Commits · f2d8d4104e276f0a9ead015a8f6fdc615a71ad29 · gaoqiong / flash-attention

".circleci/unittest/vscode:/vscode.git/clone" did not exist on "71c2ae7793de0c1ceb9cc96ef6d14bf3e302fb4f"

04 Jun, 2022 1 commit
- Don't use Smem_dp_sum in backward pass · d380e87f
  Tri Dao authored Jun 04, 2022
```
To reduce smem usage for SM75
```
  d380e87f
03 Jun, 2022 2 commits
- Reduce smem usage for Q and dO in the backward pass · b17c6fe2
  Tri Dao authored Jun 03, 2022
```
From 4KB per buffer to 2KB per buffer. This saves us 8KB of smem (each Q and dO
have 2 buffers)
```
  b17c6fe2
- Support Turing mma instructions · 2712aa4c
  Tri Dao authored Jun 02, 2022
  
  2712aa4c
02 Jun, 2022 6 commits
- Remove softmax fp16 max · 05087332
  Tri Dao authored Jun 02, 2022
  
  05087332
- Use Cutlass gemm as WarpMma · 14dc326e
  Tri Dao authored Jun 02, 2022
  
  14dc326e
- Remove old backward · e78e7c95
  Tri Dao authored Jun 02, 2022
  
  e78e7c95
- Add Cutlass as submodule · 512c98ee
  Tri Dao authored Jun 02, 2022
  
  512c98ee
- Rename src -> flash_attn · 5a61cb77
  Tri Dao authored Jun 01, 2022
  
  5a61cb77
- Support SM86 GPUs · c41479d6
  Tri Dao authored Jun 01, 2022
  
  c41479d6
26 May, 2022 1 commit
- Rename, add benchmarking script · 9dbc491a
  Tri Dao authored May 26, 2022
  
  9dbc491a
20 May, 2022 1 commit
- First release · 1fcbe6f0
  Tri Dao authored May 20, 2022
  
  1fcbe6f0