[Roadmap] 1.3.1 Patch release

[hcho3]

We are looking to release a patch release 1.3.1, primarily to incorporate pull request #6517. This patch is important to ensure that models from versions earlier than 1.0.0 can be read in the latest XGBoost.

I'd like to know if there are other pull request that should go into 1.3.1.

This issue will remain open until December 21, 2020.

@dmlc/xgboost-committer

Pull requests

• Enable loading model from <1.0.0 trained with objective='binary:logitraw' (Enable loading model from <1.0.0 trained with objective='binary:logitraw' #6517)
• Fix handling of print period in EvaluationMonitor (Fix handling of print period in EvaluationMonitor #6499)
• Move metric configuration into booster (Move metric configuration into booster. #6504)
• Fix save_best early stopping option ([backport] Fix for save_best. #6523)
• Remove cupy.array_equal, since it's not compatible with cuPy 7.8 (Remove cupy.array_equal, since it's not compatible with cuPy 7.8 #6528)

[trivialfis]

I will extract part of #6506 into 1.3 to fix #6514 .

[trivialfis]

@hcho3 I have PRs for all patches needed for back porting. Hopefully the holiday can go smoothly. Please let me know if there's anything I can help with the patch release.

[trivialfis]

@hcho3 Do you need help on updating the rel page?

[hcho3]

@trivialfis I've uploaded the package to PyPI now. If you'd like, why don't you go ahead and put up a new Release entry?

[trivialfis]

Let me do that tonight.

[trivialfis]

If you'd like, why don't you go ahead and put up a new Release entry?

Done. @hcho3 Thanks for steering the patch release.