Reputation: 2880
I've been testing a model using azureml for image classification. I see in this guide that it should be possible to view the underlying code: https://learn.microsoft.com/en-us/azure/machine-learning/v1/how-to-generate-automl-training-code?view=azureml-api-1&viewFallbackFrom=azureml-api-2
It's creating a seresnext model but when I go to the model and view generated code I just see some automl configs in script.py (automl_settings, mltable_data_json). I'd like to view the actual model code so I can adjust it in my own azureml compute session. Something like the generate_algorithm_config in the above tutorial. Is this possible?
Upvotes: 0
Views: 221
Reputation: 132
You can view the training code in the Azure Machine Learning studio UI on the Outputs + logs tab of the selected model. The file script.py
contains the model's training code that you may want to analyze with the featurization steps, specific algorithm used, and hyperparameters.
“AutoML CodeGen” Preview documentation and feedback/issues channel: GitHub repo: Azure/automl-codegen-preview: AutoML code generation (github.com)
Upvotes: 0