Releases: lucidrains/native-sparse-attention-pytorch
Releases · lucidrains/native-sparse-attention-pytorch
0.0.65
remove the get_at from inference
0.0.64
make it right for now, optimizer later
0.0.63
complete the overall idea for inference, polish up the edge cases on …
0.0.62
fix
0.0.61
start with compressed and sliding window for inference
0.0.60
precautionary
0.0.59
allow for block causal to be turned off for the triton kernel, preppi…
0.0.58
move query head group dimension to the right by one in forward triton…
0.0.57
move grouped query head dimension to the right by one and cleanup bac…
0.0.56
prepare for knocking out inference logic over the weekend, last commi…