FlashAttention

Other Free

Memory-efficient exact attention implementation for transformers.

About FlashAttention

FlashAttention improves speed and memory use in large model training.