Install TensorFlow Serving Using the Ready-to-Use Application
Table of Contents
TensorFlow Serving is designed to provide high-performance serving of machine learning models in a production environment. After using TensorFlow to train a model, you can use the TensorFlow Serving APIs to respond to client input. TensorFlow Serving is available as a ready-to-use application which can be automatically installed on a 1&1 Cloud Server when it is built. This stack comes with Inception v3 with trained data for image recognition, but it can be extended to serve other models.
Other installation options:
For more information on different installation methods for TensorFlow Serving, see the official documentation.
Install TensorFlow Serving
Log in to the Cloud Panel then go to Infrastructure > Servers.
Click + Create to create a new server.
Set a name for the server, and click the server configuration you wish to use.
Click the Applications tab.
Click the Search icon, and search for TensorFlow.
Click TensorFlow Serving.
Click Create to build the server.
The server will be built with a default Firewall Policy named TensorFlow Serving. This Firewall Policy does not allow SSH connections through port 22.
If you wish to connect to the server with SSH, you can modify the firewall policy under Network > Firewall Policies. For step-by-step instructions, see our article Editing a Firewall Policy.