[Refactor,CI] Refactor SOTA tests

ghstack-source-id: eaf75218bcb9051e738cc7af25092c56099b1af1
Pull Request resolved: #2583

.github/unittest/linux_examples/scripts/run_all.sh → .github/unittest/linux_sota/scripts/run_all.sh

@@ -106,30 +106,6 @@ python -c """import gym;import d4rl"""
 
 # install ale-py: manylinux names are broken for CentOS so we need to manually download and
 # rename them
-PY_VERSION=$(python --version)
-if [[ $PY_VERSION == *"3.7"* ]]; then
-  wget https://files.pythonhosted.org/packages/ab/fd/6615982d9460df7f476cad265af1378057eee9daaa8e0026de4cedbaffbd/ale_py-0.8.0-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
-  pip install ale_py-0.8.0-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
-  rm ale_py-0.8.0-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
-elif [[ $PY_VERSION == *"3.8"* ]]; then
-  wget https://files.pythonhosted.org/packages/0f/8a/feed20571a697588bc4bfef05d6a487429c84f31406a52f8af295a0346a2/ale_py-0.8.0-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
-  pip install ale_py-0.8.0-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
-  rm ale_py-0.8.0-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
-elif [[ $PY_VERSION == *"3.9"* ]]; then
-  wget https://files.pythonhosted.org/packages/a0/98/4316c1cedd9934f9a91b6e27a9be126043b4445594b40cfa391c8de2e5e8/ale_py-0.8.0-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
-  pip install ale_py-0.8.0-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
-  rm ale_py-0.8.0-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
-elif [[ $PY_VERSION == *"3.10"* ]]; then
-  wget https://files.pythonhosted.org/packages/60/1b/3adde7f44f79fcc50d0a00a0643255e48024c4c3977359747d149dc43500/ale_py-0.8.0-cp310-cp310-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
-  mv ale_py-0.8.0-cp310-cp310-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl ale_py-0.8.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
-  pip install ale_py-0.8.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
-  rm ale_py-0.8.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
-elif [[ $PY_VERSION == *"3.11"* ]]; then
-  wget https://files.pythonhosted.org/packages/60/1b/3adde7f44f79fcc50d0a00a0643255e48024c4c3977359747d149dc43500/ale_py-0.8.0-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
-  mv ale_py-0.8.0-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl ale_py-0.8.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
-  pip install ale_py-0.8.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
-  rm ale_py-0.8.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
-fi
 pip install "gymnasium[atari,accept-rom-license]<1.0"
 
 # ============================================================================================ #

.github/unittest/linux_sota/scripts/run_test.sh

@@ -0,0 +1,52 @@
+#!/usr/bin/env bash
+
+# Leave blank as code needs to start on line 29 for run_local.sh
+#
+#
+#
+#
+#
+#
+#
+
+#set -e
+set -v
+
+# Initialize an error flag
+error_occurred=0
+# Function to handle errors
+error_handler() {
+    echo "Error on line $1"
+    error_occurred=1
+}
+# Trap ERR to call the error_handler function with the failing line number
+trap 'error_handler $LINENO' ERR
+
+export PYTORCH_TEST_WITH_SLOW='1'
+python -m torch.utils.collect_env
+# Avoid error: "fatal: unsafe repository"
+git config --global --add safe.directory '*'
+
+root_dir="$(git rev-parse --show-toplevel)"
+env_dir="${root_dir}/env"
+lib_dir="${env_dir}/lib"
+
+# solves ImportError: /lib64/libstdc++.so.6: version `GLIBCXX_3.4.21' not found
+export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$lib_dir
+export MKL_THREADING_LAYER=GNU
+export CUDA_LAUNCH_BLOCKING=1
+
+python .github/unittest/helpers/coverage_run_parallel.py -m pytest test/smoke_test.py -v --durations 200
+
+coverage run -m pytest .github/unittest/linux_sota/scripts/run_tests.py --instafail --durations 200 -vvv --capture no --timeout=120
+
+coverage combine
+coverage xml -i
+
+# Check if any errors occurred during the script execution
+if [ "$error_occurred" -ne 0 ]; then
+    echo "Errors occurred during script execution"
+    exit 1
+else
+    echo "Script executed successfully"
+fi