https://yestino.com/entities/disable-use-flash-attention-db5bb4Every event involving Disable USE_FLASH_ATTENTIONenhttps://yestino.com/events/trunk-68d20d4ee3956ceb5fecbc1676112aa32d2ad7d9-disable-use-f-266804https://yestino.com/events/trunk-68d20d4ee3956ceb5fecbc1676112aa32d2ad7d9-disable-use-f-266804Mon, 24 Aug 2026 02:06:01 GMTUSE_FLASH_ATTENTION stays ON for any CUDA build regardless of target architecture, even though the kernels need sm80+ and can_use_flash_attention() rejects any… Sources: PyTorch Releases