- 20 Nov, 2023 3 commits
-
-
Tri Dao authored
-
Tri Dao authored
-
Driss Guessous authored
-
- 08 Oct, 2023 1 commit
-
-
Tri Dao authored
-
- 03 Oct, 2023 1 commit
-
-
Tri Dao authored
-
- 26 Sep, 2023 1 commit
-
-
Tri Dao authored
Co-authored-by:Timothee Lacroix <t@mistral.ai>
-
- 24 Sep, 2023 3 commits
- 21 Sep, 2023 1 commit
-
-
Tri Dao authored
-
- 18 Sep, 2023 3 commits
- 17 Sep, 2023 1 commit
-
-
Tri Dao authored
-
- 16 Sep, 2023 2 commits
- 13 Sep, 2023 1 commit
-
-
Tri Dao authored
-
- 12 Sep, 2023 1 commit
-
-
Tri Dao authored
-
- 11 Sep, 2023 1 commit
-
-
Tri Dao authored
-
- 04 Sep, 2023 5 commits
- 03 Sep, 2023 1 commit
-
-
Tri Dao authored
-
- 01 Sep, 2023 2 commits
-
-
Sophia Wisdom authored
* Remove lots of comments * Remove unused traits
-
Sophia Wisdom authored
-
- 30 Aug, 2023 2 commits
-
-
Aman Gupta Karmani authored
-
Tri Dao authored
-
- 29 Aug, 2023 1 commit
-
-
Tri Dao authored
-
- 28 Aug, 2023 3 commits
-
-
Tri Dao authored
-
dan_the_3rd authored
When seqlen=8136, `smem_sz = 48840`, and apparently starting the kernel returns an `invalid argument` CUDA error. `48840 < 48 * 1024` but apparently it's still above the limit somehow..? Tested on A100
-
Tri Dao authored
-
- 25 Aug, 2023 1 commit
-
-
Tri Dao authored
-
- 24 Aug, 2023 1 commit
-
-
BoxiangW authored
Support flash attention 2 with causal masking when KV's seq length is longer than Q's seq length. (#436)
-
- 17 Aug, 2023 1 commit
-
-
Tri Dao authored
-
- 16 Aug, 2023 1 commit
-
-
Tri Dao authored
-
- 13 Aug, 2023 2 commits
- 01 Aug, 2023 1 commit
-
-
Tri Dao authored
-
