Keras, best way to save state when optimizing

Question

I was just wondering what is the best way to save the state of a model while it it optimizing. I want to do this so I can run it for a while, save it, and come back to it some time later. I know there is a function to save the weights and another function to save the model as JSON. During learning I would need to save both the weights and the parameters of the model. This includes parameters like the momentum and learning rate. Is there a way to save both the model and weights in the same file. I read that it is not considered good practice to use pickle. Also would the momentums for the graident decent be included with the models JSON or in the weights?

Answer 1

from keras.models import load_model

model.save('my_model.h5')  # creates a HDF5 file 'my_model.h5'
del model  # deletes the existing model

# returns a compiled model
# identical to the previous one
model = load_model('my_model.h5')

You can use model.save(filepath) to save a Keras model into a single HDF5 file which will contain:

• the architecture of the model, allowing to re-create the model
• the weights of the model
• the training configuration (loss, optimizer)
• the state of the optimizer, allowing to resume training exactly where you left off.

You can then use keras.models.load_model(filepath) to reinstantiate your model. load_model will also take care of compiling the model using the saved training configuration (unless the model was never compiled in the first place).

Keras FAQ: How can I save a Keras model?