fit()和image_dataset_from_directory()中batch_size的区别是什么? [英] whats the difference in batch_size in fit() and image_dataset_from_directory()?

问题描述

我尝试在训练时更改批次大小，因此我在循环中运行fit方法并更改批次大小，但是我们已经在 image_dataset_from_directory()中定义了批次大小，如果我放不同这两个函数中的批处理大小，并且有一种方法可以让我用fit方法控制批处理大小，而不受 image_dataset_from_directory()的影响这是我的代码

I was trying to change batch size while training so I run the fit method in a loop and change the batch size, but we already define batch size in image_dataset_from_directory(), What if I put different batch sizes in these two functions, and is there a way that I get to control the batch size in fit method and not be affected by image_dataset_from_directory() Heres my code

  train_ds = tf.keras.preprocessing.image_dataset_from_directory(
  data_dir,
  validation_split=0.2,
  subset="training",
  seed=123,
  image_size=(img_height, img_width),
  batch_size=batch_size)

  model.fit(
      
            train_ds,
            validation_data=val_ds,
            epochs=epoch,
            batch_size=batch_size,
            steps_per_epoch=10,
            callbacks=call

            )

代码已从此张量流的教程中提取

the code has been taken from this tutorial of tensorflow

推荐答案

image_dataset_from_directory 是生成器，因此在 model.fit()中指定 batch_size 代码>不会执行任何操作.请参阅 model.fit()上的文档:

image_dataset_from_directory is a generator and so specifying batch_size in model.fit() will do nothing. See the docs on model.fit():

batch_size 整数或无.每个梯度更新的样本数.如果未指定，batch_size将默认为32.如果您的数据采用数据集，生成器或keras.utils.Sequence实例的形式(因为它们生成批次)，则不要指定batch_size.

batch_size Integer or None. Number of samples per gradient update. If unspecified, batch_size will default to 32. Do not specify the batch_size if your data is in the form of datasets, generators, or keras.utils.Sequence instances (since they generate batches).

在您所引用的示例中可以看到，Tensorflow教程未在 model.fit()中指定 batch_size :

As you can see in the example you referred, the Tensorflow tutorial doesn't specify batch_size in model.fit():

epochs=10
history = model.fit(
  train_ds,
  validation_data=val_ds,
  epochs=epochs
)

批次大小将是您在生成器中指定的大小，并且在 model.fit()

The batch size will be what you specified in the generator, and the argument will be ignored in model.fit()

这篇关于fit()和image_dataset_from_directory()中batch_size的区别是什么?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！