Merge branch 'main' into fix-configs

Author: aldro61

.github/workflows/instance_pool_ci.yml

@@ -0,0 +1,82 @@
+name: Monitor the pool of WorkArena instances
+
+on:
+  workflow_dispatch:
+  schedule:
+    - cron: "0 3 * * *"       # daily at 03:00 UTC
+
+jobs:
+
+  test-l1-tasks:
+    name: Test L1 tasks
+    runs-on: ubuntu-22.04
+
+    defaults:
+      run:
+        shell: bash -l {0}
+
+    env:
+      HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
+
+    steps:
+      - name: Checkout Repository
+        uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+          cache: 'pip'
+
+      - name: Install Python dependencies
+        working-directory: ./dev
+        run: |
+          pip install -r requirements.txt
+          pip install huggingface_hub
+
+      - name: Pip list
+        run: pip list
+
+      - name: Install Playwright
+        run: playwright install chromium --with-deps
+
+      - name: Run L1 tests
+        run: pytest -n 20 --durations=10 --slowmo 1000 -v tests/test_task_general.py
+
+
+  test-snow-instance:
+    name: Test snow instance
+    runs-on: ubuntu-22.04
+    needs: test-l1-tasks   # remove this line if you want both jobs to run in parallel
+
+    defaults:
+      run:
+        shell: bash -l {0}
+
+    env:
+      HUGGING_FACE_HUB_TOKEN: ${{ secrets.HUGGING_FACE_HUB_TOKEN }}
+
+    steps:
+      - name: Checkout Repository
+        uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+          cache: 'pip'
+
+      - name: Install Python dependencies
+        working-directory: ./dev
+        run: |
+          pip install -r requirements.txt
+          pip install huggingface_hub
+
+      - name: Pip list
+        run: pip list
+
+      - name: Install Playwright
+        run: playwright install chromium --with-deps
+
+      - name: Run snow instance tests
+        run: pytest -n 20 --durations=10 --slowmo 1000 -v tests/test_snow_instance.py

.github/workflows/unit_tests.yml

@@ -36,7 +36,7 @@ jobs:
         run: black . --check
 
   browsergym-workarena-fast:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-22.04
 
     defaults:
       run:
@@ -50,7 +50,7 @@ jobs:
       - name: Set up Python
         uses: actions/setup-python@v5
         with:
-          python-version: '3.10'
+          python-version: '3.12'
           cache: 'pip' # caching pip dependencies
 
       - name: Pip install
@@ -59,9 +59,9 @@ jobs:
 
       - name: Pip list
         run: pip list
-  
+        
       - name: Install Playwright
-        run: playwright install --with-deps
+        run: playwright install chromium --with-deps
 
       - name: Run non-slow browsergym-workarena Unit Tests
         env:

tests/test_task_general.py

@@ -34,9 +34,9 @@ def snow_instance_entry(request):
 
 @retry(
     stop=stop_after_attempt(5),
-    retry=retry_if_exception_type(TimeoutError),
+    retry=retry_if_exception_type(Exception),
     reraise=True,
-    before_sleep=lambda _: logging.info("Retrying due to a TimeoutError..."),
+    before_sleep=lambda _: logging.info("Retrying due to an exception..."),
 )
 @pytest.mark.parametrize("task_entrypoint", ATOMIC_TASKS)
 @pytest.mark.parametrize("random_seed", range(1))