Как решить? IndexError: допустимыми являются только целые числа, срезы (`:`), многоточие (`...`), numpy .newaxis (`None`) и целые или логические массивы.

Question

Я попробовал этот код Q-обучения в ноутбуке Jupyter. Это работало хорошо без ошибок. Я решил попробовать тот же код в aws, используя ноутбук Jupyter, загрузив данные из корзины S3. Я продолжаю получать ошибки (IndexError: допустимыми являются только целые числа, срезы (:), многоточие (...), numpy .newaxis (None) и целые или логические массивы). Буду очень признателен, если кто-нибудь поможет мне решить эту проблему.

#Maps indices to locations
state_to_location = dict((state,location) for location, state in location_to_state.items())

# Initialize parameters
gamma = 0.9 # Discount factor 
alpha = 0.3 # Learning rate

# Get the ending state corresponding to the ending location as given

def get_optimal_route(start_location,end_location):
    # Copy the rewards matrix to new Matrix
    rewards_new = np.copy(rewards)

    # Get the ending state corresponding to the ending location as given
    ending_state = location_to_state[end_location]

    # With the above information automatically set the priority of  
    # the given ending state to the highest one
    rewards_new["ending_state","ending_state"] = 999

    # -----------Q-Learning algorithm-----------

    # Initializing Q-Values
    Q = np.array(np.zeros([11,11]))
    print("Initialized Q matrix")
    print(Q)
    for i in range(10000):
        # Pick up a state randomly
        current_state = np.random.randint(0,11) # Python excludes the upper bound
        # For traversing through the neighbor locations in the maze
        playable_actions = []
        # Iterate through the new rewards matrix and get the actions > 0
        for j in range(11):
            if rewards_new[current_state,j] > 0:
                playable_actions.append(j)
                # Pick an action randomly from the list of playable actions
                # leading us to the next state
                next_state = np.random.choice(playable_actions)
                # Compute the temporal difference
                # The action here exactly refers to going to the next state
                TD = rewards_new[current_state,next_state] + gamma * Q[next_state, np.argmax(Q[next_state,])] - Q[current_state,next_state]              
                # Update the Q-Value using the Bellman equation
                Q[current_state,next_state] += alpha * TD

    # Initialize the optimal route with the starting location
    route = [start_location]
    # We do not know about the next location yet, so initialize with the value of 
    # starting location
    next_location = start_location

    # We don't know about the exact number of iterations
    # needed to reach to the final location hence while loop will be a good choice 
    # for iteratiing

    while(next_location != end_location):
        # Fetch the starting state
        starting_state = location_to_state[start_location]

        # Fetch the highest Q-value pertaining to starting state
        next_state = np.argmax(Q[starting_state,])

        # We got the index of the next state. But we need the corresponding letter. 
        next_location = state_to_location[next_state]
        route.append(next_location)

        # Update the starting location for the next iteration
        start_location = next_location
        print("Trained Q matrix")
        #matrix = Q / np.max(Q) * 100
        #print(matrix)
        #print("Optimal route")

    return route

print(get_optimal_route('L1','L11'))




IndexError                                Traceback (most recent call last)
<ipython-input-236-677dc76cecf2> in <module>()
----> 1 print(get_optimal_route('L1','L11'))

<ipython-input-235-59580ef2070d> in get_optimal_route(start_location, end_location)
     10     # With the above information automatically set the priority of
     11     # the given ending state to the highest one
---> 12     rewards_new["ending_state","ending_state"] = 999
     13 
     14     # -----------Q-Learning algorithm-----------

IndexError: only integers, slices (`:`), ellipsis (`...`), numpy.newaxis (`None`) and integer or 
boolean arrays are valid indices