423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-10-16 00:28:10 +08:00
Code Issues Packages Projects Releases Wiki Activity

set use_cache before saving model

Browse Source 
Former-commit-id: a6a0161f32f600f3001188ff4c7929c5f13c2a03
This commit is contained in:
hiyouga 2023-07-04 23:18:20 +08:00
parent 6df5c4ccef
commit e4e36a2d74
1 changed files with 2 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/utils/peft_trainer.py
View File

@ -105,11 +105,13 @@ class PeftTrainer(Seq2SeqTrainer):
if self.finetuning_args.finetuning_type == "lora": if self.finetuning_args.finetuning_type == "lora":
backbone_model.save_pretrained(output_dir, state_dict=get_state_dict(backbone_model)) backbone_model.save_pretrained(output_dir, state_dict=get_state_dict(backbone_model))
else: # freeze/full tuning else: # freeze/full tuning
backbone_model.config.use_cache = True
backbone_model.save_pretrained( backbone_model.save_pretrained(
output_dir, output_dir,
state_dict=get_state_dict(backbone_model), state_dict=get_state_dict(backbone_model),
safe_serialization=self.args.save_safetensors safe_serialization=self.args.save_safetensors
) )
backbone_model.config.use_cache = False
if self.tokenizer is not None: if self.tokenizer is not None:
self.tokenizer.save_pretrained(output_dir) self.tokenizer.save_pretrained(output_dir)