WeichenXu

[FR] Support Mlflow client search API filter in programming way

1

### Willingness to contribute Yes. I would be willing to contribute this feature with guidance from the MLflow community. ### Proposal Summary Example: `run_id in a_list and tag.xx == yy`,...

### Description Add config option for Ray Node Runtime to specify visible cuda devices on a node: Suppose on a machine, there are 4 GPUs, and GPU-0/ GPU-1 are used...

[pyspark] Make Xgboost estimator support using sparse matrix as optimization

1

Closes https://github.com/dmlc/xgboost/issues/8108 Make Xgboost estimator support using sparse matrix as optimization.

Xgboost spark: Support XGBRanker

6

Xgboost spark: Support XGBRanker

Xgboost pyspark: Support pyspark Sparse Vector

4

Xgboost pyspark: Support pyspark Sparse Vector

Xgboost pyspark: Support external memory mode

2

Xgboost pyspark: Support external memory mode

Apply dmatrix iteration iterface in PySpark xgboost and support external memory mode

3

Apply dmatrix iteration iterface in PySpark xgboost and support external memory mode.

Signed-off-by: Weichen Xu Upgrade xgboost dependency to be 1.6.1 version

org.apache.spark.ml.bundle.tree.decision.SparkNodeWrapper only support "gini" impurity ("gini" is hardcoded), we should update code to support all kinds of impurity. Currently, any type node in mleap will be regarded as "gini" impurity.

NaiveBayes model support Gauss type. We need update code in `ml.combust.mleap.core.classification.NaiveBayesModel` to make it also support Gauss type.