绘制的Keras模型没有给出尺寸 [英] Keras model to plot gives none, dimensions

查看:88
本文介绍了绘制的Keras模型没有给出尺寸的处理方法,对大家解决问题具有一定的参考价值,需要的朋友们下面随着小编来一起学习吧!

问题描述

我正在尝试使用Keras绘制VGG16架构图.但是,我看到下面的图像带有None,有些尺寸. None是什么意思?

I am trying to plot VGG16 architecture diagram using Keras. However, i see the below image with None, some dimension. What does None mean, is that a bug ?

from keras.utils import plot_model
model = VGG19(include_top=True, weights='imagenet')
#print(model.summary())
plot_model(model, to_file='model_Vgg19.png',show_shapes=True)

我可以看到这是Keras中的一个已知问题: https://github .com/keras-team/keras/issues/1877 .

I can see that this is a known issue in Keras: https://github.com/keras-team/keras/issues/1877.

推荐答案

形状元组中的None维度指的是批处理维度,这仅表示该层可以接受任何大小的输入.例如,数据集(例如MNIST)可能具有以下形状元组(60000, 28, 28, 1),但输入层的形状将为(None, 28, 28, 1).

The None dimension in the shape tuple refers to the batch dimension which simply means that the layer can accept input of any size. For example, a dataset (e.g. MNIST) could have the following shape tuple (60000, 28, 28, 1) but the input layer's shape would be (None, 28, 28, 1).

参考文献:

  • keras-team/keras#2054
  • Keras Documentation - Core Layers

这篇关于绘制的Keras模型没有给出尺寸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!

查看全文
登录 关闭
扫码关注1秒登录
发送“验证码”获取 | 15天全站免登陆