CNN模型训练调用model.fit时触发ValueError：输入形状不兼容

问题现象 #

训练CNN模型调用model.fit()时触发ValueError，提示输入形状与模型输入层不兼容：模型期望输入形状为(None, 500, 128)，但实际传入的输入形状是(None, 10)。

相关代码 #

DO = Denoiser()
visible = Input(shape=(500, batch_size))
my_denoiser = DO.rkhs(visible, kern, I_mat)
conv1 = Conv1D(6, kernel_size=4, activation='relu')(visible)
pool1 = MaxPooling1D(pool_size=5)(conv1)
conv2 = Conv1D(12, kernel_size=4, activation='relu')(pool1)
pool2 = MaxPooling1D(pool_size=5)(conv2)
flat = Flatten()(pool2)
hidden1 = Dense(10, activation='relu')(flat)
output = Dense(3, activation='softmax')(hidden1)
model = Model(inputs=visible, outputs=output)

model.compile(optimizer='Adam', loss='categorical_crossentropy', metrics=['accuracy'])
model.fit(new_signal, y, epochs=2, batch_size=200)

模型结构输出 #

Model: "model_3"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 input_6 (InputLayer)        [(None, 500, 128)]        0         
                                                                 
 conv1d_7 (Conv1D)           (None, 497, 6)            3078      
                                                                 
 max_pooling1d_7 (MaxPooling  (None, 99, 6)            0         
 1D)                                                             
                                                                 
 conv1d_8 (Conv1D)           (None, 96, 12)            300       
                                                                 
 max_pooling1d_8 (MaxPooling  (None, 19, 12)           0         
 1D)                                                             
                                                                 
 flatten_3 (Flatten)         (None, 228)               0         
                                                                 
 dense_6 (Dense)             (None, 10)                2290      
                                                                 
 dense_7 (Dense)             (None, 3)                 33        
                                                                 
=================================================================
Total params: 5,701
Trainable params: 5,701
Non-trainable params: 0
_________________________________________________________________

错误信息（中文翻译） #

ValueError: 用户代码中出错：

File "/usr/local/lib/python3.7/dist-packages/keras/engine/training.py", line 1021, in train_function  *
    return step_function(self, iterator)
File "/usr/local/lib/python3.7/dist-packages/keras/engine/training.py", line 1010, in step_function  **
    outputs = model.distribute_strategy.run(run_step, args=(data,))
File "/usr/local/lib/python3.7/dist-packages/keras/engine/training.py", line 1000, in run_step  **
    outputs = model.train_step(data)
File "/usr/local/lib/python3.7/dist-packages/keras/engine/training.py", line 859, in train_step
    y_pred = self(x, training=True)
File "/usr/local/lib/python3.7/dist-packages/keras/utils/traceback_utils.py", line 67, in error_handler
    raise e.with_traceback(filtered_tb) from None
File "/usr/local/lib/python3.7/dist-packages/keras/engine/input_spec.py", line 264, in assert_input_compatibility
    raise ValueError(f'Input {input_index} of layer "{layer_name}" is '

ValueError: 层"model_3"的输入0与该层不兼容：预期形状=(None, 500, 128)，实际找到形状=(None, 10)

解决方案 #

1. 修正输入层定义 #

Keras的Input层的shape参数仅需指定单个样本的特征维度，不需要包含batch维度（batch维度由model.fit()的batch_size参数控制）。代码中错误地将batch_size作为特征维度的一部分传入，导致输入层期望(500,128)的样本形状。

将输入层定义修改为：

# 假设每个时间步是单特征（比如单变量时序数据），则shape=(500,1)
# 如果是多变量时序，替换1为实际特征数
visible = Input(shape=(500, 1))

2. 对齐输入数据形状 #

确保训练数据new_signal的形状与输入层定义一致，即(样本数量, 500, 特征数)：

• 若当前new_signal形状是(样本数量,10)，说明你传入了错误的数据（比如误传了其他特征集而非目标信号），需要确认数据源；
• 若数据本身是(样本数量,500)的一维时序，需要扩展维度匹配输入层：

  new_signal = np.expand_dims(new_signal, axis=-1)
  

3. 清理冗余代码 #

代码中定义了my_denoiser = DO.rkhs(visible, kern, I_mat)但未将其加入模型计算图，属于冗余代码，可直接删除避免混淆。

内容的提问来源于stack exchange，提问作者Raheleh Mohseni