当我想在 keras 中拟合模型时，列表索引超出范围错误

Question

这是我的代码，用 keras 编写

import keras
from keras import layers
from keras import Sequential
from keras.layers import Dense, Flatten
import numpy as np
from keras.engine.topology import Input
from keras.engine.training import Model

class _Model:
  def __init__(self,state,n_dims, n_action):
    self.n_action=n_action
    self.n_dims=n_dims
    self.state=state
    self.model=self.build_model()
  def build_model(self):
    inx=model=Input((10,16))
    model=Flatten()(model)
    model=Dense(512, activation=None)(model)
    model=Dense(512, activation=None)(model)
    p_model=Dense(self.n_action, activation='sigmoid')(model)
    v_model=Dense(1, activation='tanh')(model)
    _model=Model(inx,[p_model,v_model])
    losses = ['categorical_crossentropy', 'mean_squared_error']
    _model.compile(loss=losses, optimizer='adam')
    print(_model.summary())
    return _model
  def predict(self,state):
    return self.model.predict(state)
  def train(self, state, action_probability, leaf_value):
    batch_size=11
    state=np.array(state)
    action_probability=np.array(action_probability)
    leaf_value=np.array(leaf_value)
    self.model.fit(state, [action_probability, leaf_value],batch_size=batch_size,verbose=1)
    loss=self.model.evaluate(state, [action_probability, leaf_value],batch_size=batch_size,verbose=0)
    return loss[0]

state=[ 4321432141243124,
        1423123424143213,
        4321432143213421,
        4321431241324323,
        1243121234214334,
        4123123421342314,
        4321432434212412,
        4121432121121343,
        4123413412412321,
        4123413412413431,
]
m=_Model(state,16,4)
m.train(state,[0.1,0.5,0.4,0.7],0.4)

此实现适用于 alphago 零。我正在尝试实现具有两个输出的模型。两个值 (P, v)，p 是动作概率，v 是获胜概率。 有什么问题？这段代码有什么问题？ 错误说列表超出索引，但我不知道是什么列表。我应该更改 input_shape 吗？

---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
<ipython-input-9-c5b4b76d5be5> in <module>()
     48 ]
     49 m=_Model(state,16,4)
---> 50 m.train(state,[0.1,0.5,0.4,0.7],0.4)

6 frames
<ipython-input-9-c5b4b76d5be5> in train(self, state, action_probability, leaf_value)
     32     action_probability=np.array(action_probability)
     33     leaf_value=np.array(leaf_value)
---> 34     self.model.fit(state, [action_probability, leaf_value],batch_size=batch_size,verbose=1)
     35     loss=self.model.evaluate(state, [action_probability, leaf_value],batch_size=batch_size,verbose=0)
     36     return loss[0]

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py in _method_wrapper(self, *args, **kwargs)
    106   def _method_wrapper(self, *args, **kwargs):
    107     if not self._in_multi_worker_mode():  # pylint: disable=protected-access
--> 108       return method(self, *args, **kwargs)
    109 
    110     # Running inside `run_distribute_coordinator` already.

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/training.py in fit(self, x, y, batch_size, epochs, verbose, callbacks, validation_split, validation_data, shuffle, class_weight, sample_weight, initial_epoch, steps_per_epoch, validation_steps, validation_batch_size, validation_freq, max_queue_size, workers, use_multiprocessing)
   1061           use_multiprocessing=use_multiprocessing,
   1062           model=self,
-> 1063           steps_per_execution=self._steps_per_execution)
   1064 
   1065       # Container that configures and calls `tf.keras.Callback`s.

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/data_adapter.py in __init__(self, x, y, sample_weight, batch_size, steps_per_epoch, initial_epoch, epochs, shuffle, class_weight, max_queue_size, workers, use_multiprocessing, model, steps_per_execution)
   1115         use_multiprocessing=use_multiprocessing,
   1116         distribution_strategy=ds_context.get_strategy(),
-> 1117         model=model)
   1118 
   1119     strategy = ds_context.get_strategy()

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/data_adapter.py in __init__(self, x, y, sample_weights, sample_weight_modes, batch_size, epochs, steps, shuffle, **kwargs)
    273     inputs = pack_x_y_sample_weight(x, y, sample_weights)
    274 
--> 275     num_samples = set(int(i.shape[0]) for i in nest.flatten(inputs))
    276     if len(num_samples) > 1:
    277       msg = "Data cardinality is ambiguous:\n"

/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/engine/data_adapter.py in <genexpr>(.0)
    273     inputs = pack_x_y_sample_weight(x, y, sample_weights)
    274 
--> 275     num_samples = set(int(i.shape[0]) for i in nest.flatten(inputs))
    276     if len(num_samples) > 1:
    277       msg = "Data cardinality is ambiguous:\n"

/usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/tensor_shape.py in __getitem__(self, key)
    885       else:
    886         if self._v2_behavior:
--> 887           return self._dims[key].value
    888         else:
    889           return self._dims[key]

IndexError: list index out of range

Answer 1

您将一维数组提供给 keras fit 方法，keras 需要至少 2D 的所有输入和输出。试试这个。

state = np.array(state)
state = np.reshape(state, (-1,state.shape[0]))

action_probability = np.array(action_probability)
action_probability = np.reshape(action_probability, (-1,action_probability.shape[0]))

leaf_value = np.array(leaf_value)
leaf_value = np.reshape(leaf_value, (-1,leaf_value.shape[0]))