mmarch/text-generation-webui
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Improve error handling while loading GPTQ models

Browse source
This commit is contained in:
oobabooga 2023-05-19 11:20:08 -03:00
parent 39dab18307
commit 9d5025f531
2 changed files with 5 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

2
modules/GPTQ_loader.py
View file

@ -140,7 +140,7 @@ def load_quantized(model_name):
if shared.args.model_type is None: if shared.args.model_type is None:
logging.error("The model could not be loaded because its type could not be inferred from its name.") logging.error("The model could not be loaded because its type could not be inferred from its name.")
logging.error("Please specify the type manually using the --model_type argument.") logging.error("Please specify the type manually using the --model_type argument.")
return return None
# Select the appropriate load_quant function # Select the appropriate load_quant function
model_type = shared.args.model_type.lower() model_type = shared.args.model_type.lower()

5
modules/models.py
View file

@ -97,7 +97,10 @@ def load_model(model_name):
model, tokenizer = output model, tokenizer = output
else: else:
model = output model = output
tokenizer = load_tokenizer(model_name, model) if model is None:
return None, None
else:
tokenizer = load_tokenizer(model_name, model)
# Hijack attention with xformers # Hijack attention with xformers
if any((shared.args.xformers, shared.args.sdp_attention)): if any((shared.args.xformers, shared.args.sdp_attention)):