Keras报错ValueError: Error when checking model input: the list of Numpy arrays

Keras框架使用过程遇到ValueError

ValueError: Error when checking model input: the list of Numpy arrays that you are passing to your model is not the size the model expected.

检查模型输入时出错：传递给模型的Numpy数组列表不是模型预期的大小

遇到的问题

Traceback (most recent call last):
  File "train.py", line 88, in <module>
    train()
  File "train.py", line 77, in __call__
    self.fit()
  File "train.py", line 71, in fit
    self._fit()
  File "train.py", line 58, in _fit
    validation_steps=c_d_v.get_lst_size(self.val_fileid_lst, 'Vaild') // self.BATCH_SIZE)
  File "E:\Anaconda3\envs\hw1\lib\site-packages\keras\legacy\interfaces.py", line 91, in wrapper
    return func(*args, **kwargs)
  File "E:\Anaconda3\envs\hw1\lib\site-packages\keras\engine\training.py", line 1415, in fit_generator
    initial_epoch=initial_epoch)
  File "E:\Anaconda3\envs\hw1\lib\site-packages\keras\engine\training_generator.py", line 213, in fit_generator
    class_weight=class_weight)
  File "E:\Anaconda3\envs\hw1\lib\site-packages\keras\engine\training.py", line 1209, in train_on_batch
    class_weight=class_weight)
  File "E:\Anaconda3\envs\hw1\lib\site-packages\keras\engine\training.py", line 749, in _standardize_user_data
    exception_prefix='input')
  File "E:\Anaconda3\envs\hw1\lib\site-packages\keras\engine\training_utils.py", line 101, in standardize_input_data
    str(len(data)) + ' arrays: ' + str(data)[:200] + '...')
ValueError: Error when checking model input: the list of Numpy arrays that you are passing to your model is not the size the model expected. Expected to see 2 ar
t of 1 arrays: [array([[[ 0.        ,  0.        ,  1.        ,  1.15957451],
        [-0.24468085,  0.26595744,  0.30851063,  0.09574468],
        [-0.05319149, -0.0212766 ,  0.06382979,  0.36170211],

问题原因

keras在多输入多输出模型时，训练时传入的数据不满足模型定义

解决办法：

修改model.fit或者model.fit_generator中参数，使模型定义的输入输出列表和自适应处的调用一致

模型定义的输出输出列表

model = Model(inputs=[a1, a2], outputs=[b1, b3, b3])

自适应处的调用

model.fit([a1, a2], [b1, b3, b3],
                  validation_data=([a1, a2], [b1, b3, b3]),
                  epochs=1, batch_size=64)

def gen():
	yield [a1, a2], [b1, b3, b3]
	
model.fit_generator(gen(),
                  validation_data=gen(),
                  epochs=1, batch_size=64)

踩过的坑

1. 除了训练数据，还要注意验证数据

model.fit([a1, a2], [b1, b3, b3],   # 训练数据
                  validation_data=([a1, a2], [b1, b3, b3]),   # 验证数据
                  epochs=1, batch_size=64)

2. model.fit_generator注意生成器中yield所有返回格式都要一致

        if count >= batch_size:
            X, Y = data2npy(dList, max_len)
            yield [X,Y],Y # 第一处返回
            count = 0
            dList = []
            
if dList:
    X, Y = data2npy(dList, max_len)
    yield [X,Y],Y # 第二处返回相同的格式