GPU Usage 100% at dataloading while not proceeding

[TintinWuNTUEE]

Hi, I was trying to run the model via Colab provided at the README, however I encountered two issues,

1. I was training on the bouncingball scene from dnerf and the GPU usage spiked to 100% while dataloading and the training somehow stucks there...
2. I proceeded on training on the hellwarrior dataset and getting a loss is nan while training
   (I am using a RTX 4090 with CUDA 11.6 and pytorch 1.13)
   Wondering if there is something I am doing incorrectly, thank you!

[guanjunwu]

e, most cases happened when your CUDA version not corresponds with your pytorch version.
Is your pytorch version = torch1.13+cu117?

[TintinWuNTUEE]

I am using torch1.13+cu116, does that make a difference?