Add L0 Pytorch Unittest

Author: yefubao

.github/workflows/qa_l0_pytorch.yml

@@ -75,7 +75,6 @@ jobs:
           free -h
 
       # - name: Install dependencies and build transformer_engine
-      #   env:
       #   run: |
       #     echo "\n=============== Install transformer_engine ==============="
       #     pip install --no-build-isolation -vvv . --no-deps
@@ -86,6 +85,22 @@ jobs:
         run: |
           python3 tests/pytorch/test_sanity_import.py
 
+      - name: PyTorch C++ Lint
+        id: pytorch_cpplint
+        env:
+          CPP_ONLY: 1
+          TE_PATH: ./
+        run: bash ./qa/L0_pytorch_lint/test.sh
+        continue-on-error: true
+
+      - name: PyTorch Python Lint
+        id: pytorch_pylint
+        env:
+          PYTHON_ONLY: 1
+          TE_PATH: ./
+        run: bash ./qa/L0_pytorch_lint/test.sh
+        continue-on-error: true
+
       - name: GPU Usage Check / Verification
         run: |
           source .github/workflows/scripts/gpu_check.sh
@@ -109,53 +124,42 @@ jobs:
       #     export OMP_NUM_THREADS=$((NUM_PHYSICAL_CORES / NUM_PARALLEL_JOBS))
       #     ctest --test-dir build -j$NUM_PARALLEL_JOBS
 
-      - name: L0_pytorch_debug_unittest
+      
+
+      - name: L0 Pytorch Debug Unittest
         id: L0_pytorch_debug_unittest
+        continue-on-error: true
         env:
-          PYTEST_ADDOPTS: "-v -s"
-          NVTE_TEST_NVINSPECT_FEATURE_DIRS: "transformer_engine/debug/features"
-          NVTE_TEST_NVINSPECT_CONFIGS_DIR: "tests/pytorch/debug/test_configs/"
-          NVTE_TEST_NVINSPECT_DUMMY_CONFIG_FILE: "tests/pytorch/debug/test_configs/dummy_feature.yaml"
           XML_LOG_DIR: "/logs/pytorch/debug"
+          TE_PATH: ./
         run: |
-          # Build log storage directory
-          mkdir -p "$XML_LOG_DIR"
-
-          # It is not installed as a requirement,
-          # because it is not available on PyPI.
-          pip uninstall -y nvdlfw-inspect
-          pip install git+https://github.com/NVIDIA/nvidia-dlfw-inspect.git
-          pip install pytest==8.2.1
-
-          run_pytest() {
-            local test_file=$1
-            local xml_name=$2
-            local extra_opts=${3:-""}
-            echo "🚀 Running test: $test_file"
-            pytest --junitxml="${XML_LOG_DIR}/${xml_name}" "${test_file}" "${extra_opts}"
-            if [ $? -eq 0 ]; then
-              echo "✅ Test $"xml_name" passed"
-            else
-              echo "❌ Test $"xml_name" failed"
-              exit 1
-            fi
-          }
-
-          # tests/pytorch/debug
-          run_pytest "tests/pytorch/debug/test_sanity.py" "test_sanity.xml" "--feature_dirs=$NVTE_TEST_NVINSPECT_FEATURE_DIRS"
-          run_pytest "tests/pytorch/debug/test_config.py" "test_config.xml" "--feature_dirs=$NVTE_TEST_NVINSPECT_FEATURE_DIRS"
-          run_pytest "tests/pytorch/debug/test_numerics.py" "test_numerics.xml" "--feature_dirs=$NVTE_TEST_NVINSPECT_FEATURE_DIRS"
-          run_pytest "tests/pytorch/debug/test_log.py" "test_log.xml" "--feature_dirs=$NVTE_TEST_NVINSPECT_FEATURE_DIRS --configs_dir=$NVTE_TEST_NVINSPECT_CONFIGS_DIR"
-          run_pytest "tests/pytorch/debug/test_perf.py" "test_perf.xml" "--feature_dirs=$NVTE_TEST_NVINSPECT_FEATURE_DIRS --configs_dir=$NVTE_TEST_NVINSPECT_CONFIGS_DIR"
-          # NVTE_TORCH_COMPILE=0 run_pytest "tests/pytorch/debug/test_api_features.py" "test_api_features.xml" "--feature_dirs=$NVTE_TEST_NVINSPECT_FEATURE_DIRS --configs_dir=$NVTE_TEST_NVINSPECT_CONFIGS_DIR"
-
-
-          # standard sanity and numerics tests with initialized debug
-          NVTE_TEST_NVINSPECT_ENABLED=1 NVTE_TEST_NVINSPECT_CONFIG_FILE=$NVTE_TEST_NVINSPECT_DUMMY_CONFIG_FILE NVTE_TEST_NVINSPECT_FEATURE_DIRS=$NVTE_TEST_NVINSPECT_FEATURE_DIRS PYTORCH_JIT=0 NVTE_TORCH_COMPILE=0 NVTE_ALLOW_NONDETERMINISTIC_ALGO=0 pytest --junitxml=$XML_LOG_DIR/test_sanity_2.xml tests/pytorch/test_sanity.py
-          # NVTE_TEST_NVINSPECT_ENABLED=1 NVTE_TEST_NVINSPECT_CONFIG_FILE=$NVTE_TEST_NVINSPECT_DUMMY_CONFIG_FILE NVTE_TEST_NVINSPECT_FEATURE_DIRS=$NVTE_TEST_NVINSPECT_FEATURE_DIRS PYTORCH_JIT=0 NVTE_TORCH_COMPILE=0 NVTE_ALLOW_NONDETERMINISTIC_ALGO=0 NVTE_FUSED_ATTN=0 pytest --junitxml=$XML_LOG_DIR/test_numerics_2.xml tests/pytorch/test_numerics.py
+          ignore_files=("test_api_features.py" "test_numerics.py")
+
+          for file in "${ignore_files[@]}"; do
+              sed -i 's/^.*${file}.*$/#&/' ./qa/L0_pytorch_debug_unittest/test.sh
+          done
+
+          bash ./qa/L0_pytorch_debug_unittest/test.sh
+
+      - name: L0 Pytorch Unittest
+        id: L0_pytorch_unittest
+        continue-on-error: true
+        env:
+          XML_LOG_DIR: "/logs/pytorch"
+          TE_PATH: ./
+        run: |
+          ignore_files=("test_numerics.py" "test_permutation.py" "test_attention.py" "test_checkpoint.py")
+
+          for file in "${ignore_files[@]}"; do
+              sed -i 's/^.*${file}.*$/#&/' ./qa/L0_pytorch_unittest/test.sh
+          done
+
+          bash ./qa/L0_pytorch_unittest/test.sh
 
       - name: Upload Installation Logs
-        if: failure() && steps.L0_pytorch_debug_unittest.outcome == 'failure'
+        if: |
+          (steps.L0_pytorch_debug_unittest.outcome == 'failure' || steps.L0_pytorch_unittest.outcome == 'failure')
+          && failure()
         uses: actions/upload-artifact@v4
         with:
           name: L0_pytorch-logs-${{ github.run_id }}