ML Model - Container Resource¶
The EnOS Enterprise Analytics Platform’s MI Hub provides a series of smart asset distribution centers for smart model users and data scientists, and provides a full model registration process and hosting services for model developers.
When a model is developed from an authoring laboratory or in a third-party system, the model developer can deploy the model to a production environment for prediction tasks. Based on user model update requirements, users can deploy multiple models to the production environment at the same time through canary or blue/green deployment. The MI Hub provides model developers with scientific and effective version management tools which can be shared with end users and other collaborators in a safe and controlled environment. Smart assets that other developers have explored or created can also be reused.
For more information, see MI Hub Overview.
Resource Application Scenario¶
Before using MI Hub to deploy machine learning models, you need to apply for the ML Model - Container resource.
Note
The maximum number of resources that can be applied for is 5.
Resource Specification¶
When deploying a model, you can select the CPU, RAM, and storage according to the actual business requirements, and apply for the corresponding resources to support the use of the production functions.
Specification |
Description |
|---|---|
CU |
1 CU = 1 Core CPU + 16 GB Memory. Available options are 2 - 320 CU. |
Resource Type |
Select Primary Partitions or Subpartitions.
|
RAM |
Available options are 5 - 2,000 GB. |
Permissions |
Gives the read access for HDFS and Data Warehouse when enabled. |
Note
ML model resources can only be installed on one requested resource pool under the same OU, and the Dev Console can operate on other resource pools under that OU. The ML model resource splits the resource schema into primary partitions and subpartitions, and the EAP Dev Console must be installed under the namespace associated with the primary partition.