utils.add_libraries_to_model creates a duplicated model

[pablobd]

Hello,

When I call this function,

mlflow.models.utils.add_libraries_to_model(MODEL_URI)

It register a new model into the Model Registry. Is it possible to do the same but without registering a new model?

Thanks,

[Kaniz_Fatma]

Hi @pablobd, The mlflow.models.utils.add_libraries_to_model function is used to re-log a model along with all the.... This function is typically used when you want to package your custom libraries alongside the model i....

However, this function requires a registered model URI (e.gmodels:/<model_name>/<model_version>) as .... 

This means that the model needs to be registered in the Model Registry before you can use this function.

If you want to add libraries to a model without registering it, you might need to manually package the model along with its dependencies. This could involve saving the model and its dependencies to a file, and then loading them when you want to use the model. Please note that this approach would not leverage the MLflow Model Registry and its versioning capabilities.

[pablobd]

Thanks @Kaniz_Fatma - that's exactly how I am using it, following this: https://docs.databricks.com/en/machine-learning/model-serving/private-libraries-model-serving.html

What I don't like is that in step 2 the model is registered then in step 3, the same model (with the compiled package is registered again). So, I endup with two copies of the same model.

[pablobd]

I ended up publishing the library to AWS CodeArtifact repository. Now, how can I tell MLFlow to use AWS CodeArtifact private repository instead of PyPi?