I need to register the model in unity catalog in Databricks. So, i trained the model with lightgbm.

from synapse.ml.lightgbm import LightGBMClassifier, LightGBMClassificationModel

Also, I registered the model in unity catalog as well using Mlflow. However, the problem is model inbuilt attribute is missing. I used mlflow.spark.log_model() and mlflow.pyfunc.log_model() to log & load model from unity catalog.

Except spark & pyfunc, none of the flavor didn’t support with mlflow to register the model. Finally, the issue is half of model attribute is missing when we print the model after loading from UC.

Example: I can see full list print(dir(LightGBMClassificationModel)) but half of attribute is missing when i see after loading. model = mlflow.spark.load_model(model_uri) then print(dir(model)). missing attribute example is “getFeatureImportances”.

I checked version and runtime. However, I am not able to sort this issue.

I need to get all inbuilt function(attribute) when loading the model from Unity catalog. can anyone please advise or guide me on this issue.