Env not unwrappable to BaseWorldModelWrapper

[kierad]

Hello, thanks for sharing your code. I'm having trouble running generate_expert_data.py. Here's what I'm trying:

• conda create --name irl-benchmark python=3.8 (The readme says 3.6, but 3.8 seems to be the oldest Python version supported by Conda right now).
• source activate irl-benchmark
• pip install -r requirements.txt
• python generate_expert_data.py which raises this error:

Traceback (most recent call last):
  File "generate_expert_data.py", line 20, in <module>
    expert_agent = rl_alg_factory(env)
  File "generate_expert_data.py", line 17, in rl_alg_factory
    return ValueIteration(env, {'gamma': 0.9})
  File "/Users/kieradavidson/Documents/GitHub/irl-benchmark/irl_benchmark/rl/algorithms/value_iteration.py", line 31, in __init__
    assert is_unwrappable_to(env, BaseWorldModelWrapper)
AssertionError

Adding print(env) gives this: <FrozenLakeFeatureWrapper<TimeLimit<FrozenLakeEnv<FrozenLake-v0>>>>

(And unsurprisingly, removing the assertion leads to another error:

Traceback (most recent call last):
  File "generate_expert_data.py", line 20, in <module>
    expert_agent = rl_alg_factory(env)
  File "generate_expert_data.py", line 17, in rl_alg_factory
    return ValueIteration(env, {'gamma': 0.9})
  File "/Users/kieradavidson/Documents/GitHub/irl-benchmark/irl_benchmark/rl/algorithms/value_iteration.py", line 35, in __init__
    self.model_wrapper = unwrap_env(env, BaseWorldModelWrapper)
  File "/Users/kieradavidson/Documents/GitHub/irl-benchmark/irl_benchmark/utils/wrapper.py", line 23, in unwrap_env
    raise ValueError(
ValueError: Unwrapping env did not yield an instance of class <class 'irl_benchmark.rl.model.model_wrapper.BaseWorldModelWrapper'>

)

Could you point me in the right direction? Thank you.

[HuanMinpopcorn]

I had same problem