May I know MXNET by default will store all variable in memory instead of GPU memory even I have install CUDA version MXnet, unless I specifically assign ctx=gpu(0)?
Is there any way to make it by default everything store in GPU memory, just like TensorFlow default behaviour?


When I’m storing variables to GPU I get the following notification:
C:\Jenkins\workspace\mxnet\mxnet\src\storage\ Using Pooled (Naive) StorageManager for GPU
Is this normal, or is it bad?