ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Update llama.py

Browse source
This commit is contained in:
Daniel Han 2025-02-02 02:14:13 -08:00
commit 11e43e9639
1 changed files with 0 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

1
unsloth/models/llama.py
View file

@ -254,7 +254,6 @@ def LlamaAttention_fast_forward_inference(
# pass
# Attention
print(attention_mask)
if bsz == 1:
Qn *= self.scalar # See https://github.com/ggerganov/llama.cpp/issues/7805#issuecomment-2153349963
# It seems like doing (Q * scalar) @ K is better than (Q @ K) * scalar to stop overflows