microsoft/Recursive-Cascaded-Networks

GPU memery needed for training

haoyayuzzz opened this issue · 2 comments

Thanks for sharing your code! I want to know the GPU memery needed during training. Because I always suffered the error about resource-exhaused. If I set the batchsize = 1 during traning in liver case, is a 12GB GPU enough to train VTN-10 model?

I'm not sure as we were using our custom warping operation as explained here, and a 12GB GPU is enough in that case.

Thanks for your reply. To my understanding, custom warping operation is not available for Ubuntu version, so I might try something else.