LlamaDemo Android #105
Workflow file for this run
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| # Copyright (c) Meta Platforms, Inc. and affiliates. | |
| # All rights reserved. | |
| # | |
| # This source code is licensed under the BSD-style license found in the | |
| # LICENSE file in the root directory of this source tree. | |
| name: LlamaDemo Android | |
| on: | |
| pull_request: | |
| branches: [main] | |
| paths: | |
| - 'llm/android/**' | |
| - '.github/workflows/llm-android.yml' | |
| workflow_dispatch: | |
| inputs: | |
| model_preset: | |
| description: 'Model preset to use' | |
| required: true | |
| type: choice | |
| options: | |
| - stories | |
| - llama | |
| - qwen3 | |
| - custom | |
| default: 'stories' | |
| custom_pte_url: | |
| description: 'Custom URL for model .pte file (only used when model_preset is custom)' | |
| required: false | |
| type: string | |
| custom_tokenizer_url: | |
| description: 'Custom URL for tokenizer file (only used when model_preset is custom)' | |
| required: false | |
| type: string | |
| permissions: | |
| contents: read | |
| jobs: | |
| instrumentation-test: | |
| runs-on: 8-core-ubuntu | |
| env: | |
| API_LEVEL: 34 | |
| ARCH: x86_64 | |
| EMULATOR_OPTIONS: -no-window -gpu swiftshader_indirect -noaudio -no-boot-anim -camera-back none | |
| RAM_SIZE: 16384 | |
| name: Instrumentation Test LlamaDemo | |
| steps: | |
| - name: Checkout repository | |
| uses: actions/checkout@v4 | |
| - name: Write job summary | |
| run: | | |
| echo "## Test Configuration" >> $GITHUB_STEP_SUMMARY | |
| echo "| Parameter | Value |" >> $GITHUB_STEP_SUMMARY | |
| echo "|-----------|-------|" >> $GITHUB_STEP_SUMMARY | |
| echo "| Model Preset | \`${{ inputs.model_preset || 'stories' }}\` |" >> $GITHUB_STEP_SUMMARY | |
| if [ "${{ inputs.model_preset }}" = "custom" ]; then | |
| echo "| Custom PTE URL | \`${{ inputs.custom_pte_url }}\` |" >> $GITHUB_STEP_SUMMARY | |
| echo "| Custom Tokenizer URL | \`${{ inputs.custom_tokenizer_url }}\` |" >> $GITHUB_STEP_SUMMARY | |
| fi | |
| - name: Enable KVM group perms | |
| run: | | |
| echo 'KERNEL=="kvm", GROUP="kvm", MODE="0666", OPTIONS+="static_node=kvm"' | sudo tee /etc/udev/rules.d/99-kvm4all.rules | |
| sudo udevadm control --reload-rules | |
| sudo udevadm trigger --name-match=kvm | |
| - name: Set up JDK 17 | |
| uses: actions/setup-java@v4 | |
| with: | |
| java-version: '17' | |
| distribution: 'temurin' | |
| - name: Setup Gradle | |
| uses: gradle/actions/setup-gradle@v4 | |
| - name: AVD cache | |
| uses: actions/cache@v4 | |
| id: avd-cache | |
| with: | |
| path: | | |
| ~/.android/avd/* | |
| ~/.android/adb* | |
| key: avd-${{ env.API_LEVEL }}-${{ env.ARCH }}-ram${{ env.RAM_SIZE }} | |
| - name: Create AVD and generate snapshot for caching | |
| if: steps.avd-cache.outputs.cache-hit != 'true' | |
| uses: reactivecircus/android-emulator-runner@v2 | |
| with: | |
| api-level: ${{ env.API_LEVEL }} | |
| arch: ${{ env.ARCH }} | |
| force-avd-creation: true | |
| emulator-options: ${{ env.EMULATOR_OPTIONS }} | |
| disable-animations: false | |
| working-directory: llm/android/LlamaDemo | |
| script: echo "Generated AVD snapshot for caching." | |
| - name: Configure AVD RAM | |
| run: | | |
| AVD_DIR="$HOME/.android/avd" | |
| for config in "$AVD_DIR"/*.avd/config.ini; do | |
| if [ -f "$config" ]; then | |
| echo "Updating RAM in $config" | |
| sed -i 's/hw.ramSize=.*/hw.ramSize=${{ env.RAM_SIZE }}/' "$config" || true | |
| grep -q "hw.ramSize" "$config" || echo "hw.ramSize=${{ env.RAM_SIZE }}" >> "$config" | |
| fi | |
| done | |
| - name: Download model files | |
| env: | |
| MODEL_PRESET: ${{ inputs.model_preset || 'stories' }} | |
| CUSTOM_PTE_URL: ${{ inputs.custom_pte_url }} | |
| CUSTOM_TOKENIZER_URL: ${{ inputs.custom_tokenizer_url }} | |
| run: | | |
| mkdir -p /tmp/llama_models | |
| # Determine URLs based on preset | |
| case "$MODEL_PRESET" in | |
| llama) | |
| PTE_URL="https://huggingface.co/executorch-community/Llama-3.2-1B-ET/resolve/main/llama3_2-1B.pte" | |
| TOKENIZER_URL="https://huggingface.co/executorch-community/Llama-3.2-1B-ET/resolve/main/tokenizer.model" | |
| ;; | |
| qwen3) | |
| PTE_URL="https://huggingface.co/pytorch/Qwen3-4B-INT8-INT4/resolve/main/model.pte" | |
| TOKENIZER_URL="https://huggingface.co/pytorch/Qwen3-4B-INT8-INT4/resolve/main/tokenizer.json" | |
| ;; | |
| custom) | |
| PTE_URL="$CUSTOM_PTE_URL" | |
| TOKENIZER_URL="$CUSTOM_TOKENIZER_URL" | |
| ;; | |
| *) | |
| PTE_URL="https://ossci-android.s3.amazonaws.com/executorch/stories/snapshot-20260114/stories110M.pte" | |
| TOKENIZER_URL="https://ossci-android.s3.amazonaws.com/executorch/stories/snapshot-20260114/tokenizer.model" | |
| ;; | |
| esac | |
| PTE_FILE=$(basename "$PTE_URL") | |
| TOKENIZER_FILE=$(basename "$TOKENIZER_URL") | |
| echo "Downloading model: $PTE_URL" | |
| curl -fL --progress-bar -o "/tmp/llama_models/$PTE_FILE" "$PTE_URL" | |
| echo "Downloading tokenizer: $TOKENIZER_URL" | |
| curl -fL --progress-bar -o "/tmp/llama_models/$TOKENIZER_FILE" "$TOKENIZER_URL" | |
| echo "Downloaded files:" | |
| ls -lh /tmp/llama_models/ | |
| # Export filenames for later steps | |
| echo "MODEL_FILE=$PTE_FILE" >> $GITHUB_ENV | |
| echo "TOKENIZER_FILE=$TOKENIZER_FILE" >> $GITHUB_ENV | |
| - name: Run instrumentation tests | |
| uses: reactivecircus/android-emulator-runner@v2 | |
| env: | |
| MODEL_PRESET: ${{ inputs.model_preset || 'stories' }} | |
| with: | |
| api-level: ${{ env.API_LEVEL }} | |
| arch: ${{ env.ARCH }} | |
| force-avd-creation: false | |
| emulator-options: -no-snapshot-save ${{ env.EMULATOR_OPTIONS }} | |
| disable-animations: true | |
| working-directory: llm/android/LlamaDemo | |
| script: | | |
| adb shell rm -rf /data/local/tmp/llama | |
| adb shell mkdir -p /data/local/tmp/llama | |
| echo "=== Pushing pre-downloaded model files to device ===" | |
| adb push /tmp/llama_models/* /data/local/tmp/llama/ | |
| echo "=== Model directory contents ===" | |
| adb shell ls -la /data/local/tmp/llama/ | |
| adb logcat -c && adb logcat > /tmp/logcat.txt & | |
| LOGCAT_PID=$! | |
| ./gradlew connectedCheck -PskipModelDownload=true \ | |
| -PmodelPreset="$MODEL_PRESET" \ | |
| -Pandroid.testInstrumentationRunnerArguments.modelFile="$MODEL_FILE" \ | |
| -Pandroid.testInstrumentationRunnerArguments.tokenizerFile="$TOKENIZER_FILE" | |
| TEST_EXIT_CODE=$? | |
| echo "=== Model directory after Gradle ===" | |
| adb shell ls -la /data/local/tmp/llama/ | |
| kill $LOGCAT_PID || true | |
| echo "=== Model configuration used by test ===" | |
| grep "UIWorkflowTest.*Using model" /tmp/logcat.txt || echo "Model config not found in logcat" | |
| echo "=== Searching for LLAMA_RESPONSE in logcat ===" | |
| grep "LLAMA_RESPONSE" /tmp/logcat.txt || echo "No LLAMA_RESPONSE found in logcat" | |
| grep "LLAMA_RESPONSE" /tmp/logcat.txt | sed 's/.*LLAMA_RESPONSE: //' | grep -v "BEGIN_RESPONSE\|END_RESPONSE" > /tmp/response.txt || true | |
| echo "=== Response file contents ===" | |
| cat /tmp/response.txt || echo "Response file empty or not created" | |
| adb shell rm -rf /data/local/tmp/llama | |
| exit $TEST_EXIT_CODE | |
| - name: Add model response to summary | |
| if: always() | |
| run: | | |
| if [ -f /tmp/response.txt ]; then | |
| echo "" >> $GITHUB_STEP_SUMMARY | |
| echo "## Model Response" >> $GITHUB_STEP_SUMMARY | |
| echo '```' >> $GITHUB_STEP_SUMMARY | |
| cat /tmp/response.txt >> $GITHUB_STEP_SUMMARY | |
| echo '```' >> $GITHUB_STEP_SUMMARY | |
| fi | |
| - name: Upload logcat | |
| if: always() | |
| uses: actions/upload-artifact@v4 | |
| with: | |
| name: logcat | |
| path: /tmp/logcat.txt | |
| retention-days: 7 |