Merge branch 'main' into validate_manifest

Author: Sebastian-Larsson

.ci/docker/common/install_linter.sh

@@ -13,7 +13,3 @@ source "$(dirname "${BASH_SOURCE[0]}")/utils.sh"
 # NB: Install all linter dependencies, the caching of lintrunner init could be
 # done after Executorch becomes public
 pip_install -r requirements-lintrunner.txt
-
-# Install google-java-format
-curl -L --retry 3 --retry-all-errors https://github.com/google/google-java-format/releases/download/v1.23.0/google-java-format_linux-x86-64 > /opt/google-java-format
-chmod +x /opt/google-java-format

.ci/scripts/unittest-macos-cmake.sh

@@ -12,8 +12,19 @@ set -eux
 export TORCHINDUCTOR_CACHE_DIR="$(mktemp -d "${RUNNER_TEMP:-/tmp}/torchinductor_cache_XXXXXX")"
 trap 'rm -rf "${TORCHINDUCTOR_CACHE_DIR}"' EXIT
 
-# Run pytest with coverage
-${CONDA_RUN} pytest -n auto --cov=./ --cov-report=xml
+# TODO(SS-JIA): AOTI tests hang on macOS CI runners — the thread blocks in
+# native C/C++ code (dlopen / inductor compilation) so faulthandler cannot
+# even produce a traceback. Diagnosis ongoing in #19886.
+AOTI_SKIPS=(
+  --ignore=examples/models/llama3_2_vision/preprocess/test_preprocess.py
+  --ignore=examples/models/llama3_2_vision/vision_encoder/test/test_vision_encoder.py
+  --ignore=examples/models/llama3_2_vision/text_decoder/test/test_text_decoder.py
+  --deselect=extension/llm/modules/test/test_position_embeddings.py::TilePositionalEmbeddingTest::test_tile_positional_embedding_aoti
+  --deselect=extension/llm/modules/test/test_position_embeddings.py::TiledTokenPositionalEmbeddingTest::test_tiled_token_positional_embedding_aoti
+  --deselect=extension/llm/modules/test/test_attention.py::AttentionTest::test_attention_aoti
+)
+
+${CONDA_RUN} pytest -n auto --cov=./ --cov-report=xml "${AOTI_SKIPS[@]}"
 # Run gtest
 LLVM_PROFDATA="xcrun llvm-profdata" LLVM_COV="xcrun llvm-cov" \
 ${CONDA_RUN} test/run_oss_cpp_tests.sh

.github/scripts/propose_ghstack_orig_pr.py

@@ -52,12 +52,9 @@ def extract_stack_from_body(pr_body: str) -> List[int]:
     """
 
     prs = []
-    ghstack_begin = (
-        "Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom):"
-    )
     ghstack_begin_seen = False
     for line in pr_body.splitlines():
-        if ghstack_begin in line:
+        if line.startswith("Stack from [ghstack]"):
             ghstack_begin_seen = True
         if not ghstack_begin_seen:
             continue

.github/workflows/_unittest.yml

@@ -49,6 +49,7 @@ jobs:
       python-version: '3.11'
       submodules: 'recursive'
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
+      timeout: 120
       script: |
         set -eux
         # This is needed to get the prebuilt PyTorch wheel from S3

.github/workflows/lint.yml

@@ -125,49 +125,3 @@ jobs:
     uses: ./.github/workflows/_link_check.yml
     with:
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
-
-  android-java-format:
-    runs-on: ubuntu-latest
-    permissions:
-      contents: read
-    steps:
-      - uses: actions/checkout@v4
-        with:
-          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
-
-      - uses: actions/setup-java@v4
-        with:
-          distribution: 'temurin'
-          java-version: '17'
-
-      - name: Check Java formatting
-        run: |
-          GOOGLE_JAVA_FORMAT_VERSION="1.24.0"
-          curl -sSfL "https://github.com/google/google-java-format/releases/download/v${GOOGLE_JAVA_FORMAT_VERSION}/google-java-format-${GOOGLE_JAVA_FORMAT_VERSION}-all-deps.jar" \
-            -o /tmp/google-java-format.jar
-
-          FILES_NEEDS_FORMAT=$(find extension/android/executorch_android/src/main/java/org/pytorch/executorch \
-                              extension/android/executorch_android/src/main/java/org/pytorch/executorch/extension/llm \
-                              extension/android/executorch_android/src/main/java/org/pytorch/executorch/annotations \
-                              extension/android/executorch_android/src/androidTest/java/org/pytorch/executorch \
-                              extension/benchmark/android/benchmark/app/src/main/java/org/pytorch/minibench \
-                              extension/benchmark/android/benchmark/app/src/androidTest/java/org/pytorch/minibench \
-                              -type f -name "*.java" 2>/dev/null | \
-                              xargs -r java -jar /tmp/google-java-format.jar -n)
-
-          if [ -n "$FILES_NEEDS_FORMAT" ]; then
-            echo "Warning: The following files need formatting:"
-            echo "$FILES_NEEDS_FORMAT"
-            echo ""
-            echo "Please use google-java-format from https://github.com/google/google-java-format/releases/"
-            echo ""
-            echo "To fix, run one of these commands:"
-            echo "  # Using xargs (recommended):"
-            echo "  find <paths> -type f -name '*.java' | xargs google-java-format -i"
-            echo ""
-            echo "  # Or format specific files:"
-            echo "$FILES_NEEDS_FORMAT" | while IFS= read -r file; do
-              echo "  google-java-format -i \"$file\""
-            done
-            exit 1
-          fi

.github/workflows/mlx.yml

@@ -47,6 +47,10 @@ jobs:
 
         ${CONDA_RUN} pip list
 
+        echo "::group::Install Python test requirements"
+        ${CONDA_RUN} pip install gguf
+        echo "::endgroup::"
+
         echo "::group::Build test runners"
         ${CONDA_RUN} cmake --build cmake-out --target op_test_runner multi_thread_test_runner -j$(( $(sysctl -n hw.ncpu) - 1 ))
         echo "::endgroup::"
@@ -76,6 +80,18 @@ jobs:
         ${CONDA_RUN} python -m executorch.backends.mlx.model_ops.test_gated_delta_rule run -v
         echo "::endgroup::"
 
+        echo "::group::Run tq_norm op tests"
+        ${CONDA_RUN} python -m executorch.backends.mlx.model_ops.test_tq_norm run -v
+        echo "::endgroup::"
+
+        echo "::group::Run tq4_compress op tests"
+        ${CONDA_RUN} python -m executorch.backends.mlx.model_ops.test_tq4_compress run -v
+        echo "::endgroup::"
+
+        echo "::group::Run tq_dequant op tests"
+        ${CONDA_RUN} python -m executorch.backends.mlx.model_ops.test_tq_dequant run -v
+        echo "::endgroup::"
+
   test-mlx-qwen35-moe:
     uses: pytorch/test-infra/.github/workflows/macos_job.yml@main
     with:

.github/workflows/riscv64.yml

@@ -28,21 +28,22 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        include:
-          - { model: add,        xnnpack: false, quantize: false }
-          - { model: add,        xnnpack: true,  quantize: false }
-          - { model: mv2,        xnnpack: false, quantize: false }
-          - { model: mv2,        xnnpack: true,  quantize: false }
-          - { model: mv2,        xnnpack: true,  quantize: true }
-          - { model: mobilebert, xnnpack: false, quantize: false }
-          - { model: mobilebert, xnnpack: true,  quantize: false }
-          - { model: mobilebert, xnnpack: true,  quantize: true }
-          - { model: llama2,     xnnpack: false, quantize: false }
-          - { model: llama2,     xnnpack: true,  quantize: false }
-          - { model: llama2,     xnnpack: true,  quantize: true }
-          - { model: resnet18,   xnnpack: false, quantize: false }
-          - { model: resnet18,   xnnpack: true,  quantize: false }
-          - { model: resnet18,   xnnpack: true,  quantize: true }
+        model:
+          - add
+          - mv2
+          - mobilebert
+          - llama2
+          - resnet18
+          - yolo26
+        xnnpack: [true, false]
+        quantize: [true, false]
+        exclude:
+          # We only enable quantization with XNNPACK
+          - xnnpack: false
+            quantize: true
+          # We don't test quantization for Yolo26
+          - model: yolo26
+            quantize: true
     permissions:
       id-token: write
       contents: read

.github/workflows/trunk.yml

@@ -278,6 +278,7 @@ jobs:
       matrix:
         include:
           - test_arm_backend: test_pytest_ops_vkml
+          - test_arm_backend: test_pytest_models_vkml
           - test_arm_backend: test_ootb_tests_vgf
       fail-fast: false
     with:

.gitmodules

@@ -1,6 +1,3 @@
-[submodule "backends/arm/third-party/ethos-u-core-driver"]
-	path = backends/arm/third-party/ethos-u-core-driver
-	url = https://git.gitlab.arm.com/artificial-intelligence/ethos-u/ethos-u-core-driver.git
 [submodule "backends/vulkan/third-party/Vulkan-Headers"]
 	path = backends/vulkan/third-party/Vulkan-Headers
 	url = https://github.com/KhronosGroup/Vulkan-Headers

.lintrunner.toml

@@ -112,6 +112,8 @@ include_patterns = [
     'backends/arm/**/*.cpp',
     'backends/arm/**/*.h',
     'backends/arm/**/*.hpp',
+    'backends/cortex_m/**/*.cpp',
+    'backends/cortex_m/**/*.h',
     'examples/arm/**/*.cpp',
     'examples/arm/**/*.h',
     'examples/arm/**/*.hpp',