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

Remove uint8 branch in gguf_writer.

Browse source
This commit is contained in:
Qingtao Li 2024-10-11 14:43:14 +08:00
parent f673699460
commit dfac0c4b3e
1 changed files with 1 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

4
gguf-py/gguf/gguf_writer.py
View file

@ -345,10 +345,8 @@ class GGUFWriter:
dtype = GGMLQuantizationType.I32 dtype = GGMLQuantizationType.I32
elif tensor_dtype == np.int64: elif tensor_dtype == np.int64:
dtype = GGMLQuantizationType.I64 dtype = GGMLQuantizationType.I64
elif tensor_dtype == np.uint8:
dtype = GGMLQuantizationType.I2
else: else:
raise ValueError("Only F16, F32, F64, I8, I16, I32, I64, I2 tensors are supported for now") raise ValueError("Only F16, F32, F64, I8, I16, I32, I64 tensors are supported for now")
else: else:
dtype = raw_dtype dtype = raw_dtype
if tensor_dtype == np.uint8: if tensor_dtype == np.uint8: