Add more parameters to env_config?

RLlib
1

ray.init()
env_config = {
“config_file”: trainer_config_file
}
config ={
“env”: ScimEnv,
“env_config”: env_config,
“num_gpus”: 0,
“num_workers”: 1,
“lr”: 1e-3,
“framework”: “torch”
}
results = tune.run(“DQN”, stop={“episode_reward_mean”: 20}, config=config)
df = results.results_df
ray.shutdown()

Here is my code. Ideally my environment has two inputs, one for the config file and another dictionary. However, from my understanding env_config specifically works with only the config directory in the dictionary. Is there any way to add to env_config?

2

@Yared_Kokeb,

welcome to the forum. I have a similar setup where I pass configurations via the env_config. I define my __init__() to be

class MyEnv(gym.Env):
   def __init__(self, env_config=None):
       super(MyEnv, self).__init__()
       self.config = env_config or {}
       ...

If you want to pass configuration parameters to your model (I see a trainer_config_file there) you can use in the config you defined above for example

config={
   "env_config": {
          "my_env_parameter": 10,
   },
   "model": {
          "max_seq_len": 1,
          # Or if you have a custom model or policy:
          #"custom_model_config": {
          #       "my_custom_model_parameter": 8,
          #}
   }
}

Hope this helps.
Simon