only download benchmark data for further uploading (#15945)

Currently we downaload everything created durning export and
benchmarking, including ptd, pte, benchmarking results, etc, when trying
to upload benchmarking result to pytorch hub. ptd and pte are large and
unnecessary for this stage and when we benchmarking lots of models, such
large files will cause out of disk space error.

this PR prevents those large and unnecessary files from downloading and
try to avoid out of disk space error.

Author: Gasoonjia

.github/workflows/cuda-perf.yml

@@ -259,8 +259,8 @@ jobs:
         CUDA_DRIVER_VERSION=$(nvidia-smi --query-gpu=driver_version --format=csv,noheader | head -1)
         echo "CUDA Driver Version: $CUDA_DRIVER_VERSION"
 
-        # Create results directory
-        RESULTS_DIR="${RUNNER_ARTIFACT_DIR}"
+        # Create results directory (separate from model artifacts)
+        RESULTS_DIR="benchmark_results"
         mkdir -p "$RESULTS_DIR"
 
         # Determine model name and runner command based on model
@@ -320,6 +320,15 @@ jobs:
           "workflow_run_url": "https://github.com/${{ github.repository }}/actions/runs/${{ github.run_id }}"
         }
         EOF
+
+        # Only copy benchmark results to RUNNER_ARTIFACT_DIR for upload (not the entire model)
+        # First, clean up the downloaded model artifacts from RUNNER_ARTIFACT_DIR
+        rm -rf "${RUNNER_ARTIFACT_DIR}"/*
+
+        # Then copy only the benchmark result JSON files
+        cp "$RESULTS_DIR"/*.json "${RUNNER_ARTIFACT_DIR}/"
+        echo "Benchmark results prepared for upload:"
+        ls -lah "${RUNNER_ARTIFACT_DIR}"
         echo "::endgroup::"
 
   upload-benchmark-results: