File "/home/anuj/GeneFacePlusPlus/./modules/radnerfs/encoders/gridencoder/grid.py", line 62, in forward _backend.grid_encode_forward(inputs, embeddings, offsets, outputs, B, D, C, L, S, H, dy_dx, gridtype, align_corners, interpolation) RuntimeError: inputs must be a CUDA tensor

[anujsinha72094]

Getting this error. Any reason why?

[leegang79]

I encountered the same error, found that CUDA version to high, try down-grade to CUDA11.7, now can infer successfully.

1. export CUDA_HOME=/usr/local/cuda-11.7
2. down-grade pytorch to compatiable with CUDA11.7 --- CUDA 11.7-----= pip install torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2
3. re-install pytorch3d to make sure it aslo compatible with CUDA11.7