wkcn edited a comment on issue #15624: asnumpy() took a lot of time when converting NDARRAY into numpy? URL: https://github.com/apache/incubator-mxnet/issues/15624#issuecomment-513868734 Yes, it is correct. MXNet executes the computational graph asynchronously. When asnumpy is called, it will wait until the related computational graph finishes, including the memory allocation. When using GPU context, MXNet need to allocate GPU memory in the first time. After that, MXNet can get the memory allocation from the memory pool, so it takes less time. It takes more time to allocate GPU memory than allocating CPU memory.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to 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
