Shulou(Shulou.com)06/02 Report--

This article mainly explains "what is the TensorFlow static diagram and eager mechanism". The content in the article is simple and clear, and it is easy to learn and understand. Please follow the editor's train of thought to study and learn "what is the TensorFlow static diagram and eager mechanism".

Before the advent of eager, TensorFlow has been using the static diagram mechanism, in fact, until now the static diagram of TensorFlow is still popular. This article will introduce the static graph mechanism of TensorFlow.

What is a static / dynamic diagram?

A static diagram means that after the diagram is built, the model cannot be modified at run time. The "picture" here is the meaning of the model. Generally, a graph is a model. After the diagram is built, you need freeze before running, and then do some optimization, such as image fusion, etc., and finally start session to calculate according to the previously built static diagram, the calculation process can not be changed to the static diagram.

The dynamic diagram corresponds to the static diagram, which can be modified during the operation of the model. Friends who are familiar with PyTorch should know, because PyTorch uses the dynamic graph mechanism. In general, however, the model does not need to be modified during operation.

In addition, one of the disadvantages of TensorFlow static diagrams is that it is difficult to Debug. Because static diagrams will be optimized before running after freeze, some operation will be fused, so some operation will disappear, unable to provide breakpoints and single-step debugging during the calculation phase (the diagram construction phase can single-step debugging, but only show tensor, which is not useful). Of course, breakpoints and single-step debugging functions will also affect the efficiency of the program. While PyTorch adopts dynamic graph, which naturally has the advantages of Debug, provides the function of single-step debugging, and can view the values of all tensor in the calculation process, which is very intuitive and convenient.

Therefore, at present, PyTorch is widely used in academia, while in engineering, TensorFlow is more popular because of its advantages in deployment, comprehensive operation and good ecology. For individuals, it often depends on which framework the open source code of the selected application is based on in the industry, and to build their own applications on the basis of other people's open source code, both of which need to have a certain ability to get started quickly.

OK, I hope the static diagram mechanism of TensorFlow has been explained clearly above. If there is anything unclear, please leave a message or contact me. I will update it continuously. Later, I will add code instructions, easy to understand, please follow ~

Thank you for your reading, the above is the content of "what is the TensorFlow static diagram and eager mechanism". After the study of this article, I believe you have a deeper understanding of what the TensorFlow static diagram and eager mechanism is, and the specific use needs to be verified in practice. Here is, the editor will push for you more related knowledge points of the article, welcome to follow!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.