AminRezaei0x443/memory-efficient-attention

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

PythonMIT

Issues

Improve performance via batched-matmul and fused multiplies
#7 opened 2 years ago by Birch-san
10
Feature Request: Drop In Replacement for torch.nn.MultiheadAttention
#3 opened 3 years ago by SamuelGabriel
1
feat: output_attentions
#1 opened 3 years ago by xloem
4