coach/rl_coach/environments

A custom environment implementation should look like this:

from coach.filters.input_filter import InputFilter

class CustomFilter(InputFilter):
  def __init__(self):
    ...
  def _filter(self, env_response: EnvResponse) -> EnvResponse:
    ...
  def _get_filtered_observation_space(self, input_observation_space: ObservationSpace) -> ObservationSpace:
    ...
  def _get_filtered_reward_space(self, input_reward_space: RewardSpace) -> RewardSpace:
    ...
  def _validate_input_observation_space(self, input_observation_space: ObservationSpace):
    ...
  def _reset(self):
    ...