You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
A recent update to OpenAI Gym changed the way envs behave.
gym.Env.reset() now returns a tuple of (observation, info) so I think that every instance of observation = deepcopy(env.reset()) in keras-rl.rl.core.Agent.fit() should be change to observation, info = deepcopy(env.reset()).
The text was updated successfully, but these errors were encountered:
A recent update to OpenAI Gym changed the way envs behave.
gym.Env.reset() now returns a tuple of (observation, info) so I think that every instance of
observation = deepcopy(env.reset())
inkeras-rl.rl.core.Agent.fit()
should be change toobservation, info = deepcopy(env.reset())
.The text was updated successfully, but these errors were encountered: