Make flash attention configurable in user code

#26

by YenChunChen - opened May 30, 2024

←

•

With this PR, users can specify whether to enable flash attention 2 in from_pretrain.

YenChunChen changed pull request status to open May 30, 2024

@YenChunChen default should be flash_attention in readme, user can specify to use eager if they want

@haipingwu updated default to flash attention

hi @YenChunChen , please reset config.json to original as well

done

leoxiaobin changed pull request status to merged May 31, 2024

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment