Closed Oliver0621 closed 4 months ago
This looks like an error caused by missing data in the dataloader, in other words, there is too little data. In fact, SOFA currently does not recommend training with such a small amount of data, as it is very likely to cause overfitting. The next version of SOFA may support fine-tuning with a small amount of data before inference, but in the current version, we advise users to prepare sufficient data before training.
OK, how long is the appropriate amount of data required? 30 minutes of audio in total? Or a total of 1 hour of audio?
About 30 minutes.
I have added the full_label, weak_label and no_label data normally. And binarize.py was successfully executed. The following is the result.
I have not modified the train_config.yaml configuration file.
I'm going to use a pretrained model for fine-tuning. The following result appeared, and then it paused. What is the problem?