Federated model management
In Federated Model Management, you can centrally manage all saved models and view their information, including their storage addresses on each distributed node.
After you edit a model development script, you can execute a script task to generate a federated model. You can save the trained model in Model Management and distribute it to various nodes. Operations and maintenance (O&M) engineers at each node can then use the model's storage address to retrieve the federated model and deploy it on the node.
Prerequisites
Upon completion of the task execution, the federated model generated by the script is saved.
View the model list
Log on to the Federated Modeling console and select the desired project from the drop-down list in the upper-right corner.
In the navigation pane on the left, click Model Management (
) to open the Model Management page.The Model Management page displays information about all saved federated models.
Model Name: The name of the federated model. The creator specifies this name when saving the model. Click the model name to view the model's details.
Model Tag: Tags for the federated model. Detailed information helps with future maintenance. Click the Edit (
) icon to modify the model tag.From Task: The task that generated the federated model. Click the task name to view its details, such as execution logs, execution results, and the source script. For more information about tasks, see Task management.
Associated Pre-processing Rule: The name of the pre-processing rule associated with the federated model. If this field is empty, the model is not associated with any pre-processing rule.
Number of Participating Nodes: The number of nodes that participated in the development of this model.
Creator: The user who submitted the task that generated the model.
Creation Time: The time when the federated model was generated during the execution of the source task.
Save Time: The time when the federated model was saved from the Task Management menu. For more information about how to save a model, see Save a model.
View model details
In the federated model list, click a model name to view its details. These details include the model's storage address on each node, its input features, and its associated tasks.
Basic information
On the Basic Information tab, you can view the model's creator, save time, associated pre-processing rule, and the storage address of the model file on each node.
Input features
On the Input Features tab, you can view the input features supported by the model.
Associated tasks
On the Associated Tasks tab, you can view the associated script execution tasks, task start and end times, and information about downstream tasks. Click a task name to view its details, such as execution logs, execution results, and the source script. For more information about tasks, see Task management.
A downstream task is a task whose execution script uses the federated model.
Unsave a model
On the model details page, click the Unsave button to delete the model from Model Management.
After you unsave a model, the model files on the distributed nodes are not deleted.
Query for a federated model
If the federated model list contains many models, you can use the search function to find a specific model. You can search by Model Name, From Task, or Creator.
Model Name: In the search box in the upper-right corner, enter the name of the federated model and click the Search (
) icon.From Task: In the federated model list, click the (
) icon in the From Task column. In the search box that appears, enter the task name and click the Search button.Creator: In the federated model list, click the (
) icon in the Creator column, enter the creator's username in the search box that appears, and click the Search button.
The preceding search fields support fuzzy queries.