All Products
Search
Document Center

How to Deploy Model in EAS

Last Updated: Apr 03, 2020

Elastic Algorithm Service (EAS) of Machine Learning Platform for AI (PAI) is seamlessly integrated with PAI Studio and PAI Data Science Workshop (DSW). This allows you to deploy models as RESTful APIs in multiple ways. For more information about the billing of model services, click EAS billing.

You can use the following methods to deploy models:

  1. Upload models to the console
  2. Use PAI Studio to deploy models
  3. Use DSW to deploy models
  4. Use the EASCMD client to deploy models

Upload models to the console

  • In the PAI EAS console, you can upload a trained model and package the uploaded model in the .zip format. Therefore, you can deploy a model service in a few steps.
  • If you are using a RAM user, you must first use the Alibaba Cloud account to authorize the RAM user. For more information about how to authorize the RAM user, see Authorize RAM users.
  • Click Upload and Deploy Models, and then follow the steps shown in the following figure to upload and deploy the model.consoleupload

Use PAI Studio to deploy models

  • PAI Studio is a visual modeling product that allows you to drag and drop components to build models. For more information about PAI Studio or to use it, log on to the PAI - Modelling Visualization. After you train a model in Visualization Studio, you can deploy the model to EAS.
  • The algorithms that support model deployment include GBDT binary classification, linear SVM, logistic regression for binary classification, logistic regression for multiclass classification, random forest, K-means, linear regression, GBDT regression, and TensorFlow. GBDT regression does not support input data of the integer type. Before you deploy the model, make sure that the data input to the GBDT regression component is of the double type.image.png

Use DSW to deploy models

DSW is an interactive deep learning development environment on Alibaba Cloud. It provides high-performance GPU modules and an interactive programming environment. For more information about DSW or to use it, log on to the PAI DSW console. DSW is built in with EASCMD, which allows you to deploy a model to EAS after you train the model in DSW.

  • Check your AccessKey ID and AccessKey secret on the User Management page. If you are using a RAM user, you must use your Alibaba Cloud account to log on to the RAM console, select Users, and click the RAM user to check the AccessKey information of the RAM user.
  • To deploy a model in the DSW terminal, you only need to enter one command line. For more information about the model deployment command, see EASCMD client manual.

Use the EASCMD client to deploy models

  • The EASCMD client allows you to manage online model services on your server. You can create services, delete services, modify services, and view service status. You can download the EASCMD client from the following links:
  • To deploy a model, you must provide your AccessKey ID and AccessKey secret for authentication. You can view your AccessKey information on the User Management page.
  • For more information, see EASCMD client manual.