gepu0221
Results
2
comments of
gepu0221
Hi @zlbdzhh, Have you solved the problem? There are custom operators implemented by CUDA in pointnet2. It's seemed this document doesn't support CUDA implementation. https://github.com/onnx/tutorials/blob/master/PyTorchCustomOperator/README.md
> > > 我用全连接层匹配的方式训练了一把，在私人数据集上的结果是82%，arcface是84%。 > > > > > > 请问你的setting是啥呀，BatchSize 是多少呢？ > > m,s和论文一致，bs200 可以问下具体的实验设置吗，我在id=1000的数据集上，loss没有收敛