bert-sklearn icon indicating copy to clipboard operation
bert-sklearn copied to clipboard

Published 20 hours ago • verified publisher icon charles9n

Unable to reload the saved model to retrain from last ckpt

Open sachinya00 opened this issue 5 years ago • 2 comments

image

sachinya00 avatar Aug 07 '19 10:08 sachinya00

If you want to continue training the model from a previous finetuned checkpoint you can use the 'load_at_start flag' in fit and set it to False. This isn't documented since I envisioned people would mostly use the finetuned model for inference.

So this should work for you now: new_model.fit(X_train, y_train, load_at_start=False)

Regardless, what you found is a bug which I will fix. Will get back to you when its done. Thanks!

charles9n avatar Aug 08 '19 20:08 charles9n

Thanks Charles!, it worked.

sachinya00 avatar Aug 09 '19 06:08 sachinya00