kyy/llm_trainer
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

support mllm hf inference

Browse Source
This commit is contained in:
hiyouga
2024-04-26 05:34:58 +08:00
parent c20f750d11
commit e057c8de48
27 changed files with 130 additions and 51 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/llmtuner/train/sft/workflow.py
View File

@@ -29,9 +29,9 @@ def run_sft(
callbacks: Optional[List["TrainerCallback"]] = None,
):
tokenizer_module = load_tokenizer(model_args)
dataset = get_dataset(model_args, data_args, training_args, stage="sft", **tokenizer_module)
tokenizer = tokenizer_module["tokenizer"]
model = load_model(tokenizer, model_args, finetuning_args, is_trainable=training_args.do_train)
dataset = get_dataset(model_args, data_args, training_args, stage="sft", **tokenizer_module)
model = load_model(tokenizer, model_args, finetuning_args, training_args.do_train)
if training_args.predict_with_generate:
tokenizer.padding_side = "left" # use left-padding in generation