TBNilles/LEANN
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: disable OpenMP parallelism in CI to avoid libomp crashes

Browse Source 
- Set OMP_NUM_THREADS=1 to avoid OpenMP thread synchronization issues
- Set MKL_NUM_THREADS=1 for single-threaded MKL operations
- This prevents segfaults in LayerNorm on macOS CI runners
- Addresses the libomp compatibility issues with PyTorch on Apple Silicon
This commit is contained in:
Andy Lee
2025-07-28 16:31:41 -07:00
parent f2ffcf5665
commit d8b6ae8d1a
1 changed files with 3 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
.github/workflows/build-reusable.yml vendored
View File

@@ -224,10 +224,12 @@ jobs:
HF_HUB_DISABLE_SYMLINKS: 1 HF_HUB_DISABLE_SYMLINKS: 1
TOKENIZERS_PARALLELISM: false TOKENIZERS_PARALLELISM: false
PYTORCH_ENABLE_MPS_FALLBACK: 0 # Disable MPS on macOS CI to avoid memory issues PYTORCH_ENABLE_MPS_FALLBACK: 0 # Disable MPS on macOS CI to avoid memory issues
OMP_NUM_THREADS: 1 # Disable OpenMP parallelism to avoid libomp crashes
MKL_NUM_THREADS: 1 # Single thread for MKL operations
run: | run: |
# Activate virtual environment # Activate virtual environment
source .venv/bin/activate || source .venv/Scripts/activate source .venv/bin/activate || source .venv/Scripts/activate
# Run all tests # Run all tests
pytest tests/ pytest tests/