PR #1203 perf: prefer batched matmuls for attention

perf: prefer batched matmuls for attention #1203

patil-suraj merged 1 commit into huggingface:main from Birch-san:batch-matmul-rebased

Birch-san force pushed from 220c0f05 to 4dcce807 3 years ago

pcuenca requested a review from

pcuenca 3 years ago

patil-suraj requested a review from

NouamaneTazi 3 years ago

patrickvonplaten commented on 2022-11-20

perf: prefer batched matmuls for attention. added fast-path to Decode…

f9bf1481

Birch-san force pushed from d8ac7ee2 to f9bf1481 3 years ago

patil-suraj approved these changes on 2022-11-21

pcuenca approved these changes on 2022-11-21

patil-suraj merged ad935933 into main 3 years ago

Reviewers

pcuenca

patil-suraj

patrickvonplaten

NouamaneTazi

Assignees

No one assigned

Labels

None yet

Milestone

No milestone