HuoguoML Model Registry
Guide for managing your machine learning models
Introduction
After experimenting, you want to select a model and make it accessible for further use. For this, you can use the HuoguoML Model Registry, which is the central repository for all your models.
Register Model
Open the HuoguoML dashboard and go to an experiment run. In the upper right corner, you will see a register button if the model experiment has logged model files (for logging model files see HuoguoML Tracking API). Click the button and enter the name of your registry or select the name of an existing registry:
On the dashboard, all registries are visible in a single table:
Feature of HuoguoML Registry is that each registered model has its own unique version with a stage. A model can be either in stage none, staging or production. In addition, there can be only one model in production and staging. To get a more detailed view of all models within a registry, you can click on the model:
To transfer a model to another stage, you must go to a model and click on the Change Stage button. You will be forwarded to a new page, where you can select the new stage
Access Model Registry
The model files can be retrieved using HTTP GET requests. Below are examples that show the available API endpoints.
Model files by version:
Model files by tag (production or staging):
For all endpoints, you will receive a ZIP file containing the logged files under the model directory. In the root directory you will find huoguoml_meta, which is a YAML file with the details of the source run.
Last updated
