[TODO List] Ant-XGBoost on sqlflow

[sperlingxx]

This is TODO list for milestone 1; running Ant-XGBoost locally on SQLFlow server.

SQLFlow

• design doc add design doc: xgboost on sqlflow #579
• [Dockerfile] Install ant-xgboost in SQLFlow Docker image . xgboost/sqlflow_data_source #647
• [python] SqlflowDataSource xgboost/sqlflow_data_source #647
• xgboost template setup xgboost on sqlflow #662
• A SQLFlow tutorial to run in notebook add user guide for ant-xgboost #772
• user guide add user guide for ant-xgboost #772
• SQLFlow magic supports to switch submitter in notebook, need to add SQLFLow_submitter field in the Session struct. setup xgboost on sqlflow #662
• [golang] XGBoostFiller refine. xgboostFiller refine #651
• Integrate model performance verification in CI. add training metrics verification for antxgb #767

ant-xgboost

• xgblauncher
  • move inner codes to Github
  • add README and examples
  • support pip install
  • improvements

[Yancey0623]

Some supplement:

• An SQLFlow + XGBoost example on the notebook, I think it's an e2e example, maybe we can reuse the iris dataset.
• Can switch SQLFlow_submmiter in the notebook.

[python] SqlflowModelSource
If only support local mode, we can use LocalModelSource instead.

I believe we only support local training in this milestone, how about consider dist mode in the next milestone?

[sperlingxx]

@Yancey1989 @Echo9573 @typhoonzero @tonyyang-svail @BlackPoint-CX @weiguoz
Since all tasks has been done. Can we close this TODO list?

[typhoonzero]

@sperlingxx Sure.