Kai Fricke
Kai Fricke
To use the patch from #28326 / #28327 without upgrading Ray, here is a workaround: https://github.com/ray-project/ray/issues/28326#issuecomment-1240448217
XGBoost allows setting the number of threads, so is able to parallelize across CPUs natively. Choosing more CPU parallelism instead of workers is probably a matter of taste, but one...
This is a good topic to discuss in the documentation. We should be looking into adding these, soon. Closing this for now - please let me know if you have...
Hi @eedalong, we do test elastic and non-elastic fault tolerance in the unit tests and also in ray release tests, so it should generally work. Do you have an example...
Hi @chaokunyang, I could not reproduce this error (it passes for me), which version of Ray, XGboost-Ray, and XGBoost are you using?
Hi @Matrix-World, we're preparing a short paper/poster for XGBoost-Ray anyway, so we might end up doing something like this eventually. Is there something that you're particularly interested in and find...
`gpus_per_actor` assigns GPUs to the remote worker - but you will have to tell xgboost to actually use GPUs. You can do that by passing `"tree_method": "gpu_hist"` in the xgboost...
As for the paper, it will probably take a couple of weeks
Thanks @julietnasar that would be great!