Train a contextual spell checker with new dataset

[lhlong]

Hello, I'm following this tutorial to train my spell checker. It's very clear and easy to train the sample model with new dataset.

However, I trouble some issues while training with my dataset (~300k samples):

1. How to save checkpoints automatically after some steps/epochs?
2. How can we evaluate this retrain model?
3. What is the confidence value of the result's prediction? How can we show it?

Any answer is really appropriate

[lhlong]

@albertoandreottiATgmail please

[github-actions[bot]]

This issue is stale because it has been open 120 days with no activity. Remove stale label or comment or this will be closed in 5 days