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

convert_hf : move add_type to main()

Browse source 
* convert_lora : use the GGUFWriter from Model instead of overwriting it
This commit is contained in:
Francis Couture-Harpin 2024-07-15 11:39:42 -04:00
parent 4d9ac0f375
commit d09382fac7
2 changed files with 1 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

2
convert_hf_to_gguf.py
View file

@ -99,7 +99,6 @@ class Model:
self.fname_out = fname_out.parent / fname_out.name.format(ftype_lw, outtype=ftype_lw, ftype=ftype_lw, OUTTYPE=ftype_up, FTYPE=ftype_up)
self.gguf_writer = gguf.GGUFWriter(path=None, arch=gguf.MODEL_ARCH_NAMES[self.model_arch], endianess=self.endianess, use_temp_file=self.use_temp_file,
split_max_tensors=split_max_tensors, split_max_size=split_max_size, dry_run=dry_run, small_first_shard=small_first_shard)
self.gguf_writer.add_type(gguf.GGUFType.MODEL)
@classmethod
def __init_subclass__(cls):
@ -3575,6 +3574,7 @@ def main() -> None:
small_first_shard=args.no_tensor_first_split)
logger.info("Set model parameters")
model_instance.gguf_writer.add_type(gguf.GGUFType.MODEL)
model_instance.set_gguf_parameters()
logger.info("Set model tokenizer")

1
convert_lora_to_gguf.py
View file

@ -365,7 +365,6 @@ if __name__ == '__main__':
alpha = lparams["lora_alpha"]
model_instance.gguf_writer = gguf.GGUFWriter(path=None, arch=gguf.MODEL_ARCH_NAMES[model_instance.model_arch])
model_instance.gguf_writer.add_string(gguf.Keys.General.TYPE, gguf.GGUFType.ADAPTER)
model_instance.gguf_writer.add_string(gguf.Keys.Adapter.TYPE, "lora")
model_instance.gguf_writer.add_float32(gguf.Keys.Adapter.LORA_ALPHA, float(alpha))