dmlc / xgboost

Scalable, Portable and Distributed Gradient Boosting (GBDT, GBRT or GBM) Library, for Python, R, Java, Scala, C++ and more. Runs on single machine, Hadoop, Spark, Dask, Flink and DataFlow
https://xgboost.readthedocs.io/en/stable/
Apache License 2.0
26.28k stars 8.73k forks source link

On the GPU implementation of the survival:cox objective function #8403

Open hxu296 opened 2 years ago

hxu296 commented 2 years ago

Hello, I noticed that the "GPU Support" section of the XGBoost documentation has recently been updated to exclude the table of objective functions with GPU implementations.

From the previous documentation, it seems the only objective function without GPU implementation is survival:cox. I checked the current implementation of CoxRegression in src/objective/regression_obj.cu, and it seems its GetGradient function still doesn't make use of GPU parallelism.

As someone who would like to work on implementing GPU support for survival:cox, I wonder whether the documentation change reflects a shift of development focus, or it is still encouraged to try to implement GPU support for survival:cox?

Thank you!

image

Figure: previous XGBoost documentation

trivialfis commented 2 years ago

We have new survival training objectives, see our tutorial site for more info.