Releases: Dao-AILab/flash-attention
Releases Β· Dao-AILab/flash-attention
v2.1.2.post3
Set single threaded compilation for CUDA 12.2 so CI doesn't OOM
v2.1.2.post2
Remove constexpr in launch template to fix CI compilation
v2.1.2.post1
Try switching back to Cutlass 3.2.0
v2.1.2
Bump to v2.1.2
v2.1.1
Update Cutlass to v3.2.0
v2.1.0
Change causal mask to be aligned to bottom-right instead of top-left
v2.0.9
Bump version to 2.0.9
v2.0.8
Bump to v2.0.8
v2.0.7
Bump to v2.0.7
v2.0.6.post2
[CI] Fix MATRIX_CUDA_VERSION check