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

fix #6391

Browse Source 
Former-commit-id: 067ba6e6cb4d8a1d95bba0a108f73008416a2865
This commit is contained in:
hiyouga
2024-12-19 12:16:30 +00:00
parent 0a465fc3ca
commit 0385c60177
6 changed files with 22 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
src/llamafactory/train/sft/trainer.py
View File

@@ -151,7 +151,7 @@ class CustomSeq2SeqTrainer(Seq2SeqTrainer):
return padded_tensor.contiguous() # in contiguous memory
def save_predictions(
self, dataset: "Dataset", predict_results: "PredictionOutput", gen_kwargs: Dict[str, Any]
self, dataset: "Dataset", predict_results: "PredictionOutput", skip_special_tokens: bool = True
) -> None:
r"""
Saves model predictions to `output_dir`.
@@ -179,12 +179,8 @@ class CustomSeq2SeqTrainer(Seq2SeqTrainer):
preds[i] = np.concatenate((preds[i][pad_len[0] :], preds[i][: pad_len[0]]), axis=-1)
decoded_inputs = self.processing_class.batch_decode(dataset["input_ids"], skip_special_tokens=False)
decoded_preds = self.processing_class.batch_decode(
preds, skip_special_tokens=gen_kwargs["skip_special_tokens"]
)
decoded_labels = self.processing_class.batch_decode(
labels, skip_special_tokens=gen_kwargs["skip_special_tokens"]
)
decoded_preds = self.processing_class.batch_decode(preds, skip_special_tokens=skip_special_tokens)
decoded_labels = self.processing_class.batch_decode(labels, skip_special_tokens=skip_special_tokens)
with open(output_prediction_file, "w", encoding="utf-8") as f:
for text, pred, label in zip(decoded_inputs, decoded_preds, decoded_labels):

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

@@ -91,7 +91,7 @@ def run_sft(
)
# Keyword arguments for `model.generate`
gen_kwargs = generating_args.to_dict()
gen_kwargs = generating_args.to_dict(obey_generation_config=True)
gen_kwargs["eos_token_id"] = [tokenizer.eos_token_id] + tokenizer.additional_special_tokens_ids
gen_kwargs["pad_token_id"] = tokenizer.pad_token_id
gen_kwargs["logits_processor"] = get_logits_processor()
@@ -130,7 +130,7 @@ def run_sft(
predict_results.metrics.pop("predict_loss", None)
trainer.log_metrics("predict", predict_results.metrics)
trainer.save_metrics("predict", predict_results.metrics)
trainer.save_predictions(dataset_module["eval_dataset"], predict_results, gen_kwargs)
trainer.save_predictions(dataset_module["eval_dataset"], predict_results, generating_args.skip_special_tokens)
# Create model card
create_modelcard_and_push(trainer, model_args, data_args, training_args, finetuning_args)