Merge branch 'main' into shortfin-ccache

nod-ai · Dec 5, 2024 · 5429853 · 5429853
2 parents 65bb6c5 + 57a7390
commit 5429853
Show file tree

Hide file tree

Showing 8 changed files with 83 additions and 126 deletions.
diff --git a/.github/workflows/ci-llama-large-tests.yaml b/.github/workflows/ci-llama-large-tests.yaml
@@ -33,7 +33,6 @@ jobs:
       run:
         shell: bash
     env:
-      PIP_CACHE_DIR: "${{ github.workspace }}/.pip-cache"
       VENV_DIR: ${{ github.workspace }}/.venv
     steps:
       - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
@@ -47,16 +46,12 @@ jobs:
         uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
         with:
           python-version: ${{matrix.version}}
-
-      - name: Cache Pip Packages
-        uses: actions/cache@6849a6489940f00c2f30c0fb92c6274307ccb58a # v4.1.2
-        id: cache-pip
-        with:
-          path: ${{ env.PIP_CACHE_DIR }}
-          key: pip-${{ steps.setup_python.outputs.python-version }}-${{ hashFiles('*requirements.txt') }}
+      - name: Create Python venv
+        run: python -m venv ${VENV_DIR}
 
       - name: Install pip deps
         run: |
+          source ${VENV_DIR}/bin/activate
           python -m pip install --no-compile --upgrade pip
           # Note: We install in three steps in order to satisfy requirements
           # from non default locations first. Installing the PyTorch CPU
@@ -68,14 +63,17 @@ jobs:
           pip install --no-compile -f https://iree.dev/pip-release-links.html --src deps \
             -e "git+https://github.com/iree-org/iree-turbine.git#egg=iree-turbine"
 
-
           # Test with nightly releases, not what iree-turbine uses.
           pip install -f https://iree.dev/pip-release-links.html --upgrade --pre \
             iree-base-compiler \
             iree-base-runtime
 
+          pip freeze
+
       - name: Run llama tests
-        run: pytest sharktank/tests/models/llama/benchmark_amdgpu_test.py -v -s --run-nightly-llama-tests --iree-hip-target=gfx942 --iree-device=hip://7 --html=out/llm/llama/benchmark/index.html
+        run: |
+          source ${VENV_DIR}/bin/activate
+          pytest sharktank/tests/models/llama/benchmark_amdgpu_test.py -v -s --run-nightly-llama-tests --iree-hip-target=gfx942 --iree-device=hip://7 --html=out/llm/llama/benchmark/index.html
 
       - name: Deploy to GitHub Pages
         uses: peaceiris/actions-gh-pages@4f9cc6602d3f66b9c108549d475ec49e8ef4d45e # v4.0.0

diff --git a/.github/workflows/ci-llama-quick-tests.yaml b/.github/workflows/ci-llama-quick-tests.yaml
@@ -33,7 +33,6 @@ jobs:
       run:
         shell: bash
     env:
-      PIP_CACHE_DIR: "${{ github.workspace }}/.pip-cache"
       VENV_DIR: ${{ github.workspace }}/.venv
     steps:
       - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
@@ -47,16 +46,12 @@ jobs:
         uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
         with:
           python-version: ${{matrix.version}}
-
-      - name: Cache Pip Packages
-        uses: actions/cache@6849a6489940f00c2f30c0fb92c6274307ccb58a # v4.1.2
-        id: cache-pip
-        with:
-          path: ${{ env.PIP_CACHE_DIR }}
-          key: pip-${{ steps.setup_python.outputs.python-version }}-${{ hashFiles('*requirements.txt') }}
+      - name: Create Python venv
+        run: python -m venv ${VENV_DIR}
 
       - name: Install pip deps
         run: |
+          source ${VENV_DIR}/bin/activate
           python -m pip install --no-compile --upgrade pip
           # Note: We install in three steps in order to satisfy requirements
           # from non default locations first. Installing the PyTorch CPU
@@ -68,14 +63,17 @@ jobs:
           pip install --no-compile -f https://iree.dev/pip-release-links.html --src deps \
             -e "git+https://github.com/iree-org/iree-turbine.git#egg=iree-turbine"
 
-
           # Test with nightly releases, not what iree-turbine uses.
           pip install -f https://iree.dev/pip-release-links.html --upgrade --pre \
             iree-base-compiler \
             iree-base-runtime
 
+          pip freeze
+
       - name: Run llama 8b f16 decomposed test
-        run: pytest sharktank/tests/models/llama/benchmark_amdgpu_test.py -v -s --iree-hip-target=gfx942 --iree-device=hip://0 --run-quick-llama-test
+        run: |
+          source ${VENV_DIR}/bin/activate
+          pytest sharktank/tests/models/llama/benchmark_amdgpu_test.py -v -s --iree-hip-target=gfx942 --iree-device=hip://0 --run-quick-llama-test
 
       - name: Upload llama executable files
         uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882 # v4.4.3

diff --git a/.github/workflows/ci-sdxl.yaml b/.github/workflows/ci-sdxl.yaml
@@ -35,9 +35,9 @@ env:
   LIBSHORTFIN_DIR: ${{ github.workspace }}/shortfin/
 
 jobs:
-  build-and-test:
-    name: Build and test
-    runs-on: mi300-sdxl-kernel
+  install-and-test:
+    name: Install and test
+    runs-on: mi300x-4
 
     steps:
     - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
@@ -55,53 +55,27 @@ jobs:
           sudo apt install ninja -y
         fi
 
-    - name: Checkout IREE repo
-      uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
-      with:
-        repository: iree-org/iree
-        path: ${{ env.IREE_REPO_DIR }}
-        submodules: false
-        ref: iree-3.1.0rc20241204
-
-    - name: Initalize IREE submodules
-      working-directory: ${{ env.IREE_REPO_DIR }}
-      run : |
-        git submodule update --init --depth 1 -- third_party/benchmark
-        git submodule update --init --depth 1 -- third_party/cpuinfo/
-        git submodule update --init --depth 1 -- third_party/flatcc
-        git submodule update --init --depth 1 -- third_party/googletest
-        git submodule update --init --depth 1 -- third_party/hip-build-deps/
-
     - name: Setup Python
       uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
       with:
         python-version: "3.12"
         cache: "pip"
-    - name: Install Python packages
-      # TODO: Switch to `pip install -r requirements.txt -e shortfin/`.
+
+    - name: Install requirements
       working-directory: ${{ env.LIBSHORTFIN_DIR }}
       run: |
         pip install -r requirements-tests.txt
         pip install -r requirements-iree-compiler.txt
         pip freeze
 
-    - name: Build shortfin (full)
+    - name: Install shortfin
       working-directory: ${{ env.LIBSHORTFIN_DIR }}
       run: |
-        mkdir build
-        cmake -GNinja \
-          -S. \
-          -Bbuild \
-          -DCMAKE_C_COMPILER=clang-18 \
-          -DCMAKE_CXX_COMPILER=clang++-18 \
-          -DSHORTFIN_BUNDLE_DEPS=ON \
-          -DSHORTFIN_IREE_SOURCE_DIR="${{ env.IREE_REPO_DIR }}" \
-          -DSHORTFIN_BUILD_PYTHON_BINDINGS=ON
-        cmake --build build --target all
-        pip install -v -e build/
+        pip install --no-compile -e .
 
-    - name: Test shortfin (full)
+    - name: Test apps/sd/e2e_test
       working-directory: ${{ env.LIBSHORTFIN_DIR }}
+      env:
+        HIP_VISIBLE_DEVICES: 0
       run: |
-        ctest --timeout 30 --output-on-failure --test-dir build
-        HIP_VISIBLE_DEVICES=0 pytest tests/apps/sd/e2e_test.py -v -s --system=amdgpu
+        pytest tests/apps/sd/e2e_test.py -v -s --system=amdgpu
diff --git a/.github/workflows/ci-sglang-benchmark.yml b/.github/workflows/ci-sglang-benchmark.yml
@@ -45,7 +45,7 @@ jobs:
       run:
         shell: bash
     env:
-      PIP_CACHE_DIR: "${{ github.workspace }}/.pip-cache"
+      VENV_DIR: ${{ github.workspace }}/.venv
     steps:
       - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
 
@@ -54,16 +54,12 @@ jobs:
         uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
         with:
           python-version: ${{matrix.version}}
-
-      - name: Cache Pip Packages
-        uses: actions/cache@6849a6489940f00c2f30c0fb92c6274307ccb58a # v4.1.2
-        id: cache-pip
-        with:
-          path: ${{ env.PIP_CACHE_DIR }}
-          key: pip-${{ matrix.version }}-${{ hashFiles('*requirements*.txt','shortfin/requirements*.txt','sharktank/requirements*.txt') }}
+      - name: Create Python venv
+        run: python -m venv ${VENV_DIR}
 
       - name: Install pip deps
         run: |
+          source ${VENV_DIR}/bin/activate
           python -m pip install --no-compile --upgrade pip
           # Note: We install in three steps in order to satisfy requirements
           # from non default locations first. Installing the PyTorch CPU
@@ -81,11 +77,15 @@ jobs:
             iree-base-runtime==3.1.0rc20241204 \
             "numpy<2.0"
 
-      - name: Install SGLang
-        run: pip install "git+https://github.com/nod-ai/sglang.git#subdirectory=python"
+          # Install SGLang
+          pip install "git+https://github.com/nod-ai/sglang.git#subdirectory=python"
+
+          pip freeze
 
       - name: Run Shortfin Benchmark Tests
-        run: pytest -v app_tests/benchmark_tests/llm/sglang_benchmarks/shortfin_benchmark_test.py --log-cli-level=INFO --html=shortfin_index.html --self-contained-html
+        run: |
+          source ${VENV_DIR}/bin/activate
+          pytest -v app_tests/benchmark_tests/llm/sglang_benchmarks/shortfin_benchmark_test.py --log-cli-level=INFO --html=shortfin_index.html --self-contained-html
 
       - name: Upload pytest report
         uses: actions/upload-artifact@b4b15b8c7c6ac21ea08fcf65892d2ee8f75cf882
@@ -103,8 +103,6 @@ jobs:
     defaults:
       run:
         shell: bash
-    env:
-      PIP_CACHE_DIR: "${{ github.workspace }}/.pip-cache"
     steps:
       - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
 
@@ -114,13 +112,6 @@ jobs:
         with:
           python-version: ${{matrix.version}}
 
-      - name: Cache Pip Packages
-        uses: actions/cache@6849a6489940f00c2f30c0fb92c6274307ccb58a # v4.1.2
-        id: cache-pip
-        with:
-          path: ${{ env.PIP_CACHE_DIR }}
-          key: pip-${{ matrix.version }}
-
       - name: Install SGLang
         run: |
           python -m pip install --no-compile --upgrade pip

diff --git a/.github/workflows/ci-sglang-integration-tests.yml b/.github/workflows/ci-sglang-integration-tests.yml
@@ -34,7 +34,7 @@ jobs:
       run:
         shell: bash
     env:
-      PIP_CACHE_DIR: "${{ github.workspace }}/.pip-cache"
+      VENV_DIR: ${{ github.workspace }}/.venv
     steps:
       - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
 
@@ -43,16 +43,12 @@ jobs:
         uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
         with:
           python-version: ${{matrix.version}}
-
-      - name: Cache Pip Packages
-        uses: actions/cache@6849a6489940f00c2f30c0fb92c6274307ccb58a # v4.1.2
-        id: cache-pip
-        with:
-          path: ${{ env.PIP_CACHE_DIR }}
-          key: pip-${{ steps.setup_python.outputs.python-version }}-${{ hashFiles('*requirements*.txt','shortfin/requirements*.txt','sharktank/requirements*.txt') }}
+      - name: Create Python venv
+        run: python -m venv ${VENV_DIR}
 
       - name: Install pip deps
         run: |
+          source ${VENV_DIR}/bin/activate
           python -m pip install --no-compile --upgrade pip
           # Note: We install in three steps in order to satisfy requirements
           # from non default locations first. Installing the PyTorch CPU
@@ -69,11 +65,13 @@ jobs:
             iree-base-runtime \
             "numpy<2.0"
 
-      - name: Install SGLang
-        run: pip install "git+https://github.com/nod-ai/sglang.git#subdirectory=python"
+          # Install SGLang and sentence_transformers
+          pip install "git+https://github.com/nod-ai/sglang.git#subdirectory=python"
+          pip install sentence_transformers
 
-      - name: Install sentence_transformers
-        run: pip install sentence_transformers
+          pip freeze
 
       - name: Run Integration Tests
-        run: pytest -v app_tests/integration_tests/llm/sglang --log-cli-level=INFO
+        run: |
+          source ${VENV_DIR}/bin/activate
+          pytest -v app_tests/integration_tests/llm/sglang --log-cli-level=INFO
diff --git a/.github/workflows/ci-shark-ai.yml b/.github/workflows/ci-shark-ai.yml
@@ -33,7 +33,7 @@ jobs:
       run:
         shell: bash
     env:
-      PIP_CACHE_DIR: "${{ github.workspace }}/.pip-cache"
+      VENV_DIR: ${{ github.workspace }}/.venv
     steps:
       - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
 
@@ -42,16 +42,12 @@ jobs:
         uses: actions/setup-python@0b93645e9fea7318ecaed2b359559ac225c90a2b # v5.3.0
         with:
           python-version: ${{matrix.version}}
-
-      - name: Cache Pip Packages
-        uses: actions/cache@6849a6489940f00c2f30c0fb92c6274307ccb58a # v4.1.2
-        id: cache-pip
-        with:
-          path: ${{ env.PIP_CACHE_DIR }}
-          key: pip-${{ steps.setup_python.outputs.python-version }}-${{ hashFiles('*requirements*.txt','shortfin/requirements*.txt','sharktank/requirements*.txt') }}
+      - name: Create Python venv
+        run: python -m venv ${VENV_DIR}
 
       - name: Install pip deps
         run: |
+          source ${VENV_DIR}/bin/activate
           python -m pip install --no-compile --upgrade pip
           # Note: We install in three steps in order to satisfy requirements
           # from non default locations first. Installing the PyTorch CPU
@@ -70,5 +66,9 @@ jobs:
             iree-base-compiler \
             iree-base-runtime
 
+          pip freeze
+
       - name: Run LLM Integration Tests
-        run: pytest -v app_tests/integration_tests/llm/shortfin --log-cli-level=INFO
+        run: |
+          source ${VENV_DIR}/bin/activate
+          pytest -v app_tests/integration_tests/llm/shortfin --log-cli-level=INFO