Cannot resume train.py from checkpoint

[jackhftang]

No description provided.

[flyyufelix]

First, you have to save/export the weights. You can either use model.save_weights("model.h5") or use checkpoint (see https://machinelearningmastery.com/check-point-deep-learning-models-keras/).

Once you have the exported weight you can resume training by calling the Keras fit() function with initial_epoch parameter set to the number of epoch which the training left off. For example, if you paused the training at epoch 10 and want to resume training afterwards, you can set initial_epoch at 10 (see https://keras.io/models/model/#fit).