Making FlashAttention-4 faster for inference

(modal.com) by matt_d | Jun 11, 2026 | 0 comments on HN