suetme

Hi，I also encountered this problem. Have you solved this problem？

My usage environment is: GCC 7.5.0, ubuntu18.04, cuda10.02, pythoch1.6, torchvision0.7.0. cudatoolkit10.2,torch-geometric2.1.0.post1

Hi, thank you for this excellent work.！When I trained the Partseg model using my own dataset, I encountered a situation where the acc was very low and the iou was...

Over the next iterations, the loss remains around 2.37, the acc remains around 0.38, the avgacc remains around 0.17, and the iou fluctuates between 0.85 and 0.93. By the way,...