Open
Description
Thank you for great open-sourced work!
When I train my custom dataset with COLMAP type using genColmapSceneInfo,
I faced backward error as below:
[Error occurred here]
Line 150 in 1493863
[Error message here]
Training progress: 0%| | 0/15000 [00:00<?, ?it/s]Traceback (most recent call last):
File "train.py", line 290, in <module>
training(lp.extract(args), op.extract(args), pp.extract(args), args.test_iterations, args.save_iterations, args.checkpoint_iterations, args.start_checkpoint, args.debug_from)
File "train.py", line 151, in training
loss.backward()
File "/home/ubuntu/miniforge3/envs/gaussian_surfels/lib/python3.8/site-packages/torch/_tensor.py", line 525, in backward
torch.autograd.backward(
File "/home/ubuntu/miniforge3/envs/gaussian_surfels/lib/python3.8/site-packages/torch/autograd/__init__.py", line 267, in backward
_engine_run_backward(
File "/home/ubuntu/miniforge3/envs/gaussian_surfels/lib/python3.8/site-packages/torch/autograd/graph.py", line 744, in _engine_run_backward
return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
RuntimeError: Function _RasterizeGaussiansBackward returned an invalid gradient at index 2 - got [0, 0, 3] but expected shape compatible with [0, 16, 3]
Referring to error [0, 16, 3], it seems to be related to the variables below:
# (0, 1, 3)
self._features_dc
# (0, 15, 3)
self._features_rest
Cuz, total shape of them is (0, 16, 3).
For custom images, I extracted normal using estimate_normal.py.
Metadata
Metadata
Assignees
Labels
No labels