Streambp Efficient Backprop

Enables 2.8-5.5x longer sequences during LLM training via linear decomposition of chain rule along sequence dimension, maintaining exact gradients with lower memory cost.

adu2021 Updated 7 repo stars

File contents

adu2021/skillxiv commit 901f07edc2

Frequently asked questions

npx get-skill add adu2021/streambp-efficient-backprop