423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-08-23 06:12:50 +08:00
Code Issues Packages Projects Releases Wiki Activity

Update tuner.py

Browse Source 
Former-commit-id: 13851fb04524e3a599b6c07d749f7463b8f75319
This commit is contained in:
hoshi-hiyouga 2024-05-11 23:54:53 +08:00 committed by GitHub
parent 1dd674697c
commit 17975fefd7
1 changed files with 12 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
src/llmtuner/train/tuner.py
View File

@ -15,9 +15,11 @@ from .pt import run_pt
from .rm import run_rm from .rm import run_rm
from .sft import run_sft from .sft import run_sft
if TYPE_CHECKING: if TYPE_CHECKING:
from transformers import TrainerCallback from transformers import TrainerCallback
logger = get_logger(__name__) logger = get_logger(__name__)
@ -51,8 +53,8 @@ def export_model(args: Optional[Dict[str, Any]] = None) -> None:
raise ValueError("Please merge adapters before quantizing the model.") raise ValueError("Please merge adapters before quantizing the model.")
tokenizer_module = load_tokenizer(model_args) tokenizer_module = load_tokenizer(model_args)
tokenizer = tokenizer_module['tokenizer'] tokenizer = tokenizer_module["tokenizer"]
processor = tokenizer_module['processor'] processor = tokenizer_module["processor"]
get_template_and_fix_tokenizer(tokenizer, data_args.template) get_template_and_fix_tokenizer(tokenizer, data_args.template)
model = load_model(tokenizer, model_args, finetuning_args) # must after fixing tokenizer to resize vocab model = load_model(tokenizer, model_args, finetuning_args) # must after fixing tokenizer to resize vocab
@ -63,7 +65,7 @@ def export_model(args: Optional[Dict[str, Any]] = None) -> None:
raise ValueError("The model is not a `PreTrainedModel`, export aborted.") raise ValueError("The model is not a `PreTrainedModel`, export aborted.")
if getattr(model, "quantization_method", None) is None: # cannot convert dtype of a quantized model if getattr(model, "quantization_method", None) is None: # cannot convert dtype of a quantized model
output_dtype = getattr(model.config, "torch_dtype", torch.float16) output_dtype = torch.float16
setattr(model.config, "torch_dtype", output_dtype) setattr(model.config, "torch_dtype", output_dtype)
model = model.to(output_dtype) model = model.to(output_dtype)
@ -86,10 +88,12 @@ def export_model(args: Optional[Dict[str, Any]] = None) -> None:
tokenizer.save_pretrained(model_args.export_dir) tokenizer.save_pretrained(model_args.export_dir)
if model_args.export_hub_model_id is not None: if model_args.export_hub_model_id is not None:
tokenizer.push_to_hub(model_args.export_hub_model_id, token=model_args.hf_hub_token) tokenizer.push_to_hub(model_args.export_hub_model_id, token=model_args.hf_hub_token)
if model_args.visual_inputs and processor is not None:
getattr(processor, "image_processor").save_pretrained(model_args.export_dir)
if model_args.export_hub_model_id is not None:
getattr(processor, "image_processor").push_to_hub(
model_args.export_hub_model_id, token=model_args.hf_hub_token
)
except Exception: except Exception:
logger.warning("Cannot save tokenizer, please copy the files manually.") logger.warning("Cannot save tokenizer, please copy the files manually.")
if model_args.visual_inputs:
processor.image_processor.save_pretrained(model_args.export_dir)
if model_args.export_hub_model_id is not None:
processor.image_processor.push_to_hub(model_args.export_hub_model_id, token=model_args.hf_hub_token)