Use a template for Android test spec

Author: huydhn

.github/workflows/android-perf.yml

@@ -30,10 +30,6 @@ on:
         description: The list of configs used the benchmark
         required: false
         type: string
-      test_spec:
-        description: The test spec to drive the test on AWS devices
-        required: false
-        type: string
   workflow_call:
     inputs:
       models:
@@ -60,10 +56,6 @@ on:
         description: The list of configs used the benchmark
         required: false
         type: string
-      test_spec:
-        description: The test spec to drive the test on AWS devices
-        required: false
-        type: string
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref_name }}-${{ github.ref_type == 'branch' && github.sha }}-${{ github.event_name == 'workflow_dispatch' }}-${{ github.event_name == 'schedule' }}
@@ -125,6 +117,32 @@ jobs:
           echo "devices=$(echo "$MAPPED_ARNS_JSON" | jq -c .)" >> $GITHUB_OUTPUT
           echo "delegates=$(echo $DELEGATES | jq -Rc 'split(",")')" >> $GITHUB_OUTPUT
 
+      - name: Prepare the spec
+        shell: bash
+        working-directory: extension/benchmark/android/benchmark
+        run: |
+          set -eux
+
+          # The model will be exported in the next step to this S3 path
+          MODEL_PATH="s3://gha-artifacts/${{ github.repository }}/${{ github.run_id }}/artifacts/${{ matrix.model }}_${{ matrix.delegate }}/model.zip"
+          # We could write a script to properly use jinja here, but there is only one variable,
+          # so let's just sed it
+          sed -i -e 's,{{ model_path }},'"${MODEL_PATH}"',g' android-llm-device-farm-test-spec.yml.j2
+          cp android-llm-device-farm-test-spec.yml.j2 android-llm-device-farm-test-spec.yml
+
+          # Just print the test spec for debugging
+          cat android-llm-device-farm-test-spec.yml
+
+      - name: Upload the spec
+        uses: seemethere/upload-artifact-s3@v5
+        with:
+          s3-bucket: gha-artifacts
+          s3-prefix: |
+            ${{ github.repository }}/${{ github.run_id }}/artifacts/${{ matrix.model }}_${{ matrix.delegate }}
+          retention-days: 1
+          if-no-files-found: error
+          path: extension/benchmark/android/benchmark/android-llm-device-farm-test-spec.yml
+
   export-models:
     name: export-models
     uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
@@ -231,10 +249,7 @@ jobs:
       device-pool-arn: ${{ matrix.device }}
       android-app-archive: https://gha-artifacts.s3.amazonaws.com/${{ github.repository }}/${{ github.run_id }}/artifacts/minibench/app-debug.apk
       android-test-archive: https://gha-artifacts.s3.amazonaws.com/${{ github.repository }}/${{ github.run_id }}/artifacts/minibench/app-debug-androidTest.apk
-      # NB: Need to set the default spec here so that it works for periodic too
-      test-spec: ${{ inputs.test_spec || 'https://ossci-android.s3.amazonaws.com/executorch/android-llm-device-farm-test-spec.yml' }}
-      # Uploaded to S3 from the previous job
-      extra-data: https://gha-artifacts.s3.amazonaws.com/${{ github.repository }}/${{ github.run_id }}/artifacts/${{ matrix.model }}_${{ matrix.delegate }}/model.zip
+      test-spec: ${{ https://gha-artifacts.s3.amazonaws.com/${{ github.repository }}/${{ github.run_id }}/artifacts/${{ matrix.model }}_${{ matrix.delegate }}/android-llm-device-farm-test-spec.yml }}
 
   upload-benchmark-results:
     needs:

.github/workflows/upload-android-test-specs.yml

@@ -4,13 +4,13 @@ on:
   pull_request:
     paths:
       - .github/workflows/upload-android-test-specs.yml
-      - extension/benchmark/android/benchmark/android-llm-device-farm-test-spec.yml
+      - extension/benchmark/android/benchmark/android-llm-device-farm-test-spec.yml.j2
   push:
     branches:
       - main
     paths:
       - .github/workflows/upload-android-test-specs.yml
-      - extension/benchmark/android/benchmark/android-llm-device-farm-test-spec.yml
+      - extension/benchmark/android/benchmark/android-llm-device-farm-test-spec.yml.j2
 
 concurrency:
   # NB: This concurency group needs to be different than the one used in android-perf, otherwise
@@ -19,21 +19,6 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
-  upload-android-test-spec-for-validation:
-    runs-on: linux.2xlarge
-    steps:
-      - uses: actions/checkout@v3
-
-      - name: Upload the spec as a GitHub artifact for validation
-        uses: seemethere/upload-artifact-s3@v5
-        with:
-          s3-bucket: gha-artifacts
-          s3-prefix: |
-            ${{ github.repository }}/${{ github.run_id }}/artifacts
-          retention-days: 1
-          if-no-files-found: error
-          path: extension/benchmark/android/benchmark/android-llm-device-farm-test-spec.yml
-
   validate-android-test-spec:
     needs: upload-android-test-spec-for-validation
     uses: ./.github/workflows/android-perf.yml
@@ -45,50 +30,3 @@ jobs:
       models: stories110M
       devices: samsung_galaxy_s22
       delegates: xnnpack
-      test_spec: https://gha-artifacts.s3.amazonaws.com/${{ github.repository }}/${{ github.run_id }}/artifacts/android-llm-device-farm-test-spec.yml
-
-  upload-android-test-spec:
-    needs: validate-android-test-spec
-    runs-on: ubuntu-22.04
-    timeout-minutes: 15
-    permissions:
-      id-token: write
-      contents: read
-    steps:
-      - uses: actions/checkout@v3
-
-      - uses: actions/setup-python@v4
-        with:
-          python-version: '3.11'
-          cache: pip
-
-      - name: configure aws credentials
-        uses: aws-actions/[email protected]
-        with:
-          role-to-assume: arn:aws:iam::308535385114:role/gha_executorch_upload-frameworks-android
-          aws-region: us-east-1
-
-      - name: Only push to S3 when running the workflow manually from main branch
-        if: ${{ github.ref == 'refs/heads/main' }}
-        shell: bash
-        run: |
-          set -eux
-          echo "UPLOAD_ON_MAIN=1" >> "${GITHUB_ENV}"
-
-      - name: Upload the spec to S3 ossci-android bucket
-        shell: bash
-        working-directory: extension/benchmark/android/benchmark/
-        env:
-          SPEC_FILE: android-llm-device-farm-test-spec.yml
-        run: |
-          set -eux
-
-          pip install awscli==1.32.18
-
-          AWS_CMD="aws s3 cp --dryrun"
-          if [[ "${UPLOAD_ON_MAIN:-0}" == "1" ]]; then
-            AWS_CMD="aws s3 cp"
-          fi
-
-          shasum -a 256 "${SPEC_FILE}"
-          ${AWS_CMD} "${SPEC_FILE}" s3://ossci-android/executorch/ --acl public-read

extension/benchmark/android/benchmark/android-llm-device-farm-test-spec.yml renamed to extension/benchmark/android/benchmark/android-llm-device-farm-test-spec.yml.j2

@@ -8,6 +8,14 @@ phases:
 
   pre_test:
     commands:
+      # Download the model from S3
+      - aws s3 cp {{ model_path }} model.zip
+      - unzip model.zip && ls -la
+
+      # Copy the extracted model to sdcard
+      - adb -s $DEVICEFARM_DEVICE_UDID push *.bin /sdcard
+      - adb -s $DEVICEFARM_DEVICE_UDID push *.pte /sdcard
+
       # Prepare the model and the tokenizer
       - adb -s $DEVICEFARM_DEVICE_UDID shell "ls -la /sdcard/"
       - adb -s $DEVICEFARM_DEVICE_UDID shell "mkdir -p /data/local/tmp/minibench/"