i can't get cuda working

[eliecer696]

am using conda latest version i have cudatoolkit installed on my environment and when run the text 2point cloud it chooses my CPU instead of my GPU any solutions ?

[Natotela]

Same issue here. GPU util 0%, CPU up to 77%

[Jgensel]

I had to manually install pytorch to get it to run on the GPU. Run the command from here: https://pytorch.org/get-started/locally/ Make sure you have the correct version of Cuda installed (11.7) as well

[n00mkrad]

Uninstall torch and torchvision, then try:

install torch torchvision --extra-index-url https://download.pytorch.org/whl/cu117

[Natotela]

not sure if I wanna screw anything up, at the moment I run it with certain parameters, which my 8gb seems to handle. Problem is I can't seem to allow a sufficient standard for proper embedding training. Regarding what's in that Link of @Arkay92 mention, what's that about "torch.cuda.device_of() function to determine if the input tensors are on the GPU or CPU" - is it solely for the Colab or I can use it somehow running webUI? to see utilization of GPU

[Arkay92]

not sure if I wanna screw anything up, at the moment I run it with certain parameters, which my 8gb seems to handle. Problem is I can't seem to allow a sufficient standard for proper embedding training.

Regarding what's in that Link of @Arkay92 mention, what's that about "torch.cuda.device_of() function to determine if the input tensors are on the GPU or CPU" - is it solely for the Colab or I can use it somehow running webUI? to see utilization of GPU

Hi, seems I may have overthought the gpu detection and need to rework in a new PR but with the current one it should delete old gradients in the forward pass leading to much faster compilation on point diffusion stuff check it out and let me know if you have any issues 😊