Error loading module. Only CUDA backend is supported

[sibernewf]

I have been setting this up in Ubuntu 22 in a Virtualbox Vm.

I have followed the instructions as indicated, except for the part to get the weights. Instead I curled them as below:
curl https://www.googleapis.com/storage/v1/b/aai-blog-files/o/sd-v1-4.ckpt?alt=media > sd-v1-4.ckpt

when i try: python3 scripts/dream.py (or with the -l extension), I receive the following output:

Error loading model. Only the CUDA backend is supported
Traceback (most recent call last):
File "/home/alan/stable-diffusion/ldm/simplet2i.py", line 536, in load_model
self.device = self._get_device()
File "/home/alan/stable-diffusion/ldm/simplet2i.py", line 527, in _get_device
device_type = choose_torch_device()
File "/home/alan/stable-diffusion/ldm/dream/devices.py", line 7, in choose_torch_device
if torch.backends.mps.is_available():
AttributeError: module 'torch.backends' has no attribute 'mps'

Any Ideas?

[sibernewf]

Forgot to mention after seeing the CUDA error, I installed:

sudo apt install nvidia-cuda-toolkit

[magnusviri]

This:

elif(torch.backends.mps.is_available()):

needs to be changed to this:

elif hasattr(torch.backends, 'mps') and torch.backends.mps.is_available():

If I'm faster than the others I'll actually get this pushed before someone else.

[magnusviri]

I guess everyone is asleep!