解释，然后是代码，然后是输出，最后是错误信息：

看起来Flatten层没有正常工作，输出形状取决于批量大小（当我设置BATCH_SIZE=32时，[1,32768]变成了[1,16384]）。我实在搞不明白自己哪里做错了，或者如何修复这个问题。我已经查看了Keras文档中关于Flatten和Dense层的部分。此外，我使用的是TensorFlow后端，并且这确实反映在keras.json文件中。

这是我的代码：

BATCH_SIZE = 64EPOCHS = 1000EPOCH_STEP = 50vgg = keras.applications.vgg16.VGG16(include_top=False,weights='imagenet',input_shape=(48,48,3))vgg_input = vgg.inputsvgg_output = vgg.outputs#freeze the vgg layersfor layer in vgg.layers:    layer.trainable = Falseprint('~~~~~~~~~~~~~~~~~~~~~~Tensors~~~~~~~~~~~~~~')print('vgg_output tensor:')print(vgg_output)print()model_tensor = Flatten()(vgg_output)print('flattened vgg_output tensor:')print(model_tensor)print()model_tensor = Dense(32, activation='relu')(model_tensor)print('dense FC flattened vgg_output tensor:')print(model_tensor)print('~~~~~~~~~~~~~~~~~~~~~~Tensors~~~~~~~~~~~~~~')model_tensor = Dense(2, activation='softmax')(model_tensor)model = Model(inputs=vgg_input,outputs=model_tensor)print('Model architecture made')#CHOSEN ARBITRARILY FOR NOWmodel.compile(optimizer='rmsprop',            loss='binary_crossentropy',            metrics=['accuracy'])print('Model Compiled')             print(model.summary())#train top modelval_batch, val_labels = dataGenerator.generateDataBatch(256)print('validation batch loaded')batch, labels = dataGenerator.generateDataBatch(2048)print('training batch loaded')print('t-batch shape: ' + str(batch.shape))print('t-batch lable shape: ' + str(labels.shape))        model.fit(x=batch,y=labels,batch_size=BATCH_SIZE,epochs=EPOCHS,verbose=2,validati    on_data=(val_batch,val_labels),shuffle=True)

打印输出：

Printed Info:~~~~~~~~~~~~~~~~~~~~~~Tensors~~~~~~~~~~~~~~vgg_output tensor:[<tf.Tensor 'block5_pool/MaxPool:0' shape=(?, 1, 1, 512) dtype=float32>]flattened vgg_output tensor:Tensor("flatten_1/Reshape:0", shape=(?, ?), dtype=float32)dense FC flattened vgg_output tensor:Tensor("dense_1/Relu:0", shape=(?, 32), dtype=float32)~~~~~~~~~~~~~~~~~~~~~~Tensors~~~~~~~~~~~~~~_________________________________________________________________Layer (type)                 Output Shape              Param #=================================================================input_1 (InputLayer)         (None, 48, 48, 3)         0_________________________________________________________________block1_conv1 (Conv2D)        (None, 48, 48, 64)        1792_________________________________________________________________block1_conv2 (Conv2D)        (None, 48, 48, 64)        36928_________________________________________________________________block1_pool (MaxPooling2D)   (None, 24, 24, 64)        0_________________________________________________________________block2_conv1 (Conv2D)        (None, 24, 24, 128)       73856_________________________________________________________________block2_conv2 (Conv2D)        (None, 24, 24, 128)       147584_________________________________________________________________block2_pool (MaxPooling2D)   (None, 12, 12, 128)       0_________________________________________________________________block3_conv1 (Conv2D)        (None, 12, 12, 256)       295168_________________________________________________________________block3_conv2 (Conv2D)        (None, 12, 12, 256)       590080_________________________________________________________________block3_conv3 (Conv2D)        (None, 12, 12, 256)       590080_________________________________________________________________block3_pool (MaxPooling2D)   (None, 6, 6, 256)         0_________________________________________________________________block4_conv1 (Conv2D)        (None, 6, 6, 512)         1180160_________________________________________________________________block4_conv2 (Conv2D)        (None, 6, 6, 512)         2359808_________________________________________________________________block4_conv3 (Conv2D)        (None, 6, 6, 512)         2359808_________________________________________________________________block4_pool (MaxPooling2D)   (None, 3, 3, 512)         0_________________________________________________________________block5_conv1 (Conv2D)        (None, 3, 3, 512)         2359808_________________________________________________________________block5_conv2 (Conv2D)        (None, 3, 3, 512)         2359808_________________________________________________________________block5_conv3 (Conv2D)        (None, 3, 3, 512)         2359808_________________________________________________________________block5_pool (MaxPooling2D)   (None, 1, 1, 512)         0_________________________________________________________________flatten_1 (Flatten)          (None, 512)               0_________________________________________________________________dense_1 (Dense)              (None, 32)                16416_________________________________________________________________dense_2 (Dense)              (None, 2)                 66=================================================================Total params: 14,731,170Trainable params: 16,482Non-trainable params: 14,714,688_________________________________________________________________NoneModel CompiledModel architecture madeModel Compiledvalidation batch loadedtraining batch loadedt-batch shape: (2048, 48, 48, 3)t-batch lable shape: (2048, 2)

错误信息的最后一部分（我可以添加所有内容，但其余部分看起来没有帮助）：

tensorflow.python.framework.errors_impl.InvalidArgumentError: Matrix size-incompatible: In[0]: [1,32768], In[1]: [512,32]     [[Node: dense_1/MatMul = MatMul[T=DT_FLOAT, transpose_a=false, transpose_b=false, _device="/job:localhost/replica:0/task:0/device:CPU:0"](flatten_1/Reshape, dense_1/kernel/read)]]

编辑：在model.compile()之后立即添加了print(model.summary())的输出。

回答：