I use single-instance xgboost model to train, which works fine with mlflow to log all the model-related parameters by using mlflow.xgboost.autolog(), but when i change to distributed xgboost version changing from python package to JVM package by including the xgboost4j.jar and xgboost4j-spark.jar files, and also include mlflow module into it, (mlflow.xgboost.autolog()). The mlflow cannot show all the parameters on the page. They are empty.
So I looked at the source code in the mlflow.xgboost, in line 271, def autolog(importance_types=["weight"]): says it imports the xgboost package, which i think is the single-instance xgboost model, I wonder if it is support the distributed version? Or is there any other methods to solve the problem?