Skip to content

cuda: implement shared FLASH_ATTN_BACK with direct quantized cache reads - #63

Draft
joelteply wants to merge 3 commits into
fable/p1-3b-run-accumulatorfrom
codex/cuda-flash-attn-back
Draft

joelteply wants to merge 3 commits into
fable/p1-3b-run-accumulatorfrom
codex/cuda-flash-attn-back

cuda: share online attention backward reductions

8798492
Select commit
Loading
Failed to load commit list.