shesung commented on issue #11076: [MXNET-491] Use depthwise convolution by cuDNNv7 if available, updated version URL: https://github.com/apache/incubator-mxnet/pull/11076#issuecomment-416570195 I observed a dramatic performance drop when using mxnet 1.2.1 + cuda8 + cudnn 7.2.1 on 1080ti input_size = (448, 800) mobilenet_v1 : 137 fps -> 41 fps mobilenet_v2 : 180 fps -> 57 fps And it used a lot of gpu memory Any advices?
---------------------------------------------------------------- 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
