reminisce commented on issue #11119: Feature request: please speed up asnumpy() URL: https://github.com/apache/incubator-mxnet/issues/11119#issuecomment-393767775 After looking through the script, I feel it's not the problem of `asnumpy()`. The mxnet operators are async functions in python, which means they may be still running in backend (C++) while python main thread moves to other steps. I suspect that the bottleneck is in the nested loops where ndarray slicing and slicing with assignment are triggered frequently. You can use `tpreds.wait_to_read()` to setup a barrier in the python thread to force the execution on `tpreds` to finish in backend before moving forward in frontend and time the execution.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
