p04798526
/
LLaMA-Factory-Mirror
1
0
Fork 1
Code Issues Pull Requests Packages Projects Releases Wiki Activity

lint

This commit is contained in:
hiyouga 2024-06-17 22:35:56 +08:00
parent 7857c0990b
commit 24c160df3d
2 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
scripts/pissa_init.py
View File

@ -50,7 +50,7 @@ def quantize_pissa(
lora_alpha=lora_alpha if lora_alpha is not None else lora_rank * 2, lora_alpha=lora_alpha if lora_alpha is not None else lora_rank * 2,
lora_dropout=lora_dropout, lora_dropout=lora_dropout,
target_modules=[name.strip() for name in lora_target.split(",")], target_modules=[name.strip() for name in lora_target.split(",")],
init_lora_weights="pissa" if pissa_iter == -1 else "pissa_niter_{}".format(pissa_iter) init_lora_weights="pissa" if pissa_iter == -1 else "pissa_niter_{}".format(pissa_iter),
) )
# Init PiSSA model # Init PiSSA model

2
src/llamafactory/hparams/finetuning_args.py
View File

@ -352,7 +352,7 @@ class FinetuningArguments(FreezeArguments, LoraArguments, RLHFArguments, GaloreA
self.additional_target: Optional[List[str]] = split_arg(self.additional_target) self.additional_target: Optional[List[str]] = split_arg(self.additional_target)
self.galore_target: List[str] = split_arg(self.galore_target) self.galore_target: List[str] = split_arg(self.galore_target)
self.freeze_vision_tower = self.freeze_vision_tower or self.train_mm_proj_only self.freeze_vision_tower = self.freeze_vision_tower or self.train_mm_proj_only
self.use_ref_model = (self.stage == "dpo" and self.pref_loss not in ["orpo", "simpo"]) self.use_ref_model = self.stage == "dpo" and self.pref_loss not in ["orpo", "simpo"]
assert self.finetuning_type in ["lora", "freeze", "full"], "Invalid fine-tuning method." assert self.finetuning_type in ["lora", "freeze", "full"], "Invalid fine-tuning method."
assert self.ref_model_quantization_bit in [None, 8, 4], "We only accept 4-bit or 8-bit quantization." assert self.ref_model_quantization_bit in [None, 8, 4], "We only accept 4-bit or 8-bit quantization."