Keras: How to save GAN model with optimiser state?

Asked Feb 06 '22 at 10:51

Active Feb 08 '22 at 11:11

Viewed 218 times

I know one use can be using Checkpoint and CheckpointManager tensorflow classes to save and resume training the GAN.

I am looking for a way to save the Generator (G) , Discriminator(D) and their optimisation states such that I can manipulate the model weights (for G and D). One use case for me is average the weights of different Generator in Federated Learning for GAN.

I found no way do to this using Keras or tensorflow.

edited Feb 08 '22 at 11:11

asked Feb 06 '22 at 10:51

Bhavesh Neekhra

Please provide enough code so others can better understand or reproduce the problem. – Community Feb 16 '22 at 23:09
With optimization state you mean you want to safe the model with its trained weights or that you also want to save the optimzer? – Sascha Kirch Feb 23 '22 at 17:27

Keras: How to save GAN model with optimiser state?

0 Answers0