Can't run on MacBook

[script-money]

RuntimeError: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location=torch.device('cpu') to map your storages to the CPU.

Is it possible to run on CPU?

[swankwc]

Getting the same thing

[d3vilh]

I managed to run it with following modifications:

1. add map_location='cpu' to all the torch.load functions in:
   1.1 pixelization/models/pixelization_model.py
   1.2 pixelization/models/base_model.py
   1.3 pixelization/models/c2pGen.py
   1.4 scripts/postprocessing_pixelization.py - 2 lines

   Like this:

   pixelization/models/pixelization_model.py:        state_dict = torch.load(load_path, map_location='cpu')
   pixelization/models/base_model.py:                state_dict = torch.load(load_path, map_location='cpu')
   pixelization/models/c2pGen.py:        vgg19.load_state_dict(torch.load('./pixelart_vgg19.pth', map_location='cpu'))
   scripts/postprocessing_pixelization.py:            G_A_state = torch.load(path_160_net_G_A, map_location='cpu')
   scripts/postprocessing_pixelization.py:            alias_state = torch.load(path_alias_net, map_location='cpu')

2. Patch pixelization/models/networks.py:
   2.1 Comment CUDA availability check
   2.2 Force DataParallel function to use device 'cpu' on line 113

Like this:

stable-diffusion-webui-pixelization % grep -B3 cpu pixelization/models/networks.py
    if len(gpu_ids) > 0:
      #  assert(torch.cuda.is_available())
      #  net.to(gpu_ids[0])
        net = torch.nn.DataParallel(net, 'cpu')  # multi-GPUs
stable-diffusion-webui-pixelization %

3. Restart webui.sh and Pixelize.

Enjoy.

[jpfaraco]

I modified the files and it now generates images, just not as expected:

[chariumaboat]

I am also having the same problem.

[Demacsspider]

I managed to run it with following modifications:

1. add map_location='cpu' to all the torch.load functions in:
   1.1 pixelization/models/pixelization_model.py
   1.2 pixelization/models/base_model.py
   1.3 pixelization/models/c2pGen.py
   1.4 scripts/postprocessing_pixelization.py - 2 lines
   Like this:

   pixelization/models/pixelization_model.py:        state_dict = torch.load(load_path, map_location='cpu')
   pixelization/models/base_model.py:                state_dict = torch.load(load_path, map_location='cpu')
   pixelization/models/c2pGen.py:        vgg19.load_state_dict(torch.load('./pixelart_vgg19.pth', map_location='cpu'))
   scripts/postprocessing_pixelization.py:            G_A_state = torch.load(path_160_net_G_A, map_location='cpu')
   scripts/postprocessing_pixelization.py:            alias_state = torch.load(path_alias_net, map_location='cpu')

2. Patch pixelization/models/networks.py:
   2.1 Comment CUDA availability check
   2.2 Force DataParallel function to use device 'cpu' on line 113

Like this:

stable-diffusion-webui-pixelization % grep -B3 cpu pixelization/models/networks.py
    if len(gpu_ids) > 0:
      #  assert(torch.cuda.is_available())
      #  net.to(gpu_ids[0])
        net = torch.nn.DataParallel(net, 'cpu')  # multi-GPUs
stable-diffusion-webui-pixelization %

3. Restart webui.sh and Pixelize.

Enjoy.

I tried altering the files exactly as you described, after a complete restart of Automatic1111 and terminal, the pixelation option disappear's from the webui (in the extras tab). Any idea what's happening or what I can do to get it to run (on an M1 Mac)?

[DarshanAdroja]

I managed to run it with following modifications:

1. add map_location='cpu' to all the torch.load functions in:
   1.1 pixelization/models/pixelization_model.py
   1.2 pixelization/models/base_model.py
   1.3 pixelization/models/c2pGen.py
   1.4 scripts/postprocessing_pixelization.py - 2 lines
   Like this:

   pixelization/models/pixelization_model.py:        state_dict = torch.load(load_path, map_location='cpu')
   pixelization/models/base_model.py:                state_dict = torch.load(load_path, map_location='cpu')
   pixelization/models/c2pGen.py:        vgg19.load_state_dict(torch.load('./pixelart_vgg19.pth', map_location='cpu'))
   scripts/postprocessing_pixelization.py:            G_A_state = torch.load(path_160_net_G_A, map_location='cpu')
   scripts/postprocessing_pixelization.py:            alias_state = torch.load(path_alias_net, map_location='cpu')

2. Patch pixelization/models/networks.py:
   2.1 Comment CUDA availability check
   2.2 Force DataParallel function to use device 'cpu' on line 113

Like this:

stable-diffusion-webui-pixelization % grep -B3 cpu pixelization/models/networks.py
    if len(gpu_ids) > 0:
      #  assert(torch.cuda.is_available())
      #  net.to(gpu_ids[0])
        net = torch.nn.DataParallel(net, 'cpu')  # multi-GPUs
stable-diffusion-webui-pixelization %

3. Restart webui.sh and Pixelize.

Enjoy.

This worked perfectly for me, Thanks!

[zamzamz3]

I could not get it to work this way
Did the code change?

[leouofa]

Worked for me 👍