+1, we also agree with memory optimization. When we reimplement some experiment, we always get stuck with out of memory error if we want to use enough batch size for the experiment. And we found that MXNET_GPU_MEM_POOL_TYPE is not helpful.
[ Full content available at: https://github.com/apache/incubator-mxnet/issues/12226 ] This message was relayed via gitbox.apache.org for [email protected]
