Tensorflow-2中cifar10数据的训练问题

Question

在tensorflow-2中训练cifar10数据时，我得到了以下错误。我用了这个教程。

TypeError:预期的float32传递给op‘相等’的参数'y‘，而得到'str’类型的'collections‘。错误:预期的float32，得到'str‘类型的“集合”。

我的代码看起来是：

    class Mymodel(tf.keras.Model):

        def __init__(self, class_size):
            """Initialize parameters and build model.
            """
            super(Mymodel, self).__init__()

            self.class_size =class_size
            self.conv1 = tf.keras.layers.Conv2D(32, kernel_size =3, strides =2, activation='relu')
            self.conv2 = tf.keras.layers.Conv2D(64, kernel_size =2, strides =2, activation='relu')
            self.conv3 = tf.keras.layers.Conv2D(64, kernel_size =2, strides =1, activation='relu')
            self.flat = tf.keras.layers.Flatten()
            self.d1 = tf.keras.layers.Dense(512, activation='relu')
            self.d2 = tf.keras.layers.Dense(128, activation='relu')
            self.fd =tf.keras.layers.Dense(self.class_size, activation='softmax') 

        def call(self, inputs):
            x = self.conv1(inputs)
            x = self.conv2(x)
            x = self.conv3(x)
            x = self.flat(x)
            x = self.d1(x)
            x = self.d2(x)
            return self.fd(x)

    model = Mymodel(10)

    train_images, train_labels), (test_images, test_labels) = datasets.cifar10.load_data()
    train_images, test_images = train_images / 255.0, test_images / 255.0

    train_ds = tf.data.Dataset.from_tensor_slices(
        (train_images, train_labels)).shuffle(1000).batch(32)

    test_ds = tf.data.Dataset.from_tensor_slices((test_images, test_labels)).batch(32)

    # define the training and testing objects 

    loss_object = tf.keras.losses.SparseCategoricalCrossentropy()

    optimizer = tf.keras.optimizers.Adam()

    @tf.function
    def train_step(images, labels):
      with tf.GradientTape() as tape:
        predictions = model(images)
        loss = loss_object(labels, predictions)
      gradients = tape.gradient(loss, model.trainable_variables)
      optimizer.apply_gradients(zip(gradients, model.trainable_variables))
      loss(loss)
      accuracy(labels, predictions)


tf.function
def test_step(images, labels):
  predictions = model(images)
  t_loss = loss_object(labels, predictions)
  loss(t_loss)
  accuracy(labels, predictions)


def train():
    EPOCHS = 5

    for epoch in range(EPOCHS):
      for images, labels in train_ds:
        train_step(images, labels)

      for test_images, test_labels in test_ds:
        test_step(test_images, test_labels)

      template = 'Epoch {}, Loss: {}, Accuracy: {}, Test Loss: {}, Test Accuracy: {}'
      print(template.format(epoch+1,
                        train_loss.result(),
                        train_accuracy.result()*100,
                        test_loss.result(),
                        test_accuracy.result()*100))

      # Reset the metrics for the next epoch
    train_loss.reset_states()
    train_accuracy.reset_states()
    test_loss.reset_states()
    test_accuracy.reset_states()
train()

当我替换了编译和fit函数时，它就能工作了。

model.compile(optimizer='adam',
              loss='sparse_categorical_crossentropy',
              metrics=['accuracy'])
model.fit(train_images, train_labels, batch_size= 200, epochs=6, validation_data=(test_images, test_labels))

为任何帮助拨出了大量资金。

Answer 1

在损失函数中设置from_logits=True。

tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True) = loss_object

它解决了错误！

Answer 2

我想您可以先使用input_shape param。

self.conv1 = tf.keras.layers.Conv2D(32，kernel_size =3，大步=2，激活=‘relu’，input_shape=(w，h，n_channel)