PPO2.py Update on using eval_n_iter instead of hardcoding

[haloted]

Updated PPO2.py as according to the three pointers.

[haloted]

Apologies for closing the other pull request, was experimenting with changing different default branches so accidentally deleted the other branches.

[haloted]

This latest commit can now adapt to different eval_n_iter, tested with 80 and 50 and produce recorded videos with saved weights in sync.

[jhwangbo]

looks great. I tested your branch and works fine on my laptop. Thx for your contribution

[haloted]

Hi Jwangbo, thanks for the merge, just to avoid confusion of future users, please remove the import math package line at the top, it's no longer needed. And the line comment on line 379 of PPO2.py, change evaliter=200 to eval_n_iter instead in the comments. Cheers,ted

[jhwangbo]

Done. I also made a few changes. Previously the parameter was saved after an update. Then it no longer corresponds to the video. And the code is shorter now after the change