Xgboost: [Roadmap] 1.3.0 Roadmap

[trivialfis]

@dmlc/xgboost-committer Please add your items here by editing this post. Let's ensure that

• Each item has to be associated with a ticket
• Major design/refactoring are associated with a RFC before committing the code
• Blocking issue must be marked as blocking
• Breaking change must be marked as breaking

For other contributors who have no permission to edit the post, please comment here about what you think should be in 1.3.0.

• [x] Merge rabit into XGBoost (#5995, #6023)
• [x] GPU Shap value compuation. (#6038, #6064, #6163)
• [x] Let Spark package fail gracefully (#4826, #6019)
• [x] Support for more different meta on dask. (#6130, #6132)
• [x] Support MAPE. (#6119)
• [x] GPU predict leaf function (#5847, #6187)
• [x] New callback functions (#5612, #6199)
• [x] Early stopping on dask. (#6199)
• [x] Initial support for categorical data. (#5949, #6137, #6140, #6164, #6165, #6166, #6179, #6194, #6219)
• [x] Feature weights for column sampling (exact, hist, gpu_hist) (#5962)
• [x] Slicing model. (#5531, #4052, #6302)
• [x] Default to JSON for memory snapshot. (#6027)
• [x] Support dask on reverse proxy environment. (#5765, #5408, #6142, #6343)
• [x] Support using system installation of libxgboost.so when distributing Python package. (#6362)

[trivialfis]

Moving:

• [x] Support dask on reverse proxy environment. (#5765, #5408, #6142)

off the list again. Getting it work on GKE is a bit tricky. Related:

• dask kubernetes: https://kubernetes.dask.org/en/latest/
• kubernetes
• docker image deployment
• GKE
• RABIT internal

[trivialfis]

The item is moved back in now that I can run tests myself.

[trivialfis]

RC1 is targeted at Nov 20.