The default model parameter for training is different from the pretrained checkpoint about skipping-the-frame-level

[Yujia-Yan]

Hi,
Thanks for your interest.

1. I do not see those parameter matters, as far as I know. But this view may be biased as a larger range may accommodate more harmonics for seldomly played high notes.
2. a. The curve looks under-fitting. The number of training steps should be increased accordingly. The code is using an aggressive learning rate scheduler (OneCircle) which requires the total number of iterations known beforehand. It's probably better to switch to a constant learning rate scheduler to train longer (e.g. the famous default 1e-4 learning rate) in practice if we don't know how much steps are enough.
   b. The architecture uses batch norm, which is also notorious for small batch training. That's why I have not even tried small batch like that.
   BTW, the spikes in train_f1 and others suggest you may have some issue in dataset. Have you checked the sampling rate of the data?

[seyong92]

Thank you for the fast comment! I will change the learning rate and share the results after training.

Also, I found that some of the files are not correctly resampled to 44100, as you said. Thanks!