How to adapt this code to a new environment?

[pranv]

Hi,

This looks great. How would you go about and adapt this to Open AI Gym for example?

Can you please provide a set of places where changes have to be made?

How Generic is the code to adapt to any environment?

[muupan]

Adding examples of applying A3C for OpenAI Gym's continuous tasks is on my To-Do list.

Since DoomEnv used by train_a3c_doom.py has a similar interface with gym.Env, modifying train_a3c_doom.py so that it can handle another gym.Env-like environments is straightforward. You may need to define your model that inherits a3c.A3CModel and appropriate phi function.

[pranv]

Thanks!

I will try and see how it goes.

[gowthamnatarajan]

Is it only for gym? What about other problems?

[muupan]

You only need to define an original environment class that has reset and step methods. Please check doom_env.py for example.

[gowthamnatarajan]

Thanks. I will rewrite those methods. Mine problem is not even a game.