Hi, do you have an usage example for plotting results with this code? If I don't use Container from the azure_util module then only a text file with timesteps and averaged reward would be saved.

Edit: I figured out about results logging with SimpleMonitor but I would appreciate an use case of saving and logging with Container.
Thank you kindly!

Dan

Hi, there can be a bug in wrappers. When I first ran the code, I got the following bug.

I found the reason is that wrapper' s attributions, '_reset', '_step' , are not used at all. This might because of different gym version (they change '_reset' to 'reset'). Finally, I fix the bug by setting all '_reset' , '_step', to 'reset' , 'step' in baselines/common/atari_wrappers_deprecated.py.

Can you report the exact version of gym and fix this bug ?

Thanks!

Hello LinZichuan,

I tried your train_cartpole.py, but got:

Traceback (most recent call last):
File "train_cartpole.py", line 31, in
main()
File "train_cartpole.py", line 24, in main
callback=callback
File "c:\emdqn-master\baselines\deepq\simple.py", line 235, in learn
td_errors = train(obses_t, actions, rewards, obses_tp1, dones, weights)
File "c:\emdqn-master\baselines\common\tf_util.py", line 442, in call
assert inpt in self.givens, "Missing argument " + inpt_name
AssertionError: Missing argument qec

   Thanks for your help!