[GitHub] dmidge8 opened a new issue #13125: Unsupported GPU: wrong results

[GitBox <gi...@apache.org>]

dmidge8 opened a new issue #13125: Unsupported GPU: wrong results
URL: https://github.com/apache/incubator-mxnet/issues/13125
 
 
   I tried to run the regular Mxnet example, to train on an mnist example.
   It did run. But it was training only when running on CPU. On the Nvidia GTX 745, which is unsupported, it was running. But the results where garbage, as if it was unallocated. But switching to a Nvidia GTX 1060, it works again.
   I guess that if an exception was thrown, to warn that the GPU doesn't work, would be a better way to signal that it is not working because of the GPU, wouldn't it be? Because I had to spend some time investing that problem and would probably not be the only one.
   
   ## Environment info (Required)
   
   - Mxnet version 1.2 (C++ API)
   - Windows 7
   - Cuda 9.2
   
   ## Steps to reproduce
   
   1. Run the mnist example on CPU.
   2. Run the same example on an unsupported GPU such as the GTX 745 or the GT 730.
   
   ## What have you tried to solve it?
   
   1. I changed the GPU for an GTX1060. It works.
   

----------------------------------------------------------------
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