Collapse cuda-perf export and test jobs (pytorch#22854)

### Summary

Same idea as pytorch#22848, for cuda-perf.

**Problem:** The cuda-perf workflow currently exports each model in one
job, uploads it to GitHub as an artifact, then downloads it in a second
job to build the runner and benchmark it. Nothing else ever touches
these artifacts, and across the 19 matrix cells they add up to about
**42 GB per run**.

**Solution:** I moved the export into the benchmark job so the model
never leaves the runner. `test_model_e2e.sh` already supports this with
`RUN_EXPORT=1`, and the benchmark itself didn't need to change since it
reads from the same directory. The job timeout goes from 90 to 120
minutes because one job now does both halves. A failed export also no
longer spawns a benchmark cell that just fails two minutes in.

### Test plan

CI

Author: jathu

.github/workflows/cuda-perf.yml

@@ -121,9 +121,21 @@ jobs:
           echo "Generated benchmark configs:"
           echo "$CONFIGS" | python -m json.tool
 
-  export-models:
-    name: export-models
-    needs: set-parameters
+  benchmark-cuda:
+    name: benchmark-cuda
+    needs:
+      - changed-files
+      - run-decision
+      - set-parameters
+    # Same gate as set-parameters, repeated so the job reads on its own;
+    # the needs on set-parameters already skips it when the gate is closed.
+    if: |
+      contains(needs.changed-files.outputs.changed-files, '.github/workflows/cuda-perf.yml') ||
+      contains(needs.changed-files.outputs.changed-files, '.ci/scripts/cuda_benchmark.py') ||
+      contains(needs.changed-files.outputs.changed-files, '.ci/scripts/cuda_perf_prompts') ||
+      contains(needs.changed-files.outputs.changed-files, '.ci/scripts/export_model_artifact.sh') ||
+      contains(needs.changed-files.outputs.changed-files, '.ci/scripts/test_model_e2e.sh') ||
+      needs.run-decision.outputs.is-full-run == 'true'
     uses: pytorch/test-infra/.github/workflows/linux_job_v3.yml@main
     permissions:
       id-token: write
@@ -133,14 +145,14 @@ jobs:
       matrix: ${{ fromJson(needs.set-parameters.outputs.benchmark_configs) }}
       fail-fast: false
     with:
-      timeout: 90
+      timeout: 120
       secrets-env: EXECUTORCH_HF_TOKEN
       runner: ${{ contains(matrix.model, 'Qwen3.5-35B-A3B') && 'mt-l-x86iavx512-11-125-a100' || 'mt-l-x86aavx2-29-113-a10g' }}
       gpu-arch-type: cuda
       gpu-arch-version: "13.0"
       use-custom-docker-registry: false
       submodules: recursive
-      upload-artifact: model-${{ matrix.model_safe }}-${{ matrix.quant }}
+      upload-artifact: results-${{ matrix.model_safe }}-${{ matrix.quant }}
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
       script: |
         set -eux
@@ -172,118 +184,10 @@ jobs:
         pip install --no-deps git+https://github.com/huggingface/optimum-executorch.git@${OPTIMUM_ET_VERSION}
         echo "::endgroup::"
 
-        echo "::group::Exporting model ${{ matrix.model }} with quantization ${{ matrix.quant }}"
-        OUTPUT_DIR="model_artifacts"
-        mkdir -p "$OUTPUT_DIR"
-
-        bash .ci/scripts/export_model_artifact.sh cuda "${{ matrix.model }}" "${{ matrix.quant }}" "$OUTPUT_DIR"
-
-        # Move artifacts to RUNNER_ARTIFACT_DIR for upload
-        mv "$OUTPUT_DIR"/* "${RUNNER_ARTIFACT_DIR}/"
-        ls -lah "${RUNNER_ARTIFACT_DIR}"
-        echo "::endgroup::"
-
-  benchmark-cuda:
-    name: benchmark-cuda
-    needs:
-      - changed-files
-      - run-decision
-      - set-parameters
-      - export-models
-    # Inherit the gate from set-parameters/export-models (they cascade-
-    # skip when the gate evaluates false). `always()` keeps benchmark-
-    # cuda running even when some export-models matrix cells fail —
-    # but only if the gate itself is open. Without the explicit gate
-    # here, `always()` would fire benchmark-cuda even when set-
-    # parameters was gated out.
-    if: |
-      always() &&
-      (
-        contains(needs.changed-files.outputs.changed-files, '.github/workflows/cuda-perf.yml') ||
-        contains(needs.changed-files.outputs.changed-files, '.ci/scripts/cuda_benchmark.py') ||
-        contains(needs.changed-files.outputs.changed-files, '.ci/scripts/cuda_perf_prompts') ||
-        contains(needs.changed-files.outputs.changed-files, '.ci/scripts/export_model_artifact.sh') ||
-        contains(needs.changed-files.outputs.changed-files, '.ci/scripts/test_model_e2e.sh') ||
-        needs.run-decision.outputs.is-full-run == 'true'
-      )
-    uses: pytorch/test-infra/.github/workflows/linux_job_v3.yml@main
-    permissions:
-      id-token: write
-      contents: read
-    strategy:
-      matrix: ${{ fromJson(needs.set-parameters.outputs.benchmark_configs) }}
-      fail-fast: false
-    with:
-      timeout: 90
-      runner: ${{ contains(matrix.model, 'Qwen3.5-35B-A3B') && 'mt-l-x86iavx512-11-125-a100' || 'mt-l-x86aavx2-29-113-a10g' }}
-      gpu-arch-type: cuda
-      gpu-arch-version: "13.0"
-      use-custom-docker-registry: false
-      submodules: recursive
-      download-artifact: model-${{ matrix.model_safe }}-${{ matrix.quant }}
-      upload-artifact: results-${{ matrix.model_safe }}-${{ matrix.quant }}
-      ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
-      script: |
-        set -eux
-
-        # OSDC mounts HF_HOME read-only at /mnt/hf_cache; redirect to a writable dir
-        # (RUNNER_TEMP, or /tmp when RUNNER_TEMP isn't writable inside the container).
-        export HF_HOME="${RUNNER_TEMP:-/tmp}/hf_cache"
-        mkdir -p "${HF_HOME}" 2>/dev/null || export HF_HOME=/tmp/hf_cache
-        mkdir -p "${HF_HOME}"
-
-        echo "::group::Setup environment"
-        # OSDC runners can't reach the public PyPI CDN that download.pytorch.org's
-        # transitive deps resolve to. Pre-install torch's pure-python deps from the
-        # in-cluster pypi-cache and drop the default cpu extra-index so the cuda
-        # torch wheel is the only candidate.
-        export PIP_EXTRA_INDEX_URL=
-        # fsspec is pinned to satisfy datasets' fsspec[http]<=2025.3.0 so the later
-        # examples install doesn't try to downgrade it from the public CDN.
-        pip install filelock typing-extensions "setuptools<82" sympy networkx jinja2 "fsspec[http]<=2025.3.0" numpy pillow
-        ./install_requirements.sh
-        pip list
-        echo "::endgroup::"
-
-        echo "::group::Prepare model artifacts"
-        mkdir -p model_artifacts
-        cp "${RUNNER_ARTIFACT_DIR}/model.pte" model_artifacts/model.pte
-        cp "${RUNNER_ARTIFACT_DIR}/aoti_cuda_blob.ptd" model_artifacts/aoti_cuda_blob.ptd
-
-        # Copy additional files if they exist
-        if [ -f "${RUNNER_ARTIFACT_DIR}/voxtral_preprocessor.pte" ]; then
-          cp "${RUNNER_ARTIFACT_DIR}/voxtral_preprocessor.pte" model_artifacts/
-        fi
-        if [ -f "${RUNNER_ARTIFACT_DIR}/whisper_preprocessor.pte" ]; then
-          cp "${RUNNER_ARTIFACT_DIR}/whisper_preprocessor.pte" model_artifacts/
-        fi
-        if [ -f "${RUNNER_ARTIFACT_DIR}/tekken.json" ]; then
-          cp "${RUNNER_ARTIFACT_DIR}/tekken.json" model_artifacts/
-        fi
-        if [ -f "${RUNNER_ARTIFACT_DIR}/poem.wav" ]; then
-          cp "${RUNNER_ARTIFACT_DIR}/poem.wav" model_artifacts/
-        fi
-        if [ -f "${RUNNER_ARTIFACT_DIR}/output.wav" ]; then
-          cp "${RUNNER_ARTIFACT_DIR}/output.wav" model_artifacts/
-        fi
-        if [ -f "${RUNNER_ARTIFACT_DIR}/tokenizer.model" ]; then
-          cp "${RUNNER_ARTIFACT_DIR}/tokenizer.model" model_artifacts/
-        fi
-        if [ -f "${RUNNER_ARTIFACT_DIR}/test_audio.wav" ]; then
-          cp "${RUNNER_ARTIFACT_DIR}/test_audio.wav" model_artifacts/
-        fi
-        # Copy tokenizer files
-        for file in tokenizer.json tokenizer_config.json special_tokens_map.json; do
-          if [ -f "${RUNNER_ARTIFACT_DIR}/$file" ]; then
-            cp "${RUNNER_ARTIFACT_DIR}/$file" model_artifacts/
-          fi
-        done
-
-        ls -lah model_artifacts/
-        echo "::endgroup::"
-
-        echo "::group::Build runner"
-        bash .ci/scripts/test_model_e2e.sh cuda "${{ matrix.model }}" "${{ matrix.quant }}" model_artifacts
+        echo "::group::Export model and build runner"
+        # Export and benchmark in the same job so the model never leaves the runner
+        # (no GitHub Actions artifact upload/download).
+        RUN_EXPORT=1 bash .ci/scripts/test_model_e2e.sh cuda "${{ matrix.model }}" "${{ matrix.quant }}" model_artifacts
         echo "::endgroup::"
 
         echo "::group::Running benchmark for ${{ matrix.model }} (${{ matrix.quant }}) with ${{ matrix.num_runs }} runs"
@@ -377,11 +281,7 @@ jobs:
         }
         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 "${RUNNER_ARTIFACT_DIR}"/*
-
-        # Then copy only the benchmark result JSON files
+        # Upload only the benchmark result JSON files
         cp "$RESULTS_DIR"/*.json "${RUNNER_ARTIFACT_DIR}/"
         echo "Benchmark results prepared for upload:"
         ls -lah "${RUNNER_ARTIFACT_DIR}"