Commits · ac3b684cdb2d3262a0dff2b7952ee1519ffcdc9c · gaoqiong / flash-attention

18 Apr, 2023 1 commit
- Have a separate nn.Dropout module in SelfAttention module · ac3b684c
  Tri Dao authored Apr 17, 2023
  
  ac3b684c
16 Apr, 2023 2 commits
- Bump to v1.0.2 · df1344f8
  Tri Dao authored Apr 15, 2023
  
  df1344f8
- Merge pull request #166 from ksivaman/enable_cuda_graph_capture · 635f159e
  Tri Dao authored Apr 16, 2023
```
Enable CUDA graph capture
```
  635f159e
15 Apr, 2023 3 commits
- only 1 thread writes to global mem in fprop · 45567a25
  Kirthi Shankar Sivamani authored Apr 15, 2023
```
Signed-off-by: Kirthi Shankar Sivamani <ksivamani@nvidia.com>
```
  45567a25
- Merge branch 'HazyResearch:main' into enable_cuda_graph_capture · a0997bc7
  Kirthi Shankar Sivamani authored Apr 14, 2023
  
  a0997bc7
- [Docs] Link to Forbes article · 221a39fd
  Tri Dao authored Apr 14, 2023
  
  221a39fd
14 Apr, 2023 4 commits
- [Gen] Fix FT kernel when using CG · 605655bc
  Tri Dao authored Apr 14, 2023
  
  605655bc
- Merge pull request #170 from CrustaceanJ/dependencies · dceb2687
  Tri Dao authored Apr 14, 2023
```
Missing module in `setup.py`
```
  dceb2687
- add missed module · 72629ac9
  Pavel Shvets authored Apr 14, 2023
  
  72629ac9
- Merge branch 'HazyResearch:main' into enable_cuda_graph_capture · 081c2b01
  Kirthi Shankar Sivamani authored Apr 13, 2023
  
  081c2b01
13 Apr, 2023 6 commits
- [Gen] Measure prompt processing + decoding time, not just decoding · 1c9ef9b3
  Tri Dao authored Apr 13, 2023
  
  1c9ef9b3
- [FusedDense] Enable sqrelu activation in FusedMLP · 6f6e9a9a
  Tri Dao authored Apr 13, 2023
  
  6f6e9a9a
- Handle FlashAttnQKVPackedSplitFunc by making rng_state optional in backward · 7d25a4ec
  Kirthi Shankar Sivamani authored Apr 13, 2023
```
Signed-off-by: Kirthi Shankar Sivamani <ksivamani@nvidia.com>
```
  7d25a4ec
- Merge branch 'HazyResearch:main' into enable_cuda_graph_capture · 315fd31f
  Kirthi Shankar Sivamani authored Apr 12, 2023
  
  315fd31f
- Merge pull request #164 from ZhiyuanChen/patch-1 · 5cee0714
  Tri Dao authored Apr 12, 2023
```
make mlp hidden_features defaults to 4*in_features
```
  5cee0714
- make mlp hidden_features defaults to 4*in_features · 8c424156
  Zhiyuan Chen authored Apr 13, 2023
  
  8c424156
12 Apr, 2023 3 commits
- Support CUDA graph capture · 31018c5f
  Kirthi Shankar Sivamani authored Apr 12, 2023
```
Signed-off-by: Kirthi Shankar Sivamani <ksivamani@nvidia.com>
```
  31018c5f
- Bump version to v1.0.1, fix Cutlass version · 853ff729
  Tri Dao authored Apr 12, 2023
  
  853ff729
- Bump version to 1.0.0 · 74af0233
  Tri Dao authored Apr 11, 2023
  
  74af0233
07 Apr, 2023 1 commit
- [FusedDense] Set workspace size to 32M for Hopper and 4M for others · dec4f2e9
  Tri Dao authored Apr 06, 2023
  
  dec4f2e9
04 Apr, 2023 2 commits
- Merge pull request #154 from kuizhiqing/usage · d478eeec
  Tri Dao authored Apr 04, 2023
```
add paddlepaddle in usage
```
  d478eeec
- add paddlepaddle in usage · c5be8d3a
  kuizhiqing authored Apr 04, 2023
  
  c5be8d3a
31 Mar, 2023 3 commits
- Merge pull request #147 from ksivaman/add_deterministic_execution_option · d6fc8605
  Tri Dao authored Mar 31, 2023
```
Add option for deterministic execution
```
  d6fc8605
- [LayerNorm] Implement LN with parallel residual, support dim 8k · 393882bc
  Tri Dao authored Mar 29, 2023
  
  393882bc
- Add option for deterministic execution · b6aa059b
  Kirthi Shankar Sivamani authored Mar 30, 2023
  
  b6aa059b
29 Mar, 2023 3 commits
- [Training] Fix lightning _PATH import · 009a3e71
  Tri Dao authored Mar 29, 2023
  
  009a3e71
- Implement GPT-NeoX · 993d1244
  Tri Dao authored Mar 29, 2023
  
  993d1244
- [FT] Fix FT's single query attention for bf16 hdim128 rotary · f5d0fbd4
  Tri Dao authored Mar 28, 2023
  
  f5d0fbd4
22 Mar, 2023 3 commits
- Implement GPT-J · 4d87e4d8
  Tri Dao authored Mar 22, 2023
  
  4d87e4d8
- [Triton] Fix benchmark_causal.py · 4360cfc6
  Tri Dao authored Mar 22, 2023
  
  4360cfc6
- [Triton] Fix benchmark_causal, mention Triton version · 5d079fdd
  Tri Dao authored Mar 22, 2023
  
  5d079fdd
15 Mar, 2023 4 commits
- Support H100 for other CUDA extensions · dc08ea1c
  Tri Dao authored Mar 15, 2023
  
  dc08ea1c
- Support H100 · 1b18f1b7
  Tri Dao authored Mar 15, 2023
  
  1b18f1b7
- Merge pull request #140 from VikParuchuri/main · 318e2f1b
  Tri Dao authored Mar 15, 2023
```
Remove unused kwargs like device in FlashAttention
```
  318e2f1b
- Remove unused kwargs in flashattention · 31653980
  Vik Paruchuri authored Mar 15, 2023
  
  31653980
14 Mar, 2023 1 commit
- [Rotary] Implement GPT-J style (interleaved) rotary · e45a46a5
  Tri Dao authored Mar 14, 2023
  
  e45a46a5
13 Mar, 2023 1 commit
- Update README on requirements (nvcc and Pytorch) · f28d61cb
  Tri Dao authored Mar 13, 2023
  
  f28d61cb
15 Feb, 2023 1 commit
- Merge pull request #94 from calebthomas259/main · 57ee6181
  Tri Dao authored Feb 14, 2023
```
Add a simple tutorial to README.md
```
  57ee6181
09 Feb, 2023 1 commit
- Update roadmap · 2dc2a195
  Tri Dao authored Feb 09, 2023
  
  2dc2a195
27 Jan, 2023 1 commit
- Merge pull request #110 from eltociear/patch-1 · 06da275b
  Tri Dao authored Jan 27, 2023
```
fix typo in default.yaml
```
  06da275b