Commits · 08124c8f9cf88ba327e2c455ed2a302979f06c91 · gaoqiong / flash-attention

17 Dec, 2023 1 commit
- [CrossEntropy] Implement logit_scale option · 08124c8f
  Tri Dao authored Dec 16, 2023
  
  08124c8f
01 Dec, 2023 1 commit
- [LayerNorm] Implement layer_norm_linear · 9356a1c0
  Tri Dao authored Nov 30, 2023
  
  9356a1c0
28 Nov, 2023 2 commits
- Bump to v2.3.6 · 92dd5703
  Tri Dao authored Nov 27, 2023
  
  92dd5703
- [CI] Only compile for CUDA 11.8 & 12.2, MAX_JOBS=2,add torch-nightly · d4a7c8ff
  Tri Dao authored Nov 27, 2023
  
  d4a7c8ff
27 Nov, 2023 3 commits
- Allow varlen_fwd to take optional seqused_k (#647) · ce3e7280
  Jeremy Reizenstein authored Nov 27, 2023
```
Co-authored-by: bottler <bottler@users.noreply.github.com>
```
  ce3e7280
- Bump to v2.3.5 · 23b77c81
  Tri Dao authored Nov 26, 2023
  
  23b77c81
- Fix performance regression with causal · b4bf9cc1
  Tri Dao authored Nov 26, 2023
  
  b4bf9cc1
20 Nov, 2023 6 commits
- Bump to v2.3.4 · 2c3baba4
  Tri Dao authored Nov 19, 2023
  
  2c3baba4
- [CrossEntropy] Simplify the case of large vocab with Tensor Parallel · aaa14741
  Tri Dao authored Nov 19, 2023
  
  aaa14741
- fix flash ce mp large vocab (#673) · abf04a56
  Shijie authored Nov 20, 2023
  
  abf04a56
- Write zero to out / grad if seqlen_q or seqlen_k is zero · db2f8069
  Tri Dao authored Nov 19, 2023
  
  db2f8069
- Update cutlass to 3.2.2 · 43bb6d8a
  Tri Dao authored Nov 19, 2023
  
  43bb6d8a
- add checks (#640) · dc4b9ad6
  Driss Guessous authored Nov 19, 2023
  
  dc4b9ad6
14 Nov, 2023 1 commit
- [LayerNorm] Add postnorm residual + LayerNorm/RMSNorm in Triton · 01771645
  Tri Dao authored Nov 13, 2023
  
  01771645
13 Nov, 2023 1 commit
- [LayerNorm] Implement residual + LayerNorm/RMSNorm in Triton · 79bd1a2d
  Tri Dao authored Nov 13, 2023
  
  79bd1a2d
09 Nov, 2023 1 commit
- Fix typo in RotaryEmbedding forward output type (#666) · 3566596a
  Antony Frolov authored Nov 09, 2023
  
  3566596a
24 Oct, 2023 2 commits
- Bump to v2.3.3 · 83aef842
  Tri Dao authored Oct 24, 2023
  
  83aef842
- [CrossEntropy] Fix triton cross_entropy_loss IMA for >=2B elements · c79de85f
  Tri Dao authored Oct 24, 2023
  
  c79de85f
12 Oct, 2023 1 commit
- Clarify inference README is a placeholder · 02ac572f
  Tri Dao authored Oct 12, 2023
  
  02ac572f
09 Oct, 2023 1 commit
- Bump to v2.3.2 · 7f31e7c1
  Tri Dao authored Oct 08, 2023
  
  7f31e7c1
08 Oct, 2023 1 commit
- Change constexpr int to constexpr static int · 5a834254
  Tri Dao authored Oct 08, 2023
  
  5a834254
05 Oct, 2023 2 commits
- Add change log · 3a9fe7b0
  Tri Dao authored Oct 05, 2023
  
  3a9fe7b0
- Clarify that Windows is not supported right now · aa4fd2d1
  Tri Dao authored Oct 05, 2023
  
  aa4fd2d1
04 Oct, 2023 2 commits
- [CI] Use official Pytorch 2.1, add CUDA 11.8 for Pytorch 2.1 · 5e525a8d
  Tri Dao authored Oct 03, 2023
  
  5e525a8d
- Bump to v2.3.1 · 21c3b0d8
  Tri Dao authored Oct 03, 2023
  
  21c3b0d8
03 Oct, 2023 1 commit
- [Gen] Accept cache_batch_idx to index into the KV cache · e279bf8e
  Tri Dao authored Oct 03, 2023
  
  e279bf8e
27 Sep, 2023 1 commit
- Bump to v2.3.0 · 601b4dc4
  Tri Dao authored Sep 26, 2023
  
  601b4dc4
26 Sep, 2023 1 commit
- Implement local attention · 083e8f52
  Tri Dao authored Sep 24, 2023
```
Co-authored-by: Timothee Lacroix <t@mistral.ai>
```
  083e8f52
25 Sep, 2023 1 commit
- Fix NameError and typo in ApplyRotaryEmbQKV_ (#569) · 4c8ff915
  Katherine Crowson authored Sep 25, 2023
  
  4c8ff915
24 Sep, 2023 4 commits
- Bump to v2.2.5 · 0a1d03c7
  Tri Dao authored Sep 24, 2023
  
  0a1d03c7
- Switch cutlass to newer commit to avoid compilation warning · 812cb1c9
  Tri Dao authored Sep 24, 2023
  
  812cb1c9
- Don't over-allocate dq_accum in case of varlen · 65c234ed
  Tri Dao authored Sep 24, 2023
  
  65c234ed
- Reduce number of templates for headdim > 128 · 1879e089
  Tri Dao authored Sep 23, 2023
  
  1879e089
22 Sep, 2023 2 commits
- Add placeholder for inference example · dd9a6fa4
  Tri Dao authored Sep 22, 2023
  
  dd9a6fa4
- Re-enable compilation for Hopper · bff31471
  Tri Dao authored Sep 21, 2023
  
  bff31471
21 Sep, 2023 4 commits
- Fix E1136 (#563) · 187c2a06
  Yuchao Dai authored Sep 22, 2023
  
  187c2a06
- Bump to v2.2.4 · 229080b9
  Tri Dao authored Sep 20, 2023
  
  229080b9
- Swap seqlen_q and ngroups when seqlen_q=1 (h/t Daniel Haziza) · 2d8ea9a5
  Tri Dao authored Sep 20, 2023
  
  2d8ea9a5
- [Llama] Fix some tests, add tests for Llama 2 and CodeLlama · 0705d271
  Tri Dao authored Sep 20, 2023
  
  0705d271
20 Sep, 2023 1 commit
- [Gen] Simplify decode_speculative · e0fbaa70
  Tri Dao authored Sep 18, 2023
  
  e0fbaa70