Not utilizing the GPU at all, but hammering away on CPU regardless of settings

[Rudolph2109]

My RTX 4070 TI is being utilized by 0% while my 12700K is working very hard whenever I generate anything from the program. This is regardless of I set it to use the GPU or the CPU, and I even forced the utilization of the GPU through windows graphics settings, by adding my browser to the list, just to make sure it wasn't a Windows issue.

[leonkhoo123]

I encoutered same issue, its due the pytorch version.

But first make sure u installed nvidia cuda toolkit , can be check by using

nvcc --version

if prompt command not found, u need to install from here : https://developer.nvidia.com/cuda-downloads after installing, try to run the testing code below :

import torch
torch.zeros(1).cuda()

if it show something like : tensor([0.], device='cuda:0') means your gpu cuda should be working now.

if prompt error, depending on the error u might need to find your own solution.

Mine is due to the pytorch installed do not compiled with cuda. So, i uninstall the pytorch first with pip uninstall torch torchvision then depending on your cuda version, check using nvidia-smi download the respective pytorch from here: https://pytorch.org/ after installing, try to run the cuda testing code again. If it dint show error, u should be able to use GPU now.

[TijuanaKez]

I'm getting this too. Console says device:cuda, but nvtop shows it's not being used, but CPU %100, then it usually crashes.

Loading VAE weights specified in settings: /mnt/AI/stable-diffusion-webui/models/VAE/sdxl-fp16.vae.safetensors
Applying attention optimization: Doggettx... done.
Model loaded in 3.7s (load weights from disk: 0.9s, create model: 0.4s, apply weights to model: 2.1s).
DepthMap v0.4.7 (ee54b550)
device: cuda
Loading model(s) ..
Loading model weights from  ./models/leres/res101.pth
initialize network with normal
loading the model from ./models/pix2pix/latest_net_G.pth
Computing output(s) ..
  0%|                                                              | 0/1 [00:00<?, ?it/s]wholeImage being processed in : 1568
Adjust factor is: 1.024247491638796
Selecting patches ...
Target resolution:  (3212, 2198, 3)
Resulting depthmap resolution will be : (4864, 3328)
patches to process: 19
     processing patch 0 / 18 | [  73   73 3247 3247]
     processing patch 1 / 18 | [  73  581 3247 3247]
     processing patch 2 / 18 | [ 145 1163 3101 3101]
     processing patch 3 / 18 | [ 485 2011 2423 2423]
     processing patch 4 / 18 | [   0    0 2374 2374]
     processing patch 5 / 18 | [   0  509 2374 2374]
     processing patch 6 / 18 | [   0 1018 2374 2374]
     processing patch 7 / 18 | [   0 1526 2374 2374]
     processing patch 8 / 18 | [ 509    0 2374 2374]
     processing patch 9 / 18 | [1090   73 2229 2229]
     processing patch 10 / 18 | [1090  581 2229 2229]
     processing patch 11 / 18 | [1090 1090 2229 2229]
     processing patch 12 / 18 | [1090 1599 2229 2229]
     processing patch 13 / 18 | [ 170 2205 2035 2035]
     processing patch 14 / 18 | [   0  509 1357 1357]
     processing patch 15 / 18 | [   0 2544 1357 1357]
     processing patch 16 / 18 | [ 509    0 1357 1357]
     processing patch 17 / 18 | [1018    0 1357 1357]
     processing patch 18 / 18 | [1526    0 1357 1357]
100%|██████████████████████████████████████████████████████| 1/1 [00:16<00:00, 16.69s/it]
Computing output(s) done.
Running 3D Photo Inpainting .. 
Loading edge model ..
Loading depth model ..
Loading rgb model ..
  0%|                                                                                                                                          | 0/1 [00:00<?, ?it/s]
Generating inpainted mesh .. (go make some coffee) ..
                                                                                                                                                           ./webui.sh: line 297:  7383 Killed                  "${python_cmd}" -u "${LAUNCH_SCRIPT}" "$@"                                         | 0/7 [00:00<?, ?it/s]

PyTorch 2.3.0. Torch is working fine on my 3090 for generating images inside A1111.

Version: 2.3.0
Summary: Tensors and Dynamic neural networks in Python with strong GPU acceleration
Home-page: https://pytorch.org/
Author: PyTorch Team
Author-email: packages@pytorch.org
License: BSD-3