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

gguf-py : fix and simplify quantized shape round-trip

Browse source
This commit is contained in:
Francis Couture-Harpin 2024-05-22 23:40:41 -04:00
parent cd93a28cb1
commit 2ff601fc32
5 changed files with 27 additions and 13 deletions
Download patch file Download diff file Expand all files Collapse all files

4
gguf-py/scripts/gguf-new-metadata.py
View file

@ -118,9 +118,7 @@ def copy_with_new_metadata(reader: gguf.GGUFReader, writer: gguf.GGUFWriter, new
for tensor in reader.tensors:
total_bytes += tensor.n_bytes
# Dimensions are written in reverse order, so flip them first
shape = np.flipud(tensor.shape).tolist()
writer.add_tensor_info(tensor.name, shape, tensor.data.dtype, tensor.data.nbytes, tensor.tensor_type)
writer.add_tensor_info(tensor.name, tensor.data.shape, tensor.data.dtype, tensor.data.nbytes, tensor.tensor_type)
bar = tqdm(desc="Writing", total=total_bytes, unit="byte", unit_scale=True)