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 02:18:37 +11:00
commit a77c448939
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
unsloth/models/llama.py
View file

@ -160,8 +160,8 @@ def LlamaAttention_fast_forward_inference(
# Knn = Knn.reshape(1, n_heads, cached_len, head_dim)
# Vnn = Vnn.reshape(1, n_heads, cached_len, head_dim)
new_seq_len = seq_len + 1
Knn = Kn[:, :, None, :].expand(n_kv_heads, n_groups, new_seq_len, head_dim)
Vnn = Vn[:, :, None, :].expand(n_kv_heads, n_groups, new_seq_len, head_dim)
Knn = Kn[:, None, :, :].expand(n_kv_heads, n_groups, new_seq_len, head_dim)
Vnn = Vn[:, None, :, :].expand(n_kv_heads, n_groups, new_seq_len, head_dim)
Knn = Knn.reshape(n_heads, new_seq_len, head_dim)
Vnn = Vnn.reshape(n_heads, new_seq_len, head_dim)
else: