vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

fix W_pack for-loop

Browse source
This commit is contained in:
jameswu2014 2023-09-11 12:05:35 +08:00
parent 39c4b8540d
commit 306d71be42
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
convert-baichuan-hf-to-gguf.py
View file

@ -232,7 +232,7 @@ for part_name in part_names:
model_part = torch.load(f"{dir_model}/{part_name}", map_location="cpu")
tmp=model_part
for i in itertools.count():
for i in range(block_count):
if f"model.layers.{i}.self_attn.W_pack.weight" in model_part:
print(f"Unpacking and permuting layer {i}")
tmp[f"model.layers.{i}.self_attn.q_proj.weight"]=reverse_hf_permute_part(model_part[f"model.layers.{i}.self_attn.W_pack.weight"],0,head_count,head_count)