Access to neural network model

Question

We have built numerous diagnostic models which can be reduced to equations and code that will allow us to repeat the work. We have the code physically available to us, so it can be installed in our own software.

Now I would like to use artificial neural networks to build a prediction model. After I build that model, will I be able to take that model and transfer it to our own software environment? My concern is that the prediction model will just be a black box. Thanks

Answer 1

@OliverBathe-8330 Please follow the below Deployment scenarios. If possible can you please add more details about the use case.

Option A: Use the DevOps pipeline integration to rollout to production Using same approach as in the MLOps repo, set up a release trigger for your DevOps release pipeline listening from your dev workspace model registry but then deploy to your production workspace (requires registering again in Prod model registry, call model.deploy() in the Prod workspace

Option B: Use the AML pipeline to rollout to production Following same example as above, add additional PythonScriptStep in your AML pipeline to register and deploy model in the Production workspace

https://learn.microsoft.com/en-us/azure/machine-learning/how-to-deploy-and-where