memory-efficient-attention icon indicating copy to clipboard operation
memory-efficient-attention copied to clipboard

Published 20 hours ago • verified publisher icon AminRezaei0x443
→

Metadata

Memory Efficient Attention (O(sqrt(n)) for Jax and PyTorch

Results 1 memory-efficient-attention issues
Sort by recently updated
recently updated
newest added

Improve performance via batched-matmul and fused multiplies

11 comment

Many thanks for providing this reference implementation. I tried integrating this into stable-diffusion / diffusers. A fix was required to make it work on Mac (PyTorch MPS backend): https://github.com/Birch-san/diffusers/pull/1/commits/04372140a25d7f53549175f1f196599c3e9bf3a5 Knowing...

Birch-san avatar Birch-san

← Metadata

121
Stars
12
Forks
Watchers

Owner

verified publisher icon AminRezaei0x443

Metadata

Memory Efficient Attention (O(sqrt(n)) for Jax and PyTorch

Back