fixing path issue

Author: namanlalitnyu

.github/scripts/run_vllm_profiling.sh

@@ -35,10 +35,9 @@ install_dependencies() {
 }
 
 setup_workspace() {
-    # Ensure we're in the workspace directory, but don't go into vllm source
-    cd /tmp/workspace
+    WORKSPACE_DIR="/tmp/workspace"
+    cd "${WORKSPACE_DIR}"
 
-    # Create the profiling directory
     echo "Creating profiling directory: ${VLLM_TORCH_PROFILER_DIR}"
     mkdir -p "${VLLM_TORCH_PROFILER_DIR}"
     chmod 755 "${VLLM_TORCH_PROFILER_DIR}"
@@ -173,11 +172,11 @@ main() {
     # Setup phase
     print_configuration
     install_dependencies
-    # setup_workspace
+    setup_workspace
 
     # Determine the profiling test file based on device type
     local device_name="${DEVICE_NAME:-cuda}"
-    local profiling_test_file="vllm-profiling/${device_name}/profiling-tests.json"
+    local profiling_test_file="${WORKSPACE_DIR}/vllm-profiling/${device_name}/profiling-tests.json"
 
     echo "Looking for profiling test file: $profiling_test_file"
 
@@ -186,8 +185,8 @@ main() {
         run_profiling_tests "$profiling_test_file"
     else
         echo "Error: No profiling test file found at $profiling_test_file"
-        echo "Available files in vllm-profiling/:"
-        find vllm-profiling/ -name "*.json" 2>/dev/null || echo "No JSON files found"
+        echo "Available files in ${WORKSPACE_DIR}/vllm-profiling/:"
+        find "${WORKSPACE_DIR}/vllm-profiling/" -name "*.json" 2>/dev/null || echo "No JSON files found"
         exit 1
     fi
 

.github/workflows/vllm-profiling.yml

@@ -2,7 +2,6 @@ name: vLLM Profiling
 
 on:
   schedule:
-    # Run every week on Sunday at midnight
     - cron: '0 0 * * 0'
   workflow_dispatch:
     inputs:
@@ -40,7 +39,6 @@ jobs:
       fail-fast: false
       matrix:
         include:
-          # TODO: Figure out later if we need to scale this up to multiple runners
           - runs-on: linux.aws.h100.4
             device-name: cuda
     runs-on: ${{ matrix.runs-on }}
@@ -61,7 +59,6 @@ jobs:
           fetch-depth: 0
 
       - uses: actions/setup-python@v5
-        # Amazon Linux fails on this step
         continue-on-error: true
         with:
           python-version: '3.12'
@@ -118,7 +115,6 @@ jobs:
         run: |
           set -eux
 
-          # Mimic the logic from vllm ci-infra test template
           if [[ "${HEAD_BRANCH}" == "main" ]]; then
             DOCKER_IMAGE_PREFIX=public.ecr.aws/q9t5s3a7/vllm-ci-postmerge-repo
           else
@@ -143,11 +139,8 @@ jobs:
           set -eux
 
           if [[ -z "${HEAD_SHA}" ]]; then
-            # Looking back the latest 100 commits is enough
             for i in {0..99}
             do
-              # Check if the image is there, if it doesn't then check an older one
-              # because the commit is too recent
               HEAD_SHA=$(git rev-parse --verify HEAD~${i})
               DOCKER_IMAGE="${DOCKER_IMAGE_PREFIX}:${HEAD_SHA}${DOCKER_IMAGE_SUFFIX}"
 
@@ -159,8 +152,6 @@ jobs:
           fi
 
           echo "HEAD_SHA=$HEAD_SHA" >> $GITHUB_ENV
-
-          # Print the profiling commit for rereference
           echo "### Run profiling on [${HEAD_SHA}](https://github.com/vllm-project/vllm/commit/${HEAD_SHA})" >> "${GITHUB_STEP_SUMMARY}"
 
       - name: Setup CUDA GPU_FLAG for docker run
@@ -182,7 +173,6 @@ jobs:
           SCCACHE_REGION: us-east-1
           HF_TOKEN: ${{ secrets.HF_TOKEN }}
           DOCKER_IMAGE: ${{ env.DOCKER_IMAGE_PREFIX }}:${{ env.HEAD_SHA }}${{ env.DOCKER_IMAGE_SUFFIX }}
-          # vLLM-related environment variables
           VLLM_USE_MODELSCOPE: false
           VLLM_TORCH_PROFILER_DIR: /tmp/workspace/vllm-profiling/profiling-results
           CUDA_VISIBLE_DEVICES: 0
@@ -222,7 +212,6 @@ jobs:
           )
           docker exec -t "${container_name}" bash -c "cd vllm-profiling && bash ../.github/scripts/run_vllm_profiling.sh"
 
-      # Keep a copy of the profiling results on GitHub for reference
       - uses: actions/upload-artifact@v4
         with:
           name: profiling-results--${{ env.DEVICE_TYPE }}