列车功能内部和外部的结果不同

Question

我正在玩张量流2。我做了我自己的模型，类似于这里的做法。

然后我创建了自己的拟合函数。现在我得到了有史以来最奇怪的事情。以下是我进行测试的笔记本的精确复制/粘贴输出：

def fit(x_train, y_train, learning_rate=0.01, epochs=10, batch_size=100, normal=True, verbose=True, display_freq=100):

    if normal:

        x_train = normalize(x_train)  # TODO: This normalize could be a bit different for each and be bad.

    num_tr_iter = int(len(y_train) / batch_size)  # Number of training iterations in each epoch

    if verbose:

        print("Starting training...")

    for epoch in range(epochs):

        # Randomly shuffle the training data at the beginning of each epoch

        x_train, y_train = randomize(x_train, y_train)

        for iteration in range(num_tr_iter):

            # Get the batch

            start = iteration * batch_size

            end = (iteration + 1) * batch_size

            x_batch, y_batch = get_next_batch(x_train, y_train, start, end)

            # Run optimization op (backpropagation)

            # import pdb; pdb.set_trace()

            if verbose and (epoch * batch_size + iteration) % display_freq == 0:

                current_loss = _apply_loss(y_train, model(x_train, training=True))

                current_acc = evaluate_accuracy(x_train, y_train)

                print("Epoch: {0}/{1}; batch {2}/{3}; loss: {4:.4f}; accuracy: {5:.2f} %"

                      .format(epoch, epochs, iteration, num_tr_iter, current_loss, current_acc*100))

            train_step(x_batch, y_batch, learning_rate)

    current_loss = _apply_loss(y_train, model(x_train, training=True))

    current_acc = evaluate_accuracy(x_train, y_train)

    print("End: loss: {0:.4f}; accuracy: {1:.2f} %".format(current_loss, current_acc*100))

import logging

logging.getLogger('tensorflow').disabled = True

fit(x_train, y_train)

current_loss = _apply_loss(y_train, model(x_train, training=True))

current_acc = evaluate_accuracy(x_train, y_train)

print("End: loss: {0:.4f}; accuracy: {1:.2f} %".format(current_loss, current_acc*100))

现在我的问题是，我如何在最后2行上得到不同的值！？我在做同样的事情对吧？我在这里完全困惑。我甚至不知道如何谷歌这个。

Answer 1

所以问题只是愚蠢的。这是由于我在火车示例开始时所做的规范化操作！已将其删除并开始工作 正常。