Getting errors while exporting

[jitesh3023]

Too few inliers for pose initialization
0%| | 1709/100000000 [01:37<1152:55:35, 24.09it/s, Focal:752.5,Keyframes:220,Gaussians:158264,AnchorToo few inliers for pose initialization
Too few inliers for pose initialization
0%| | 1989/100000000 [01:49<1072:35:33, 25.90it/s, Focal:752.5,Keyframes:239,Gaussians:195596,AnchorException in thread Thread-241 (optimization_loop):
Traceback (most recent call last):
File "/home/forrealnew/miniconda3/envs/onthefly_nvs/lib/python3.12/threading.py", line 1075, in _bootstrap_inner
self.run()
File "/home/forrealnew/miniconda3/envs/onthefly_nvs/lib/python3.12/threading.py", line 1012, in run
self._target(*self._args, **self._kwargs)
File "/home/forrealnew/repos/gpu-server-dev/gpuserver/on-the-fly-nvs/scene/scene_model.py", line 337, in optimization_loop
self.optimization_step()
File "/home/forrealnew/repos/gpu-server-dev/gpuserver/on-the-fly-nvs/scene/scene_model.py", line 302, in optimization_step
l1_loss = (image - gt_image).abs().mean()
~~~~~~^~~~~~~~~~
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!
0%| | 2493/100000000 [02:13<1490:06:34, 18.64it/s, Focal:752.5,Keyframes:292,Gaussians:132881,Anchor
Saving the reconstruction to: results/000000
num anchors: 2, num keyframes: 292, time: 133.738, FPS: 747731.355

[ameuleman]

Hi,
This seems to be a problem with how we handle training in the live stream mode. I will look into it.
However, results should still be stored in results/000000