RuntimeError: CUDA error: device-side assert triggered

[wangalong-ahpu]

loading images from path: ./input
loading image from path: ./input/s1.png
loading image from path: ./input/s2.png
loading image from path: ./input/s3.png
loading image from path: ./input/s4.png
loading image from path: ./input/s5.png
loading image from path: ./input/s6.png
loading image from path: ./input/s7.png
detecting landmarks using: fan
/root/autodl-tmp/NextFace/landmarksfan.py:35: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /opt/conda/conda-bld/pytorch_1639180544224/work/torch/csrc/utils/tensor_new.cpp:201.)
return torch.tensor(landmarks, device = self.device)
init camera pose...
1/3 => Optimizing head pose and expressions using landmarks...
100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 2000/2000 [00:16<00:00, 123.93it/s]
2/3 => Optimizing shape, statistical albedos, expression, head pose and scene light...
0%| | 0/401 [00:00<?, ?it/s]/opt/conda/conda-bld/pytorch_1639180544224/work/aten/src/ATen/native/cuda/IndexKernel.cu:93: operator(): block: [1457,0,0], thread: [0,0,0] Assertion index >= -sizes[i] && index < sizes[i] && "index out of bounds" failed.
| 0/401 [00:00<?, ?it/s]
Traceback (most recent call last):
File "optimizer.py", line 490, in
optimizer.run(inputDir,
File "optimizer.py", line 438, in run
self.runStep2()
File "optimizer.py", line 241, in runStep2
diffuseTextures = self.pipeline.morphableModel.generateTextureFromAlbedo(diffAlbedo)
File "/root/autodl-tmp/NextFace/morphablemodel.py", line 173, in generateTextureFromAlbedo
texture = (baryCenterWeights[:, neighboors, None] * albedo[:, self.faces[oFaces[:, None], neighboors]]).sum(dim=-2)
RuntimeError: CUDA error: device-side assert triggered