Move HF LLM to nightly

Author: winskuo-quic

.github/workflows/nightly.yml

@@ -36,3 +36,37 @@ jobs:
     uses: ./.github/workflows/_link_check.yml
     with:
       ref: ${{ github.sha }}
+    
+  test-static-hf-llm-qnn-linux:
+    name: test-static-hf-llm-qnn-linux
+    uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main
+    permissions:
+      id-token: write
+      contents: read
+    strategy:
+      matrix:
+        task: [smollm2_135m]
+      fail-fast: false
+    with:
+      runner: linux.24xlarge
+      docker-image: ci-image:executorch-ubuntu-22.04-qnn-sdk
+      submodules: 'recursive'
+      ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
+      timeout: 900
+      script: |
+        # The generic Linux job chooses to use base env, not the one setup by the image
+        CONDA_ENV=$(conda env list --json | jq -r ".envs | .[-1]")
+        conda activate "${CONDA_ENV}"
+
+        BUILD_TOOL="cmake"
+
+        PYTHON_EXECUTABLE=python bash .ci/scripts/setup-qnn-deps.sh
+        PYTHON_EXECUTABLE=python bash .ci/scripts/build-qnn-sdk.sh
+
+        # Setup executorch
+        PYTHON_EXECUTABLE=python bash .ci/scripts/setup-linux.sh --build-tool "${BUILD_TOOL}"
+
+        # Setup install_requirements for llama
+        PYTHON_EXECUTABLE=python bash examples/models/llama/install_requirements.sh
+
+        PYTHON_EXECUTABLE=python bash .ci/scripts/test_qnn_static_llm.sh ${{ matrix.task }}

.github/workflows/pull.yml

@@ -559,22 +559,22 @@ jobs:
         # Test llama2
         PYTHON_EXECUTABLE=python bash .ci/scripts/test_llama.sh -model stories110M -build_tool "${BUILD_TOOL}" -mode "${MODE}" -dtype "${DTYPE}" -pt2e_quantize "${PT2E_QUANTIZE}"
 
-  test-static-llm-qnn-linux:
-    name: test-static-llm-qnn-linux
+  test-static-llama-qnn-linux:
+    name: test-static-llama-qnn-linux
     uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main
     permissions:
       id-token: write
       contents: read
     strategy:
       matrix:
-        task: [stories_110m, stories_260k_bc, smollm2_135m]
+        task: [stories_110m, stories_260k_bc]
       fail-fast: false
     with:
-      runner: linux.24xlarge
+      runner: linux.2xlarge
       docker-image: ci-image:executorch-ubuntu-22.04-qnn-sdk
       submodules: 'recursive'
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
-      timeout: 900
+      timeout: 180
       script: |
         # The generic Linux job chooses to use base env, not the one setup by the image
         CONDA_ENV=$(conda env list --json | jq -r ".envs | .[-1]")