423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-17 20:30:36 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix #6499

Browse Source
This commit is contained in:
hiyouga
2025-01-02 11:17:29 +00:00
parent f8e80d566f
commit 1800f8c72d
5 changed files with 17 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files

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

@@ -117,8 +117,6 @@ def run_sft(
# Evaluation
if training_args.do_eval:
metrics = trainer.evaluate(metric_key_prefix="eval", **gen_kwargs)
if training_args.predict_with_generate: # eval_loss will be wrong if predict_with_generate is enabled
metrics.pop("eval_loss", None)
trainer.log_metrics("eval", metrics)
trainer.save_metrics("eval", metrics)
@@ -126,8 +124,6 @@ def run_sft(
if training_args.do_predict:
logger.warning_once("Batch generation can be very slow. Consider using `scripts/vllm_infer.py` instead.")
predict_results = trainer.predict(dataset_module["eval_dataset"], metric_key_prefix="predict", **gen_kwargs)
if training_args.predict_with_generate: # predict_loss will be wrong if predict_with_generate is enabled
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, generating_args.skip_special_tokens)