TiagoCortinhal / SalsaNext

Uncertainty-aware Semantic Segmentation of LiDAR Point Clouds for Autonomous Driving
MIT License
413 stars 102 forks source link

About the problems that arise during training #75

Open SSSSSSStar opened 1 year ago

SSSSSSStar commented 1 year ago

Hi, I run train.py, but every time after the validation of the first epoch, there is an error out of memory when I proceed to the second epoch. With 3080/3090, the memory suddenly increases at the beginning of the second epoch Is there any solution to this problem