flashattention-4

Tag

Cards List
#flashattention-4

@PyTorch: New from @Meta Engineering: FlashAttention-4 extended with MXFP8 support for @nvidia Blackwell—from forward and backwar…

X AI KOLs Timeline · 5d ago Cached

Meta Engineering extends FlashAttention-4 with MXFP8 support for NVIDIA Blackwell, achieving up to 2.85 PFLOP/s forward performance and integrating into production training workflows like GEM.

0 favorites 0 likes
#flashattention-4

@Hass_Abdallah11: We have two new Colfax articles on optimizing FlashAttention-4, featuring work by myself (decode) and Jack Carlisle (bw…

X AI KOLs Timeline · 2026-09-08 Cached

The article details optimization techniques for FlashAttention-4, including an S/P ping-pong method for decode to overlap operations, achieving up to 16% performance gain on NVIDIA Blackwell GPUs.

0 favorites 0 likes
← Back to home

Submit Feedback