvm.org/LICENSE.txt for license information.
# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception

name: PkgCI Torch Tests

permissions:
  contents: read

on:
  workflow_call:
    inputs:
      artifact_run_id:
        type: string
        default: ""
  workflow_dispatch:
    inputs:
      artifact_run_id:
        type: string
        default: ""

jobs:
  test_torch_ops:
    name: "test_torch_ops :: ${{ matrix.name }}"
    runs-on: ${{ matrix.runs-on }}
    strategy:
      fail-fast: false
      matrix:
        include:
          # CPU
          - name: cpu_task
            markers: "cpu"
            config-file: torch_ops_cpu_llvm_sync.json
            runs-on: ubuntu-24.04
          - name: amdgpu_hip_gfx1100_O3
            config-file:  torch_ops_gpu_hip_gfx1100_O3.json
            runs-on: [Linux, X64, gfx1100]
          - name: amdgpu_hip_gfx1201_O3
            config-file: torch_ops_gpu_hip_gfx1201_O3.json
            runs-on: [Linux, X64, gfx1201]
          - name: amdgpu_vulkan_O3
            config-file: torch_ops_gpu_vulkan_O3.json
            # TODO(#22579): Remove `shark10-ci` label. There are vulkan driver issues on other runners.
            runs-on: [Linux, X64, rdna3, shark10-ci]
          - name: amdgpu_rocm_mi300_gfx942_O3
            config-file: torch_ops_gpu_hip_gfx942_O3.json
            runs-on: linux-mi325-1gpu-ossci-iree-org
    env:
      PACKAGE_DOWNLOAD_DIR: ${{ github.workspace }}/.packages
      LOG_FILE_PATH: /tmp/test_torch_ops_${{ matrix.name }}_logs.json
      VENV_DIR: ${{ github.workspace }}/venv
      GH_TOKEN: ${{ github.token }}
      CONFIG_FILE_PATH: tests/external/iree-test-suites/torch_ops/${{ matrix.config-file }}
    steps:
      - name: Checking out IREE repository
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
        with:
          submodules: false
      - uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6.2.0
        with:
          # Must match the subset of versions built in pkgci_build_packages.
          python-version: "3.11"
      - uses: actions/download-artifact@37930b1c2abaa49bbe596cd826c3c89aef350131 # v7.0.0
        if: ${{ inputs.artifact_run_id == '' }}
        with:
          name: linux_x86_64_release_packages
          path: ${{ env.PACKAGE_DOWNLOAD_DIR }}
      - name: Setup venv
        run: |
          ./build_tools/pkgci/setup_venv.py ${VENV_DIR} \
            --artifact-path=${PACKAGE_DOWNLOAD_DIR} \
            --fetch-gh-workflow=${{ inputs.artifact_run_id }}

      - name: Checkout test suites repository
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
        with:
          repository: iree-org/iree-test-suites
          ref: c3b48ed923548e74b85329e201c12fe69cb1df2f
          path: iree-test-suites
          lfs: false
      - name: Install Torch ops test suite requirements
        run: |
          source ${VENV_DIR}/bin/activate
          python -m pip install -r iree-test-suites/torch_ops/requirements.txt
      - name: Run Torch ops test suite
        env:
          # Enable hip logging on errors.
          AMD_LOG_LEVEL: 1
        run: |
          source ${VENV_DIR}/bin/activate
          pytest iree-test-suites/torch_ops/ \
              -rpfE \
              --timeout=30 \
              --durations=20 \
              --report-log=${LOG_FILE_PATH} \
              --config-files=${CONFIG_FILE_PATH}

  tests:
    name: "torch_models tests :: ${{ matrix.name }}"
    runs-on: ${{ matrix.runs-on }}
    strategy:
      fail-fast: false
      matrix:
        include:
          # CPU
          - name: cpu_task
            markers: "cpu"
            cache-dir: /home/nod/iree_tests_cache
            summary-file: torch_models_cpu_task_summary.json
            runs-on:
              - self-hosted # must come first
              - persistent-cache
              - Linux
              - X64
              - threadripper
          # MI325
          - name: amdgpu_mi325_gfx942
            markers: "gfx942 or mi325"
            cache-dir: /shark-cache/data/iree-regression-cache
            summary-file: torch_models_amdgpu_mi325_summary.json
            runs-on: linux-mi325-1gpu-ossci-iree-org
    env:
      VENV_DIR: ${{ github.workspace }}/.venv
      PACKAGE_DOWNLOAD_DIR: ${{ github.workspace }}/.packages
    steps:
      - name: Checking out IREE repository
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
        with:
          submodules: false

      - uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v5.4.0
        with:
          # Must match the subset of versions built in pkgci_build_packages.
          python-version: "3.11"

      - uses: actions/download-artifact@37930b1c2abaa49bbe596cd826c3c89aef350131 # v7.0.0
        if: ${{ inputs.artifact_run_id == '' }}
        with:
          name: linux_x86_64_release_packages
          path: ${{ env.PACKAGE_DOWNLOAD_DIR }}

      - name: Checkout test suites repository
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
        with:
          repository: iree-org/iree-test-suites
          ref: c3b48ed923548e74b85329e201c12fe69cb1df2f
          path: iree-test-suites
          # Don't need lfs for torch models yet.
          lfs: false

      - name: Setup Python virtual environment
        run: |
          ./build_tools/pkgci/setup_venv.py ${VENV_DIR} \
            --artifact-path=${PACKAGE_DOWNLOAD_DIR} \
            --fetch-gh-workflow=${{ inputs.artifact_run_id }}
          source ${VENV_DIR}/bin/activate
          pip install -r ${{ github.workspace }}/iree-test-suites/torch_models/requirements.txt
        env:
          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}

      - name: Run Torch Model tests
        env:
          # Enable hip logging on errors.
          AMD_LOG_LEVEL: 1
        run: |
          source ${VENV_DIR}/bin/activate
          pytest \
            iree-test-suites/torch_models \
            -rpFe \
            --log-cli-level=info \
            -o log_cli=True \
            --durations=0 \
            --timeout=1200 \
            --capture=no \
            --test-file-directory=${{ github.workspace }}/tests/external/iree-test-suites/torch_models \
            --external-file-directory=${{ github.workspace }}/tests/external/iree-test-suites/test_suite_files \
            --module-directory=${{ github.workspace }}/tests/external/iree-test-suites/torch_models \
            --artifact-directory=${{ matrix.cache-dir }}/torch_models/artifacts \
            --job-summary-path=${{ github.workspace }}/iree-test-suites \
            -m "${{ matrix.markers }}"

      - name: Upload Torch Model Job Summary
        if: always()
        run: |
          cat ${{ github.workspace }}/iree-test-suites/job_summary.md >> $GITHUB_STEP_SUMMARY

      - name: Upload Torch Models JSON Summary
        if: always()
        uses: actions/upload-artifact@b7c566a772e6b6bfb58ed0dc250532a479d7789f # v6.0.0
        with:
          name: ${{ matrix.summary-file }}
          path: ${{ github.workspace }}/iree-test-suites/job_summary.json