Removing logging messages in ray.train

Ray Libraries (Data, Train, Tune, Serve)
1

when running ray.train, there are lots of logging messages like:

  RayTrainWorker pid=131328) Starting execution of Dataset. Full logs are in 
 /tmp/ray/session_2024-07-03_17-46-11_385350_130700/logs/ray-data [repeated 58x 
 across cluster]
  (RayTrainWorker pid=131328) Execution plan of Dataset: InputDataBuffer[Input] -> 
  TaskPoolMapOperator[Write] [repeated 46x across cluster]                        
  (RayTrainWorker pid=131328) Execution plan of Dataset: InputDataBuffer[Input] -> 
 TaskPoolMapOperator[ReadCSV] [repeated 12x across cluster]

I tried to turn them off by running the following commands but they make no effects:

    ray_data_logger = logging.getLogger("ray.data")
    ray_rllib_logger = logging.getLogger("ray.rllib")
    ray_serve_logger = logging.getLogger("ray.serve")
    ray_data_logger.setLevel(logging.CRITICAL)
    ray_rllib_logger.setLevel(logging.CRITICAL)
    ray_serve_logger.setLevel(logging.CRITICAL)

Any help would be appreciated

2

Since you’re running ray.train you have to set the ray_train_logger (you set everything else done EXCEPT for Train)

IE:
ray_train_logger = logging.getLogger(“ray.train”)
ray_train_logger.setLevel(logging.WARNING) # or logging.ERROR for even fewer logs

3

Adding “ray.train” logger as suggest doesn’t silent those messages. Further inputs would be appreciated.