[RFC] 1.3.0 Release Candidate

[hcho3]

Roadmap: #6031

We are about to release version 1.3.0 of XGBoost. In the next two weeks, we invite everyone to try out the release candidate (RC).

Feedback period: until the end of December 7, 2020. No new feature will be added to the release; only critical bug fixes will be added.

@dmlc/xgboost-committer

Now available

Python package. RC1 available on PyPI. Try it out with the command

python -m pip install xgboost==1.3.0rc1

• R package. RC1 available from the Releases section. Download the tarball file xgboost_1.3.0.1.tar.gz and run

R CMD INSTALL xgboost_1.3.0.1.tar.gz

• JVM packages. RC1 available from our Maven repository. Add XGBoost4J as dependency to your Java application.

Show instructions (Maven/SBT)

Maven

<dependencies>
  ...
  <dependency>
      <groupId>ml.dmlc</groupId>
      <artifactId>xgboost4j_2.12</artifactId>
      <version>1.3.0-RC1</version>
  </dependency>
  <dependency>
      <groupId>ml.dmlc</groupId>
      <artifactId>xgboost4j-spark_2.12</artifactId>
      <version>1.3.0-RC1</version>
  </dependency>
</dependencies>

<repositories>
  <repository>
    <id>XGBoost4J Release Repo</id>
    <name>XGBoost4J Release Repo</name>
    <url>https://s3-us-west-2.amazonaws.com/xgboost-maven-repo/release/</url>
  </repository>
</repositories>

SBT

libraryDependencies ++= Seq(
  "ml.dmlc" %% "xgboost4j" % "1.3.0-RC1",
  "ml.dmlc" %% "xgboost4j-spark" % "1.3.0-RC1"
)
resolvers += ("XGBoost4J Release Repo"
              at "https://s3-us-west-2.amazonaws.com/xgboost-maven-repo/release/")

Starting from 1.2.0, XGBoost4J-Spark supports training with NVIDIA GPUs. To enable this capability, download artifacts suffixed with -gpu, as follows:

Show instructions (Maven/SBT)

Maven

<dependencies>
  ...
  <dependency>
      <groupId>ml.dmlc</groupId>
      <artifactId>xgboost4j-gpu_2.12</artifactId>
      <version>1.3.0-RC1</version>
  </dependency>
  <dependency>
      <groupId>ml.dmlc</groupId>
      <artifactId>xgboost4j-spark-gpu_2.12</artifactId>
      <version>1.3.0-RC1</version>
  </dependency>
</dependencies>

<repositories>
  <repository>
    <id>XGBoost4J Release Repo</id>
    <name>XGBoost4J Release Repo</name>
    <url>https://s3-us-west-2.amazonaws.com/xgboost-maven-repo/release/</url>
  </repository>
</repositories>

SBT

libraryDependencies ++= Seq(
  "ml.dmlc" %% "xgboost4j-gpu" % "1.3.0-RC1",
  "ml.dmlc" %% "xgboost4j-spark-gpu" % "1.3.0-RC1"
)
resolvers += ("XGBoost4J Release Repo"
              at "https://s3-us-west-2.amazonaws.com/xgboost-maven-repo/release/")

Breaking changes:

• Starting from 1.3.0 release, XGBoost no longer supports MacOS 10.13 (High Sierra).

Known issue:

• Python packaging does not yet work with MacOS 11.0 (Big Sur): add support for macos 11.0, arm64, universal2 pypa/packaging#319. Users of Big Sur cannot use the binary wheel and will need to build XGBoost from the source.

Outstanding patches that should make it into the 1.3.0 release:

To be merged after RC1:

• [R] Fix R package installation via CMake #6423
• Thread local memory allocation for BuildHist #6358
• [CI] Upgrade cuDF and RMM to 0.17 nightlies #6434
• Fix CLI ranking demo. #6439
• Fix period = 1 in evaluation monitor. #6441
• Enforce row-major order in cuPy array #6459
• Fix broken links. #6455
• Optionaly fail when gpu_id is set to invalid value #6342
• Vendor libgomp in the manylinux Python wheel #6461
• Fix filtering callable objects in skl xgb param. #6466
• [CI] Fix CentOS 6 Docker images #6467
• Fix dask ip resolution. #6475

[trivialfis]

Is there an update on github release page?

[7starsea]

I hope that the hidden api XGBoosterPredictFromDense could be included in the file c_api.h and we can clearly document that cache_id must be zero currently.
As I learned from python api inplace_predict, XGBoosterPredictFromDense is very useful for multiple threads prediction in c/cpp application.

[trivialfis]

It will be in the next release.

[terrytangyuan]

"Starting from 1.3.0 release, XGBoost no longer supports MacOS 10.13 (High Sierra)."

This won't be a deprecation notice though since we are dropping the support in this release.

[hcho3]

@trivialfis I put up the release page for 1.3.0 RC1.
@terrytangyuan Thanks. Revised it to say "Breaking changes."

[hcho3]

Should #6342 be part of 1.3.0? @trivialfis