BLaST: High Performance Inference and Pretraining using BLock Sparse Transformers

Re-assign community