Closed BlueFelix closed 4 years ago
I use 2 executors and run the xgboost demo. I tried to change spark.sql.files.maxPartitionBytes and maxRowsPerChunk but it always has 1 partition and run on single executor and single GPU. How to use multiple GPU?
It's controled by numWorkers of xgboost's params
How many data partitions do you have? You need to specify 2 executors and 2 workers.
Was this issue resolved (numWorkers = number of GPUs)?
I use 2 executors and run the xgboost demo. I tried to change spark.sql.files.maxPartitionBytes and maxRowsPerChunk but it always has 1 partition and run on single executor and single GPU. How to use multiple GPU?