huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-09-11 04:14:52 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fix window_size_left for flash attention v1

Browse Source 
This fixes flash attention v1 by not raising NotImplementedError when window_size_left is left at its default of -1
This commit is contained in:
Peter Lowrance 2023-10-02 11:46:04 -04:00 committed by GitHub
parent 5ba53d44a1
commit 891b18a0c3
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
server/text_generation_server/utils/flash_attn.py
View File

@ -80,7 +80,7 @@ def attention(
) )
if HAS_FLASH_ATTN: if HAS_FLASH_ATTN:
if window_size_left != 0: if window_size_left != -1:
raise NotImplementedError( raise NotImplementedError(
"window_size_left is only available with flash attn v2" "window_size_left is only available with flash attn v2"
) )