0

I am working on a deep reinforcement problem, I am new to this. I am writing a snippet of code and errors I am getting.

Broker_Node_Map is a list of values present in different positions in a machine. I don't know how to present these values as integers. This is the state I have as it's changing too accordingly. Please suggest what should I do. Please be kind, I am pretty new and trying to get hold of things.

def __init__(self):
    super(BrokerEnv2, self).__init__()
    reward = 0 
    self.action_space = DiscreteActions.get_action_space()
    self.observation_space = DiscreteObservations.get_observation_space() 

def reset(self):
    observed_State = self.Broker_Node_Map
    return observed_State

While checking env on stable baselines - check_env(env) **Error** - AssertionError: The observation returned by reset() method must be an int

EDIT 1 - Very careless of me. Changed the space to box space but now another error emerged. AssertionError: The observation returned by the reset() method does not match the given observation space

This is what my reset() is returning -

<class 'numpy.ndarray'> (46,) [26 33  0 50  0  0 73 26  0 29  0 34 27 67  0  0  0  0 35 60  0  0 24 22
  0  0  0  0 25  0 17  0  0  0 21  0  0 53 68 40 51  0 62  0 56  0]

This is how I have defined my observation space -

self.observation_space = spaces.Box(low=1, high=73, shape=(46,), dtype=np.int64)

Please help me out why this error is coming?

SURABHI SHARMA
  • 37
  • 1
  • 6

2 Answers2

2

So I ran into the same problem and the fix for me was to set the bounds of space correctly. For example in your case, the state contains zeros but you've set the low bound to 1.

NaWeeD
  • 561
  • 5
  • 15
1

I ran into a problem like this and I fixed it by making the dtype to be correct

First you know the dtype by going like print(observed_State.dtype) and that will tell you the type of the array they you change the dtype arg to be the correct one, thx

ZiyadCodes
  • 379
  • 3
  • 10