Infllm V2 Dense Sparse Switchable Attention

Implement dense-sparse switchable attention enabling LLMs to scale from short to long sequences with 4× speedup and 98-99.7% performance retention, requiring no extra parameters by reusing pretrained attention weights through trainable sparse pattern selection.

adu2021 Updated 7 repo stars

File contents

adu2021/skillxiv commit dcaeba5555

Frequently asked questions

npx get-skill add adu2021/infllm-v2-dense-sparse-switchable-attention