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

fix baichuan1-13b W_pack convert bug

Browse source
This commit is contained in:
jameswu2014 2023-09-05 17:34:36 +08:00
parent 37873ae77a
commit 9fd5aa7ecb
1 changed files with 0 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

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

@ -239,8 +239,6 @@ for part_name in part_names:
tmp[f"model.layers.{i}.self_attn.k_proj.weight"]=reverse_hf_permute_part(model_part[f"model.layers.{i}.self_attn.W_pack.weight"],1,head_count,head_count_kv) tmp[f"model.layers.{i}.self_attn.k_proj.weight"]=reverse_hf_permute_part(model_part[f"model.layers.{i}.self_attn.W_pack.weight"],1,head_count,head_count_kv)
tmp[f"model.layers.{i}.self_attn.v_proj.weight"]=reverse_hf_part(model_part[f"model.layers.{i}.self_attn.W_pack.weight"],2) tmp[f"model.layers.{i}.self_attn.v_proj.weight"]=reverse_hf_part(model_part[f"model.layers.{i}.self_attn.W_pack.weight"],2)
del tmp[f"model.layers.{i}.self_attn.W_pack.weight"] del tmp[f"model.layers.{i}.self_attn.W_pack.weight"]
else:
break
for name in model_part.keys(): for name in model_part.keys():
data = model_part[name] data = model_part[name]