Elastic Algorithm Service (EAS) provides built-in processors for common frameworks such as PMML and XGBOOST. You can also build a custom processor.
How it works
Prepare model files, processor files, and other deployment artifacts, and upload them to a cloud storage service such as Object Storage Service (OSS) or NAS. EAS deploys the service by mounting the storage service to access these artifacts.
EAS processor deployment workflow:
Key points of processor deployment:
-
PAI provides built-in processors for common use cases. To meet specific business requirements, develop custom models and processor files, and then upload them to a cloud storage service such as OSS or NAS.
-
Store model and processor files separately. During deployment, configure the model mount path. In the processor file, use the get_model_path parameter to retrieve the configured model path. This simplifies model replacement when frequent updates are required.
-
When a service is deployed with a processor, EAS pulls the official environment image for the selected inference framework. Based on the processor file, it deploys an HTTP server to handle service requests.
Image deployment is recommended. Processor deployment requires the model inference framework and processor file to meet specific environment requirements. This method is less flexible than image deployment.
Parameters
Parameters specific to processor-based deployment in the console. For other parameters, see Custom Deployment.
|
Parameter |
Description |
|
Model Settings |
Configure the model file by using one of the following methods:
|
|
Processor Type |
Built-in and custom processors are both supported. For more information about built-in processors, see Built-in processors.
|