423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-17 12:20:37 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix generation in seq2seq.py

Browse Source
This commit is contained in:
hiyouga
2023-06-26 18:07:06 +08:00
parent cec9760eb8
commit 1175948029
2 changed files with 10 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
src/train_sft.py
View File

@@ -25,7 +25,10 @@ def main():
dataset = prepare_data(model_args, data_args)
model, tokenizer = load_pretrained(model_args, finetuning_args, training_args.do_train, stage="sft")
dataset = preprocess_data(dataset, tokenizer, data_args, training_args, stage="sft")
data_collator = DynamicDataCollatorWithPadding(tokenizer, data_args.ignore_pad_token_for_loss)
data_collator = DynamicDataCollatorWithPadding(
tokenizer=tokenizer,
ignore_pad_token_for_loss=(data_args.ignore_pad_token_for_loss and not training_args.predict_with_generate)
)
# Override the decoding parameters of Seq2SeqTrainer
training_args.generation_max_length = training_args.generation_max_length if \