search

MG2033 / A2C

A Clearer and Simpler Synchronous Advantage Actor Critic (A2C) Implementation in TensorFlow
Apache License 2.0
183 stars 37 forks source link

Help running code #3

Closed ahsteven closed 6 years ago

ahsteven commented 6 years ago

I am not sure what i am doing wrong but I am in the A2C folder and when I run:

(gym) teves@teves:~/A2C$ python main.py config/breakout.json usage: main.py [-h] [--version] [--config CONFIG] main.py: error: unrecognized arguments: config/breakout.json Add a config file using '--config file_name.json'

or if I run:

(gym) teves@teves:~/A2C$ python main.py --config config/breakout.json Add a config file using '--config file_name.json'

How shall I run this?

MG2033 commented 6 years ago

There was an issue regarding the bunch library. It's fixed now. You should be able to run.

Tell me if you face any more issues.

ahsteven commented 6 years ago

Yes it works now running the command: python main.py --config config/pong.json Thank you

ahsteven commented 6 years ago

Hi Mostafa,

I am trying to understand A2C in detail and I have read through your code as well as the openai baselines code. I have a couple of questions.

Why do we need a train model? Could we not just update the step model after the rollout? Or does the train model serve as some sort of placehoder? Also, why are the parameters in the train model set to be reused and the step model not. I am also a little confused on how the step model parameters get updated.

I appreciate any help you can give me on this.

MG2033 commented 6 years ago

Ok. I will do it in the next few days.