[pyspark] Add param validation for "objective" and "eval_metric" param, and remove invalid booster params #8173
Conversation
WeichenXu123
changed the title
|
Please fix the formatting error: https://github.com/dmlc/xgboost/runs/7824147126?check_suite_focus=true . |
|
Any update? |
| @@ -267,6 +267,18 @@ def _validate_params(self): | |||
| "If features_cols param set, then features_col param is ignored." | |||
| ) | |||
|
|
|||
| if self.getOrDefault(self.objective) is not None: | |||
There was a problem hiding this comment.
It does not make sense (to me) to check the type of a parameter that already has the typeConverter
There was a problem hiding this comment.
The "objective" param is automatically generated, currently it does not have type converter.
There was a problem hiding this comment.
I still hope we can find a common solution for the type converter in the following PR
Signed-off-by: Weichen Xu <weichen.xu@databricks.com>
Signed-off-by: Ubuntu <weichen.xu@ip-10-110-30-183.us-west-2.compute.internal>
python-package/xgboost/spark/core.py
Outdated
| if not gpu_per_task: | ||
| # For spark cluster with GPU config, the "spark.task.resource.gpu.amount" | ||
| # config might not be set. In this case, the default gpu per task is 1. | ||
| # If running on no GPU available, in spark task calling `_get_gpu_id` will | ||
| # raise gpu not found error. | ||
| gpu_per_task = "1" |
There was a problem hiding this comment.
@wbo4958 A notable fix.
There was a problem hiding this comment.
My fault, reverted this.
|
CC @trivialfis @wbo4958 Done. |
|
@trivialfis @wbo4958 We can merge this. The test failures are not related to my PR. |
xgboost.trainspark.task.resource.gpu.amountconfig is not set.