Update GitHub Actions workflow for testing on GPUs

Author: l1l1l1l

.github/workflows/test.yml

@@ -1,13 +1,95 @@
-name: Actions Runner Controller Demo
-on: [push]
+name: Run Tests on gpu and iluvatar
+on: push
 
 jobs:
-  Explore-GitHub-Actions:
-    # You need to use the INSTALLATION_NAME from the previous step
-    runs-on: bj-gpu-03
+  upload_on_gpu:
+    name: Upload on bj-gpu-03 runner
+    runs-on: [bj-gpu-03]
     container:
       image: ccr-2vdh3abv-pub.cnc.bj.baidubce.com/device/paddle-ixuca:latest
     steps:
-    - run: echo "🎉 This job uses runner scale set runners!"
-    - run: sleep 3600
-    - run: echo 'pwd'
+      - name: Checkout code
+        uses: actions/checkout@v4
+
+      - name: Setup Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.10"
+
+      - name: Install dependencies (bj-gpu-03)
+        shell: bash
+        run: |
+          set -euxo pipefail
+          pip install --pre paddlepaddle-gpu -i https://www.paddlepaddle.org.cn/packages/nightly/cu118/
+          pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
+          pip install func_timeout pandas pebble pynvml pyyaml
+      - name: Run EngineV2 (upload on bj-gpu-03)
+        shell: bash
+        run: |
+          set -euxo pipefail
+          python engineV2.py \
+            --custom_device_vs_gpu=True \
+            --custom_device_vs_gpu_mode=upload \
+            --random_seed=712 \
+            --api_config_file="tests/test.txt" \
+            --num_gpus=-1 \
+            --num_workers_per_gpu=1 \
+            --required_memory=10.0
+      - name: Package logs (bj-gpu-03)
+        if: always()
+        shell: bash
+        run: |
+          set -euxo pipefail
+          tar -czf test_log.tar.gz tester/api_config/test_log
+      - name: Upload test logs artifact (bj-gpu-03)
+        if: always()
+        uses: actions/upload-artifact@v4
+        with:
+          name: test_log_bj-gpu-03
+          path: test_log.tar.gz
+
+  download_on_iluvatar:
+    name: Download on iluvatar-gpu-2-1 runner
+    runs-on: [iluvatar-gpu-2-1]
+    needs: upload_on_gpu
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+
+      - name: Setup Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.10"
+
+      - name: Install dependencies (iluvatar-gpu-2-1)
+        shell: bash
+        run: |
+          set -euxo pipefail
+          pip install func_timeout pandas pebble pynvml pyyaml
+          python -m pip install paddlepaddle==3.2.1 -i https://www.paddlepaddle.org.cn/packages/stable/cpu/
+          python -m pip install paddle-iluvatar-gpu==3.2.1 -i https://www.paddlepaddle.org.cn/packages/stable/ixuca/
+          pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu
+      - name: Run EngineV2 (download on iluvatar-gpu-2-1)
+        shell: bash
+        run: |
+          set -euxo pipefail
+          python engineV2.py \
+            --custom_device_vs_gpu=True \
+            --custom_device_vs_gpu_mode=download \
+            --random_seed=712 \
+            --api_config_file="tests/test.txt" \
+            --num_gpus=-1 \
+            --num_workers_per_gpu=1 \
+            --required_memory=10.0
+      - name: Package logs (iluvatar-gpu-2-1)
+        if: always()
+        shell: bash
+        run: |
+          set -euxo pipefail
+          tar -czf test_log.tar.gz tester/api_config/test_log
+      - name: Upload test logs artifact (iluvatar-gpu-2-1)
+        if: always()
+        uses: actions/upload-artifact@v4
+        with:
+          name: test_log_iluvatar-gpu-2-1
+          path: test_log.tar.gz