Keras functional api输入形状

Question

我的模型定义是：

inputs = keras.Input(shape=(28,28))
dense = keras.layers.Dense(64, activation="relu")
x = dense(inputs)
x = keras.layers.Dense(64, activation="relu")(x)
outputs = keras.layers.Dense(10)(x)
model = keras.Model(inputs=inputs, outputs=outputs, name="mnist_model")

我将使用以下内容来训练网络：

model.compile(optimizer="sgd", loss=keras.losses.SparseCategoricalCrossentropy(from_logits=True), metrics=['accuracy'])
history = model.fit(x_train, y_train, batch_size=10, epochs=30, verbose=2)

x_train的形状是(55000，28，28)，而y_train的形状是(55000，)，我得到了这个错误：

ValueError: Shape mismatch: The shape of labels (received (10, 1)) should equal the shape of logits except for the last dimension (received (10, 28, 10))

Answer 1

@Marco Cerliani所说的意思是您应该将您的代码更改为

inputs = keras.Input(shape=(28,28))
output = keras.layers.Flatten()(inputs)
output = keras.layers.Dense(64, activation="relu")(output)
output = keras.layers.Dense(64, activation="relu")(output)
outputs = keras.layers.Dense(10)(output)
model = keras.Model(inputs=inputs, outputs=outputs, name="mnist_model")

如果您是初学者，请阅读tensorflow tutorials和guide。如果你阅读this，你会找到问题的答案。