Segmentation Fault when train model=2

RenYurui / StructureFlow

Code for paper "StructureFlow: Image Inpainting via Structure-aware Appearance Flow"

Other

222 stars 42 forks source link

Segmentation Fault when train model=2 #8

Closed ReSpike closed 4 years ago

ReSpike commented 5 years ago

I installed the new layer follow README and try to train the inpaint model. The structure model was training perfectly, but I got Segmentation Fault in the 2nd inpaint model. The version of my pytorch is 1.0.1. Are there anybody else meet the same problem? Would you please to help me to check it?

RenYurui commented 5 years ago

Could you please tell me your gup and cuda vision? Is there any problem when you build the Gaussian sampling layer?