evil-master

我也是，请问怎么解决的?

numwork设置为0，关闭pin_memory

https://github.com/chenyuntc/simple-faster-rcnn-pytorch/issues/157#issuecomment-549062091 @PeterBishop0

> > > > @young147 > > > > Hi, I want to train with my own datasets without image and calib . But I met an error which is:...

请问,有在训练过程中有遇到这个错误吗,我强制转换成float仍然不起作用.我的torch版本是1.1 spconv是1.0 File "/home/neousys/cjg/CenterPoint-KITTI/pcdet/models/dense_heads/centerpoint_head_single.py", line 687, in gaussian_focal_loss pos_loss = (-(pred + eps).log() * (1 - pred).pow(alpha) * pos_weights).float() RuntimeError: expected backend CUDA and dtype Float but got backend CUDA...

I solve it by change int to float,maybe cause by torch 1.1. def gaussian_focal_loss(pred, gaussian_target, alpha=2.0, gamma=4.0): eps = 1e-12 pos_weights = gaussian_target.eq(1) neg_weights = (1 - gaussian_target).pow(gamma) pos_loss =...

> Thank you for your kind release of the codes, I have successfully applied it to my own data and got good performace. But may I ask how to export...