search

pokaxpoka / sunrise

SUNRISE: A Simple Unified Framework for Ensemble Learning in Deep Reinforcement Learning
119 stars 29 forks source link

UnboundLocalError: local variable 'env' referenced before assignment #3

Closed sweetice closed 3 years ago

sweetice commented 3 years ago

Error:

Traceback (most recent call last):
  File "/home/mirror/PycharmProjects/sunrise/OpenAIGym_SAC/examples/sunrise.py", line 197, in <module>
    experiment(variant)
  File "/home/mirror/PycharmProjects/sunrise/OpenAIGym_SAC/examples/sunrise.py", line 50, in experiment
    expl_env = NormalizedBoxEnv(get_env(variant['env'], variant['seed']))
  File "/home/mirror/PycharmProjects/sunrise/OpenAIGym_SAC/examples/sunrise.py", line 46, in get_env
    env = env(env_name=env_name, rand_seed=seed, misc_info={'reset_type': 'gym'})
UnboundLocalError: local variable 'env' referenced before assignment

Seems that the error comes from this line

def get_env(env_name, seed):
    if env_name in ['gym_walker2d', 'gym_hopper',
                    'gym_cheetah', 'gym_ant']:
        from mbbl.env.gym_env.walker import env
    env = env(env_name=env_name, rand_seed=seed, misc_info={'reset_type': 'gym'})
    return env

The env was used but not be defined. I don't know how to fix this bug due to the complexity of the codebase. Would you mind fixing this bug in this repo?

pokaxpoka commented 3 years ago

hm, hard to understand the problem at glance. Could you elaborate more? what was the env_name you used?

sweetice commented 3 years ago

Well, I use pubullet envs. Specifically, I use "HopperBulletEnv-v0" env. . And then the error described occurs. Seems that the error is due to the program don't process the env (Because "HopperBulletEnv-v0" is not in ['gym_walker2d', 'gym_hopper', 'gym_cheetah', 'gym_ant'])

I want to apply your code on another env.

sweetice commented 3 years ago

Thanks for your reply! I have solved the issue! :)