[GitHub] leonid-pishchulin commented on issue #7518: V0.11.0

[gi...@git.apache.org]

leonid-pishchulin commented on issue #7518: V0.11.0
URL: https://github.com/apache/incubator-mxnet/pull/7518#issuecomment-334301600
 
 
   I have a large discrepancy in ResNet-18 run-time on Jetson TX2 for MXNet v0.9.5 vs. v0.11.0 when running with batch=1: 91 ms/fr vs. 124 ms/fr. nvvp shows that the same conv kernels are called and they take ~2x longer compared in case of v0.11.0. Do you have an idea why that might happen?
   Experimental setup:
   - Jetson TX2 with Jetpack 3.1 (cuDNN 6.0, CUDA 8.0)
   - ResNet-18 based on implementation https://github.com/apache/incubator-mxnet/blob/master/example/image-classification/symbols/resnet.py
   - input resolution 640x480
   - batch = 1
   
   nvvp output:
   v0.9.5: https://drive.google.com/open?id=0Bwb8o6lldJgBSFpmb1R5SmpYdnM
   v0.11.0: https://drive.google.com/open?id=0Bwb8o6lldJgBYzBSY1RGa1RsMnc
   
   running with batch = 8 results into similar run-time in both cases
   Adding @DickJC123 and @ptrendx as they helped to resolve another TX2 specific issue ([#5897](https://github.com/apache/incubator-mxnet/pull/5897)) I had before
 
----------------------------------------------------------------
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:
users@infra.apache.org


With regards,
Apache Git Services