Incorrect number of arguments from call to env.step(action)

[JRod-Seed]

Incorrect number of arguments from call to env.step(action)

[pseudo-rnd-thoughts]

With v0.26 you can use this code

from nes_py.wrappers import JoypadSpace
import gym_super_mario_bros
from gym_super_mario_bros.actions import SIMPLE_MOVEMENT
import gym

env = gym.make('SuperMarioBros-v0', apply_api_compatibility=True, render_mode="human")
env = JoypadSpace(env, SIMPLE_MOVEMENT)

done = True
env.reset()
for step in range(5000):
    action = env.action_space.sample()
    obs, reward, terminated, truncated, info = env.step(action)
    done = terminated or truncated

    if done:
       state = env.reset()

env.close()

You can read about the migration guide for the new API https://gymnasium.farama.org/content/migration-guide/

[JRod-Seed]

It would appear that the issue stems from a mismatch in the expected return from env.step. gym/wrappers/time_limit.py returns 5 parameters (observation, reward, terminated, truncated, info) but the invocation is only expecting 4 (observation, reward, terminated, info)

Likely related to: Kautenja/nes-py#85

[Robwc000]

I'm having the same issue what is the fix?

[zxdclyz]

Try to use gym==0.25.1, it works for me

[PowderFan87]

I've debugged the step functions and it seems that nes-py returns 4 or 5 length tuple.
They have a code snippet that checks for if(Len(result)==4) and for 5.
That seems to be the "way to go" it just is not implemented in the gym/wrappers/time_limit.py

I adapted the snipped and set truncated to False as a default:

def step(self, action):
    """Steps through the environment and if the number of steps elapsed exceeds ``max_episode_steps`` then truncate.

    Args:
        action: The environment step action

    Returns:
        The environment step ``(observation, reward, terminated, truncated, info)`` with `truncated=True`
        if the number of steps elapsed >= max episode steps

    """

    truncated = False
    result = self.env.step(action)

    if (len(result) == 4):
        observation, reward, terminated, info = self.env.step(action)
    elif (len(result) == 5):
        observation, reward, terminated, truncated, info = self.env.step(action)
    else:
        raise Exception('Tupel length of step return was neither 4 nore 5. Stop run.')

    self._elapsed_steps += 1

    if self._elapsed_steps >= self._max_episode_steps:
        truncated = True

    return observation, reward, terminated, truncated, info

so instead of just reading 5 every time I do the same check and this works fine.
But I realise changing the lib is bad mojo so maybe this could be a fix ;) with a proper Exception added...