[HN Gopher] Jagged Flash Attention Optimization
___________________________________________________________________
Jagged Flash Attention Optimization
Author : tullie
Score : 13 points
Date : 2025-03-18 17:45 UTC (3 days ago)
(HTM) web link (www.shaped.ai)
(TXT) w3m dump (www.shaped.ai)
| platers wrote:
| Flash attention natively supports packing multiple variable
| length sequences into a single call, what is the advantage of
| jagged flash attention?
___________________________________________________________________
(page generated 2025-03-21 23:00 UTC)