python - How to save tensorflow model with best weights

Question

I'm training an LSTM, and for portion of epochs (range not consistent) during the 300 epoch training sequence it exhibits very good accuracy, for a significant length of time. I want to be able to save the entire model at this point, but can't seem to find a way to do it except to use checkpoints and then restore and save the new model.

Is there a way I can use model.save to only capture the model when it had its best weights?

Answer 1

Use save _best_only argument:

model_checkpoint_callback = tf.keras.callbacks.ModelCheckpoint(
    filepath=checkpoint_filepath,
    save_weights_only=True,
    monitor='val_accuracy',
    mode='max',
    save_best_only=True)