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-Chen 2024-01-23 01:44:10 +11:00
commit 9a5062e6c4
1 changed files with 4 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

6
unsloth/models/llama.py
View file

@ -144,8 +144,10 @@ def LlamaAttention_fast_forward_inference(
Kn *= cos; Kn.addcmul_(RH_K, sin);
# New KV cache
Kn = torch.cat([K1, Kn], dim = 2)
Vn = torch.cat([V1, Vn], dim = 2)
# Kn = torch.cat([K1, Kn], dim = 2)
# Vn = torch.cat([V1, Vn], dim = 2)
Kn = torch.cat([K1, Kn], dim = 1)
Vn = torch.cat([V1, Vn], dim = 1)
# Grouped query attention
if n_groups != 1: