Closed pandeydeep9 closed 4 years ago
Makes perfect sense and yes you are right. I noticed it some time ago and forgot to update the code. Feel free to push a PR. I'll review it.
On Fri, 17 Jul 2020, 01:51 Deep Shankar Pandey, notifications@github.com wrote:
Hi there,
Thanks for the great code.
I was experimenting with checkpoints/resuming experiments from specified epoch for maml experiments and noticed that the optimizer state dict is not saved.
I think since outer loop uses adam optimizer we should save the optimizer state dict along with the model state dict i.e. in few_shot_learning_system.py save_model, adding state['optimizer'] = self.optimizer.state_dict() and in load_model adding self.optimizer.load_state_dict(state['optimizer']).
Let me know if this does not make sense.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/AntreasAntoniou/HowToTrainYourMAMLPytorch/issues/31, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACSK4NX52N3EZZ27AMLKNFDR36OBFANCNFSM4O5HKCHQ .
Hi there,
Thanks for the great code.
I was experimenting with checkpoints/resuming experiments from specified epoch for maml experiments and noticed that the optimizer state dict is not saved.
I think since outer loop uses adam optimizer we should save the optimizer state dict along with the model state dict i.e. in few_shot_learning_system.py save_model, adding state['optimizer'] = self.optimizer.state_dict() and in load_model adding self.optimizer.load_state_dict(state['optimizer']).
Let me know if this does not make sense.