model.fit неверный синтаксис машинного обучения Python - PullRequest
Новая
       9

model.fit неверный синтаксис машинного обучения Python

0 голосов
/ 24 декабря 2018

Я смотрел много учебников по машинному обучению для работы в университете, я следовал учебнику, который предоставил этот код.Я получаю неверную синтаксическую ошибку, и я не уверен, почему, надеюсь, кто-то может помочь с этой ошибкой.

Строка, которой она не нравится: model.fit (np.array (x_batch), np.array (y_batch), batch_size = len (x_batch), подробный = 0) 

import gym
import keras
import random
import math
import tensorflow as tf
import numpy as np
from collections import deque

#Training Parameters
n_episodes = 1000
n_win_ticks = 195
max_env_steps = None

gamma = 1.0 #Discount Factor
epsilon = 1.0 #Exploration Factor
epsilon_min = 0.01
epsilon_decay = 0.995
alpha = 0.01 #The Learning Rate
alpha_decay = 0.01

batch_size = 64
monitor = False
quiet = False

#Environment Parameters
memory = deque(maxlen=100000)
env = gym.make('CartPole-v0')
if max_env_steps is not None: env.max_episode_steps = max_env_steps

#Building Neural Network
from keras.models import Sequential
from keras.layers import Dense
from keras.optimizers import Adam

#Model Definition
model = Sequential()
model.add(Dense(24, input_dim=4, activation='relu'))
model.add(Dense(48, activation='relu'))
model.add(Dense(2, activation='relu'))
model.compile(loss='mse', optimizer=Adam(lr=alpha, decay=alpha_decay))

#Define Functions

def remember(state, action, reward, next_state, done):
    memory.append((state, action, reward, next_state, done))

def choose_action(state, epsilon):
    return env.action_space.sample() if (np.random.random() <= epsilon) else np.argmax(model.predict(state))

def get_epsilon(t):
    return max(epsilon_min, min(epsilon, 1.0 = math.log10((t+1)*epsilon_decay)))

def preprocess_state(state):
    return np.reshape(state, [1, 4])

def replay(batch_size, epsilon):
    x_batch, y_batch = [], []
    minibatch = random.sample(memory, min(len(memory), batch_size))

    for state, action, reward, next_state, done in minibatch:
        y_target = model.predict(state)
        y_target[0][action] = reward if done else reward + gamma * np.max(model.predict(next_state)[0])
        x_batch.append(state[0])
        y_batch.append(y_target[0]

    model.fit(np.array(x_batch), np.array(y_batch), batch_size=len(x_batch), verbose=0)

    if epsilon > epsilon_min:
        epsilon *= epsilon_decay

#Define the function to run the program
def run():
    scores = deque(maxlen=100)

    for e in range(n_episodes):
        state = preprocess_state(env.reset())
        done = False
        i = 0
        while not done:
            action = choose_action(state, get_epsilon(e))
            next_state, reward, done _ = env.step(action)
            env.render()
            next_state = preprocess_state(next_state)
            remember(state, action, reward, next_state, done)
            state = next_state
            i += 1
        scores.append(i)
        mean_score = np.mean(scores)

        if mean_score >= n_win_ticks and e >= 100:
            if not quiet: print('Ran {} episodes. Solved after {} trails'.format(e, e-100))
            return e - 100
        if e % 20 == 0 and not quiet:
            print('[Episode {}] - Mean Survival time over last 100 episodes was {} ticks.format'(e, mean_score))

        replay(batch_size, epsilon)

    if not quiet: print('Did not solve after {} episodes'.format(e)
    return e

run()
...