jin

I solved this issue by simply adding this in `if __name__ == '__main__'` of `run.py` before `parser = ...` ```python os.environ['CUDA_VISIBLE_DEVICES'] = '0' ``` This way, you can control which...

Looks to me that you may have not installed `atari-py`. It is not officially supported on Windows, but this wonderful person made a Windows port of it (https://stackoverflow.com/questions/42605769/openai-gym-atari-on-windows)! Try after...

That is very odd 😨 I'll look into it! I'm actually working on optimization, so I'll let you know if I find anything wrong with the code. Thanks for the...

It seems like there's something broken overall... :( I'll make sure to let you know after it's fixed!

Hello, Thank you so much for the comment, I actually recently went through the code and decided that there are more bugs than I expected. I think I should have...

Don't worry about it :) I actually decided to flush signal on feedforward so that I can get rid of recurrency actually. I think I made the problem a little...

That's nice to hear! I'm hoping to continue as a graduate student once I graduate as well! It's great to have at least a little bit of pressure to motivate...

That's a good point, actually! I'll work on that at some point :) Thanks for the idea!

Evolution strategies do not use backpropagation, but they differentiate (approximately) across a population of solutions, rather than across parameters _within_ a solution. Perhaps reading [this blog post](http://blog.otoro.net/2017/10/29/visual-evolution-strategies/) is useful :stuck_out_tongue: