Machine Learning Designer is seamlessly integrated with Elastic Algorithm Service (EAS). After you train and evaluate a model offline in Machine Learning Designer, you can deploy the model to EAS as an online service. This topic describes how to deploy a model that is trained in Machine Learning Designer to EAS as an online service.
Prerequisites
A model is trained and the accuracy of the model is verified. For more information, see Model training.
Use one-click deployment
Algorithms that support one-click deployment
Component | Generated model format | EAS processor | Description |
Logistic Regression for Binary Classification | PMML | PMML | Before you train a model, you need to select Whether to Generate PMML on the Fields Setting tab of the component. |
GBDT Binary Classification | PMML | PMML | |
Linear SVM | PMML | PMML | |
Logistic Regression for Multiclass Classification | PMML | PMML | |
Random Forest | PMML | PMML | |
Naive Bayes | PMML | PMML | |
K-means Clustering | PMML | PMML | |
GBDT Regression | PMML | PMML | |
Linear Regression | PMML | PMML | |
Scorecard Training | PMML | PMML | |
Text Summarization | TGZ package | EasyNLP | EasyNLP provided by Platform for AI (PAI) in a public Object Storage Service (OSS) bucket is automatically specified. |
image classification (torch) | TGZ package | EasyCV | EasyCV provided by PAI in a public OSS bucket is automatically specified. |
PyAlink Script | AlinkModel | Alink | For more information, see PyAlink Script. |
XGboost Train | XGBoost | XGBoost | For more information, see XGboost Train. |
For components that are used to train a PMML model, you need to perform the steps in the following figure to select Whether to Generate PMML on the Fields Setting tab of the component and rerun the corresponding node.
Procedure
Go to the Machine Learning Designer page.
Log on to the PAI console.
In the left-side navigation pane, click Workspaces. On the Workspaces page, click the name of the workspace that you want to manage.
In the left-side navigation pane, choose to go to the Machine Learning Designer page.
On the Pipelines tab, double-click the pipeline that you want to manage.
In the upper-left corner of the canvas, click Models.
The system automatically detects the trained model on the canvas, matches the model with a processor, and then redirects you to the EAS-Online Model Services page to complete the deployment. For more information, see Deploy a model service by using Machine Learning Designer.
Manually deploy a model
After you use the algorithm components in the following table to train a model, you need to use the Model Export component to assemble the model, and export the model to an OSS bucket, and then manually deploy the model. This method does not support one-click deployment.
Component | Generated model format | EAS processor | Manual deployment process |
PS-SMART Binary Classification Training | PS | PS algorithm | Connect the output port of the component to the Model Export component. |
PS-SMART Multiclass Classification | |||
PS-SMART Regression |
For more information about how to manually deploy a model after you export the model to an OSS bucket, see Model service deployment by using the PAI console.
FAQ
What do I do if some nodes are dimmed and cannot be selected when I deploy a model in one-click mode?
Open the Fields Setting tab of the component on the right side, select Whether to Generate PMML, and then rerun the corresponding node. For more information, see Algorithms that support one-click deployment.
References
You can go to the EAS-Online Model Services page to view the status of the deployed services or manage services. For more information, see Manage online model services in EAS.
After you deploy a model service, you can use the Update EAS Service (Beta) component provided by Machine Learning Designer to update the service on a regular basis. For more information, see Periodically update online model services.
Machine Learning Designer allows you to deploy a batch data processing pipeline to EAS as an online service after you package the pipeline as a model. For more information, see Deploy a pipeline as an online service.
You can use the prediction component provided by Machine Learning Designer to perform batch offline prediction for a model. For more information, see Implement batch prediction.