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

[llama.cpp] update convert_hf_to_gguf.py

Browse source
This commit is contained in:
kalineid 2024-10-14 17:31:48 +08:00 • committed by Qingtao Li
parent 6bb4acae7c
commit b266290700
1 changed files with 3 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

6
convert_hf_to_gguf.py
View file

@ -331,10 +331,10 @@ class Model:
data_torch = torch.from_numpy(data.reshape(old_shape))
if self.ftype == gguf.LlamaFileType.MOSTLY_F16:
data_torch = data_torch.to(torch.float16)
else:
return self.modify_tensors(data_torch, name, bid)
return [(self.map_tensor_name(name), data_torch)]
return [(self.map_tensor_name(name), data_torch)]
return self.modify_tensors(data_torch, name, bid)
def tensor_force_quant(self, name: str, new_name: str, bid: int | None, n_dims: int) -> gguf.GGMLQuantizationType | bool:
del name, new_name, bid, n_dims # unused