Songyang Sun

@Tianjin3 it works! when you fail about something error of nvcc, you can do environment variable of cuda again, then add CUDA_HOME=/usr/local/cuda. command: vim ~/.bashrc add at last(this include environment...

i try it

if you use coco format, please check class. then you should confirm the path you train,i make a mistake on training path.

https://github.com/WXinlong/SOLO/blob/master/docs/GETTING_STARTED.md Useful tools -> Analyze logs

> if you use configs/solo/solo_r50_fpn_8gpu_1x.py, you change "workflow = [('train',1),('val',1)]" in configs/solo/solo_r50_fpn_8gpu_1x.py. I try it,but it can't do eval in SOLO. Can you do eval and print mAP result?

> > if you use configs/solo/solo_r50_fpn_8gpu_1x.py, you change "workflow = [('train',1),('val',1)]" in configs/solo/solo_r50_fpn_8gpu_1x.py. > > I try it,but it can't do eval in SOLO. Can you do eval and print...