Time cosumming and GPU memory #10
Description
After I transformed the mxnet model into caffe model using the provided codes, I found the time comsumming and GPU memory of caffe model was much larger than the mxnet model. Is that normal?
{{ message }}