"--no-cuda" does not work

[KIVix]

When using the --no-cuda argument, it returns an error.

(env) λ python pix2tex.py --no-cuda
Traceback (most recent call last):
  File "H:\pytlat\ocr\pix2tex.py", line 84, in <module>
    args, model, tokenizer = initialize(args)
  File "H:\pytlat\ocr\pix2tex.py", line 33, in initialize
    model.load_state_dict(torch.load(args.checkpoint))
  File "H:\pytlat\env\lib\site-packages\torch\serialization.py", line 594, in load
    return _load(opened_zipfile, map_location, pickle_module, **pickle_load_args)
  File "H:\pytlat\env\lib\site-packages\torch\serialization.py", line 853, in _load
    result = unpickler.load()
  File "H:\pytlat\env\lib\site-packages\torch\serialization.py", line 845, in persistent_load
    load_tensor(data_type, size, key, _maybe_decode_ascii(location))
  File "H:\pytlat\env\lib\site-packages\torch\serialization.py", line 834, in load_tensor
    loaded_storages[key] = restore_location(storage, location)
  File "H:\pytlat\env\lib\site-packages\torch\serialization.py", line 175, in default_restore_location
    result = fn(storage, location)
  File "H:\pytlat\env\lib\site-packages\torch\serialization.py", line 151, in _cuda_deserialize
    device = validate_cuda_device(location)
  File "H:\pytlat\env\lib\site-packages\torch\serialization.py", line 135, in validate_cuda_device
    raise RuntimeError('Attempting to deserialize object on a CUDA '
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.

I use torch 1.7.+cpu, cuda version is not installed, and can't use cuda.

[lukas-blecher]

Thank you!
I missed an argument. It should work now.
Let me know if there are any other problems.

[lukas-blecher]

Feel free to reopen if the issue is not resolved.

[aleksandar-vuckovic]

I am sorry if this is the wrong place to ask this, but I have a similar issue. When I start latexocr (GUI) or the command line program, I get the following error message:

UserWarning: Failed to load image Python extension: libc10_cuda.so: cannot open shared object file: No such file or directory
regardless if I start them with --no-cuda or not.

I should add that I installed the package via pip, did no additional training, and am running Arch with Python v3.10.5 .

[lukas-blecher]

Looks more like a broken installation to me. Can you confirm the pytorch installation is valid?
This particular issue is resolved, I believe.

[aleksandar-vuckovic]

I think so, at least running to commands provided in https://pytorch.org/get-started/locally/ worked fine. For the GUI application, it starts while throwing this error message, but hangs when trying to snip a formula.

[lukas-blecher]

How about

import torch
print(torch.randn(2, device='cuda'))

and does the cli work?

[aleksandar-vuckovic]

First of all, I want to say that I appreciate you answering this fast and helping me.
The python commands print

>>> import torch
>>> print(torch.randn(2, device='cuda'))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python3.10/site-packages/torch/cuda/__init__.py", line 211, in _lazy_init
    raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled

which is to be expected, as my GPU is an AMD R9 290 and not nvidia, which is why I tried the "--no-cuda" flag.
The CLI output is

 $ pix2tex -f Screenshot.png --no-cuda 
/home/username/.local/lib/python3.10/site-packages/torchvision/io/image.py:13: UserWarning: Failed to load image Python extension: libc10_cuda.so: cannot open shared object file: No such file or directory
  warn(f"Failed to load image Python extension: {e}")

Predict LaTeX code for image ("?"/"h" for help). 

It seemingly ignores that I already specified a file. The only thing which does not just print the prompt again is when I enter the absolute path of the file, where it segfaults.

[lukas-blecher]

No problem!
The thing with the file is a bug right now. I've discovered it a while back, but I'm currently not really allowed to commit to this repo.
If you don't have a nvidia GPU in your system, that is handeled automatically (--no-cuda isn't needed, but also doesn't change anything)

From the error message it looks like torchvision is the problem here.
Try to reinstall that package

[aleksandar-vuckovic]

Oh you're right, I don't know how I missed that torchvision was the problem.
The issue was that I had the Arch-package python-pytorch installed, while torch and torchvision were installed via pip. It seems that it tried to use the system-provided torch and the pip-provided torchvision.
Uninstalling the python-pytorch package and reinstalling torch and torchvision via pip fixed it, and everything seems to work now. Thanks.