Commits · 88dc2040a073b270773a4d2c71d91c78e9d837cb · gaoqiong / flash-attention

11 Sep, 2022 1 commit
- Relax assert to allow both bf16 and fp16 · 13403e81
  Tri Dao authored Sep 11, 2022
  
  13403e81
06 Sep, 2022 1 commit
- Update flash_attention.py · b410d14f
  eric-tc-wong authored Sep 06, 2022
```
Recasting query and key after rotary_emb()
```
  b410d14f
09 Aug, 2022 1 commit
- Add back need_weights in FlashMHA · 19d12610
  Tri Dao authored Aug 09, 2022
  
  19d12610
05 Aug, 2022 2 commits
- Support index_first_axis with more than 2 dimensions · 6cc73425
  Tri Dao authored Aug 05, 2022
  
  6cc73425
- Allow headdim 128 in FlashMHA interface · 713ea302
  Tri Dao authored Aug 05, 2022
  
  713ea302
04 Jul, 2022 2 commits
- Do P * dP (pointwise) in the bwd in fp32 instead of fp16 · a5559a0e
  Tri Dao authored Jul 03, 2022
  
  a5559a0e
- Implement cross attention · 6c3a8c65
  Tri Dao authored Jun 30, 2022
  
  6c3a8c65
03 Jul, 2022 1 commit
- Add missing __init__.py · af4a9ce0
  Gustaf authored Jul 03, 2022
  
  af4a9ce0
02 Jun, 2022 1 commit
- Rename src -> flash_attn · 5a61cb77
  Tri Dao authored Jun 01, 2022
  
  5a61cb77