search

ClementPerroud / Rainbow-Agent

Replicate of Reinforcement Learning Rainbow with Tensorflow 2 from paper "Rainbow: Combining Improvements in Deep Reinforcement Learning"
MIT License
10 stars 7 forks source link

Model saving and reloading #3

Closed bagumamartin closed 5 months ago

bagumamartin commented 5 months ago

Looks like the logic for saving and reloading the model; both in agents.py and utils.save.py are partially implemented. Anyone with complete work arounds.

bagumamartin commented 5 months ago

Actually, the saving Logic in agent.py is complete, just not for re-training.

For retraining, I had to modify it a little bit. The issue for retraining is the replay_memory, that has to be restored in a way.