Xuhan Sheng

I create the project environment on Anaconda with python=3.7, and while following the requirements.txt to install packages there occurred some problem while installing torch-scatter, torch-sparse, torch-cluster, torch-spline-conv. Finally, I use...

关于训练终止的问题

9

我设置total iter为12w，训练在9w的时候输出"End of training"，中间没有出现任何报错。 起初我以为是代码内部有EarlyStop机制，但是现在看来好像不是。 以下是我两次训练的输出文档。total iter都设置为12w，在9w终止训练，性能最好的模型分别对应迭代次数为6w和9w。 这有些奇怪。  

It's a fancy work. We hope the COMPLETE instruction will be ready soon.

About the pretrained weights loading problem in training

7

While loading **UNet2DConditionModel** from pretrained SD-2, **ValueError** appears, as parameter weights related to image_attentions are missed. > Traceback (most recent call last): > File "train_seesr.py", line 680, in > unet...

Experimental reproduction of "train" scene has different brightness compared with original 3dGS

2

 The test view rendered by gsplat is much darker. The following is the rendered results from original 3dGS: