tslgithub
tslgithub
之前只是给了一个很简单的版本的DenseNet，容易出bug，我没有时间去做调试，所以干脆diss啦
这个代码会预先把所有的数据加载到内存，所以数据太大了，会出现内存不足的现象（不是2080Ti的现存），如果你一定要用Densenet，你可以自己改一下代码，对于Densenet的代码，我只是注释了，你把注释打开，自己调一下
具体那种模型效果好，也要看情况，我这边训练，有的数据集用resnet效果好点，有的数据集用inception效果好点，有的用VGG好点
把mnist数据转换成图片，然后按照给的例子改
there is no pretrained model. and you should train your own model
when i run the command: ns-render camera-path --load-config outputs/poster/nerfacto/2024-04-19_201438/config.yml --camera-path-filename /home/accer2/project/people/nerfstudio/data/nerfstudio/poster/camera_paths/2024-04-19-20-14-45.json --output-path renders/poster/2024-04-19-20-14-45.mp4 and it runs: Loading latest checkpoint from load_dir ✅ Done loading checkpoint from outputs/poster/nerfacto/2024-04-19_201438/nerfstudio_models/step-000059999.ckpt camera_path -> {'default_fov':...
> when i run the command: ns-render camera-path --load-config outputs/poster/nerfacto/2024-04-19_201438/config.yml --camera-path-filename /home/accer2/project/people/nerfstudio/data/nerfstudio/poster/camera_paths/2024-04-19-20-14-45.json --output-path renders/poster/2024-04-19-20-14-45.mp4 > > and it runs: Loading latest checkpoint from load_dir ✅ Done loading checkpoint from outputs/poster/nerfacto/2024-04-19_201438/nerfstudio_models/step-000059999.ckpt...
> torch.tensor(camera["camera_to_world"]).view(4, 4)[:3] I found is my "camera_path": []", camera_path is empty, but what "camera_path" means ???
The detected CUDA version (11.6) mismatches the version that was used to compile PyTorch (12.1). Please make sure to use the same CUDA versions.