Geek Answers Handbook

Different output of GoogLeNet model in virtual machine and Ubuntu 14.04

I am using the GoogleNet model for binary image classification. I used to use a virtual machine, and now I use Ubuntu 14.04. Both give me different results. I tried to find out where the problem was, but I could not pinpoint it.

I trained two models separately in Ubuntu 14.04 and the other in a virtual machine. Both models use a processor. cuDNN is not used in both cases. As for the BLAS library, I use ATLAS by default.

Any suggestions would be of great help.

+4
source share
1 answer

, random_seed solver.prototxt, , caffe . , - .
, , random_seed solver.prototxt.

+4

More articles:


All Articles