Share via

Register a dataset as model in azure

Anonymous
2020-12-14T11:56:21.47+00:00

How is it possible to register a dataset or blob container directly as a model?

At the moment I have built a pipeline that outputs my trained models and registers them as datasets. I want to perform inference by deploying the models to for example AKS.

At the moment it appears the only way to register a model is by first downloading the dataset. Note I cannot register the model based on the runID because my model output is saved directly to blob storage through OutputFileDatasetConfig.

The motivation here is to avoid having to download and then upload large models.

Azure Machine Learning
Azure Machine Learning
An Azure machine learning service for building and deploying models.
2,922 questions
0 comments

1 answer

Sort by: Most helpful
Most helpful Newest Oldest
  1. romungi-MSFT 46,301 Reputation points Microsoft Employee
    2020-12-15T08:22:28.227+00:00

    @Anonymous You can try to register your model/file which is in the output directory of the run by using the azure ML cli command? Once you setup the configuration of your Azure account or workspace you can simply use this command with the run id and path of the output to register the model.

    Example:
    az ml model register -n sklearn_mnist --asset-path outputs/sklearn_mnist_model.pkl --experiment-name myexperiment --run-id myrunid --tag area=mnist

Your answer

Answers can be marked as Accepted Answers by the question author, which helps users to know the answer solved the author's problem.