[FEATURE] Overtrain/overfit limiters

[MarcoRavich]

Hi there,
I'm honestly not sure if these features have already been implemented in NAM (if so, suddently close this issue, sorry), but it would be extremely useful to have a more efficient learning - as @IAHispano managed to do in their Applio for example - if possible.

Some interesting articles about:

• How to Guide: Overcoming overfitting in your ML models;
• A Gentle Introduction to Early Stopping to Avoid Overtraining Neural Networks;
• Overfitting in Machine Learning: What It Is and How to Prevent It;
• ...

Hope that inspires !

[sdatkinson]

Hi there.

The trainer already uses checkpointing. See here:

• neural-amp-modeler/nam/train/core.py

  Line 1236 in 619fb4d

  def _get_callbacks(

• neural-amp-modeler/nam/train/core.py

  Line 1431 in 619fb4d

  best_checkpoint = trainer.checkpoint_callback.best_model_path