Regarding to policy/model/weights #18

Activity

1. zyzhang1130 commented on Feb 3, 2020

   

   zyzhang1130

   on Feb 3, 2020

   Author

   More actions

   Hi, I found this piece of info quite relevant! Kaixhin/Rainbow#58
   However, I am not proficient to incorporate it into the rainbow code of yours (judging based on the time it was added, your current version rainbow probably does not have this feature). Do you mind to take a look at it?

   Thank you so much
2. zyzhang1130 commented on Feb 3, 2020

   

   zyzhang1130

   on Feb 3, 2020

   Author

   More actions

   Actually the contributor of Rainbow says these lines save the model weights:
   https://github.com/Kaixhin/Rainbow/blob/d3afb5ad570137d675d6c7c903c050c8a19db084/main.py#L179-L181
   How do you think I should incorporate it into cups-lr Rainbow code?

   Thank you.
3. beduffy commented on Feb 11, 2020

   

   beduffy

   on Feb 11, 2020

   Last edited by beduffy

   Contributor

   More actions

   Line 79 in rainbow/test.py which is called periodically from the main.py file is the code which saves the model:

   # Save model parameters if improved
   if avg_reward > best_avg_reward:
       best_avg_reward = avg_reward
       dqn.save(path='weights', filename='rainbow_{}.pt'.format(num_steps))
   

   This is therefore saved in the rainbow directory from which is related to #17. Set your evaluation-interval lower. If it still doesn't save the model after args.evaluation_interval steps let me know.
4. zyzhang1130 commented on Feb 20, 2020

   

   zyzhang1130

   on Feb 20, 2020

   Author

   More actions

   Sorry, can I check does the current repo allow resume training feature? Thank you.