在 rllib 中注册健身房环境时传递可选的 arguments

Question

I've been following the helpful example here to create a custom environment in gym, which I then want to train in rllib.

My environment has some optional parameters which I would like to select when training. I think the relevant part of code is in train.py here:

    # start Ray -- add `local_mode=True` here for debugging
    ray.init(ignore_reinit_error=True)

    # register the custom environment
    select_env = "example-v0"
    register_env(select_env, lambda config: Example_v0())

I've tried some obvious things, like

    register_env(select_env, lambda config: Example_v0(optional_arg=n))

but nothing has seemed to work. Is there a way to pass different arguments before training?

Answer 1

I think you should use env_config in the constructor and then pass the dictionary env_config = {'optional_arg': n}