Merge branch 'ROCm:main' into main

Author: vllmellm

.githooks/pre-commit

@@ -58,7 +58,7 @@ done
 if command -v clang-format >/dev/null; then
     for file in $files; do
     [[ -L $file ]] && continue
-       if [[ -e $file ]] && echo $file | grep -Eq '\.c$|\.h$|\.hpp$|\.cpp$|\.cl$|\.h\.in$|\.hpp\.in$|\.cpp\.in$'; then
+       if [[ -e $file ]] && echo $file | grep -Eq '\.c$|\.h$|\.cuh$|\.hpp$|\.cpp$|\.cl$|\.h\.in$|\.hpp\.in$|\.cpp\.in$'; then
             echo "clang-format $file"
             clang-format -i -style=file "$file"
             git add -u "$file"

.github/workflows/sglang_downstream.yaml

@@ -0,0 +1,102 @@
+name: Sglang Downstream Test
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+    branches: [main]  # Triggers on PRs targeting `main`
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  sglang:
+    name: sglang integration
+    runs-on: mi300x-1gpu
+    env:
+      SGL_BRANCH: v0.4.10
+      GPU_ARCH: gfx942
+      HF_TOKEN: ${{ secrets.HF_TOKEN_TEST }}
+
+    steps:
+      - name: Checkout aiter repo
+        uses: actions/checkout@v4
+
+      - name: Clone sglang repo
+        run: |
+          git clone https://github.com/sgl-project/sglang.git
+
+      # TODO: Currently, we use PAT, it should be replaced by repo token
+      - name: Make authenticated API request
+        run: |
+          curl -sSfL https://api.github.com/repos/casey/just/releases/latest
+      
+      - name: Workaround to change the connection of unbuntu from http to https
+        run: |
+          cd sglang
+          sed -i "79iRUN sed -i 's|http://|https://|g' /etc/apt/sources.list.d/* && sed -i 's|http://|https://|g' /etc/apt/sources.list && apt update -o Acquire::https::Verify-Peer=false && apt install -o Acquire::https::Verify-Peer=false ca-certificates -y && echo 'Acquire::https::Verify-Peer \"false\";' > /etc/apt/apt.conf.d/99insecure" docker/Dockerfile.rocm
+
+      - name: Build sglang_aiter_test image
+        run: |
+          cd sglang
+          sed -i '/ENV AITER_COMMIT="v0.1.4"/c\ARG AITER_COMMIT="v0.1.4"\nENV AITER_COMMIT=${AITER_COMMIT}' docker/Dockerfile.rocm
+          docker build --no-cache \
+          --build-arg SGL_BRANCH=${SGL_BRANCH} \
+          --build-arg GPU_ARCH=${GPU_ARCH} \
+          --build-arg AITER_REPO=${GITHUB_REPO_URL} \
+          --build-arg AITER_COMMIT=${GITHUB_COMMIT_SHA} \
+          -t sglang_aiter_test:ci -f docker/Dockerfile.rocm .
+        env:
+          GITHUB_REPO_URL: ${{ github.event.pull_request.head.repo.clone_url || 'https://github.com/ROCm/aiter.git' }}
+          GITHUB_COMMIT_SHA: ${{ github.event.pull_request.head.sha || github.event.head_commit.id }}
+
+      - name: Start CI container
+        run: |
+          echo "Clean up containers..."
+          docker ps -aq -f name=sglang_aiter_test | xargs -r docker stop | xargs -r docker rm
+
+          if [ -f "/etc/podinfo/gha-render-devices" ]; then
+            DEVICE_FLAG=$(cat /etc/podinfo/gha-render-devices)
+          else
+            DEVICE_FLAG="--device /dev/dri"
+          fi
+
+          echo "Starting container: sglang_aiter_test:ci"
+          docker run -dt --user root --device=/dev/kfd $DEVICE_FLAG \
+          -v "${GITHUB_WORKSPACE:-$PWD}/sglang:/sglang-checkout" \
+          --ipc=host --group-add video \
+          --shm-size 32g \
+          --cap-add=SYS_PTRACE \
+          -e HF_TOKEN="${HF_TOKEN:-}" \
+          --security-opt seccomp=unconfined \
+          -w /sglang-checkout \
+          --name sglang_aiter_test \
+          sglang_aiter_test:ci
+        env:
+          GITHUB_WORKSPACE: ${{ github.workspace }}
+
+      - name: Install wget
+        run: sudo apt-get update && sudo apt-get install -y wget
+
+      - name: Install dependencies
+        run: |
+          cd sglang
+          sed -i 's/ci_sglang/sglang_aiter_test/g' scripts/ci/amd_ci_install_dependency.sh
+          bash scripts/ci/amd_ci_install_dependency.sh
+
+      - name: Evaluate Accuracy
+        timeout-minutes: 60
+        run: |
+          cd sglang
+          sed -i 's/ci_sglang/sglang_aiter_test/g' scripts/ci/amd_ci_exec.sh
+          bash scripts/ci/amd_ci_exec.sh pip show aiter || true
+          bash scripts/ci/amd_ci_exec.sh -e SGLANG_USE_AITER=0 python3 test_eval_accuracy_large.py
+          bash scripts/ci/amd_ci_exec.sh python3 test_eval_fp8_accuracy.py
+          bash scripts/ci/amd_ci_exec.sh python3 models/test_qwen_models.py
+
+      # TODO: Clean up because some dependencies are installed under root user which can't be removed by runner, these dependencies should be installed as a non-root user
+      - name: Clean Up
+        if: always()
+        run:
+          docker exec -u root sglang_aiter_test bash -c "rm -rf /sglang-checkout/sgl-kernel; rm -rf /sglang-checkout/python"

3rdparty/composable_kernel

@@ -26,7 +26,17 @@ def getLogger():
             )
         console_handler.setFormatter(formatter)
         console_handler.setLevel(logging.INFO)
+
         logger.addHandler(console_handler)
+        if hasattr(torch._dynamo.config, "ignore_logger_methods"):
+            torch._dynamo.config.ignore_logger_methods = (
+                logging.Logger.info,
+                logging.Logger.warning,
+                logging.Logger.debug,
+                logger.warning,
+                logger.info,
+                logger.debug,
+            )
 
     return logger
 
@@ -51,6 +61,7 @@ def getLogger():
 from .ops.attention import *
 from .ops.custom import *
 from .ops.custom_all_reduce import *
+from .ops.quick_all_reduce import *
 from .ops.moe_op import *
 from .ops.moe_sorting import *
 from .ops.pos_encoding import *