train /loss_pose is not decreasing on metallic texturless custom dataset

Hello @YangHai-1218 , Thanks for the great work.

I have encountered a problem that train /loss_pose is not decreasing on metallic texturless custom dataset even after 35000 iterations. Is there any changes I need to make on the config files given. Additionally I have to mention the ply models didn't have any texture as they are metallic textureless industrial objects.

Hello @AjinJayan , It seems you have successfully run the training code. I'd like to ask a question: I have tried many versions but still cannot get mmcv-full to be compatible, and thus fail to set up an environment where the training code can run properly. Could you share the versions of your environment? I am using Python 3.9, CUDA 11.3, and PyTorch 1.12.1, which is for the purpose of installing PyTorch3D.

ImportError: /root/miniconda3/envs/py3d/lib/python3.9/site-packages/mmcv/_ext.cpython-39-x86_64-linux-gnu.so: undefined symbol: _ZN2at5sliceERKNS_6TensorElN3c108optionalIlEES5_l