jiepengwang / NeuRIS

MIT License
227 stars 16 forks source link

What is Torch or PyTorch version required? #12

Closed ashish-kubade closed 1 year ago

ashish-kubade commented 1 year ago

Hi @jiepengwang, thanks for making this work publicly available. I tried the installation instructions. However, during the training, I am getting an error self.intrinsics_all_inv = torch.inverse(self.intrinsics_all) # n, 4, 4 RuntimeError: CUDA error: no kernel image is available for execution on the device

I am using NVIDIA GeForce RTX 3080 Ti Laptop GPU and I have cuda 11.1 installed. The Pytorch version 1.8.0+cu111.

In the instructions, in Setup section, we are installing Pytorch 1.9.0 and in the requirements.txt, we use torch=1.8.0 My question is, do we need pytorch 1.9.0 explicitly? And for torch=1.8.0, should it be tagged with certain cuda version? (e.g. torch 1.8.0 + cu111)

Apologies, in case I missed any point.

ashish-kubade commented 1 year ago

Update: I am able to run with PyTorch 1.8.0+cu111.