torch dependency results in error

[maxjeblick]

I'm getting the error below when running any torch code. This is probably due to an incompatible cuda version (requirements.txt specifies cu121).

I would suggest to

• ether add a comment in REAME.md about the epxected cuda version
• unpin cuda version in requirement.txt (I'm curious why dev20240723 is selected? At least the first .generate example works fine with 2.4)

(venv)  max@max  ~/PycharmProjects/cold-compress   main  python3 -mtorch.utils.collect_env
Traceback (most recent call last):
  File "<frozen runpy>", line 189, in _run_module_as_main
  File "<frozen runpy>", line 112, in _get_module_details
  File "/home/max/PycharmProjects/cold-compress/venv/lib/python3.11/site-packages/torch/__init__.py", line 364, in <module>
    _load_global_deps()
  File "/home/max/PycharmProjects/cold-compress/venv/lib/python3.11/site-packages/torch/__init__.py", line 320, in _load_global_deps
    raise err
  File "/home/max/PycharmProjects/cold-compress/venv/lib/python3.11/site-packages/torch/__init__.py", line 298, in _load_global_deps
    ctypes.CDLL(global_deps_lib_path, mode=ctypes.RTLD_GLOBAL)
  File "/usr/lib/python3.11/ctypes/__init__.py", line 376, in __init__
    self._handle = _dlopen(self._name, mode)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^
OSError: /home/max/PycharmProjects/cold-compress/venv/lib/python3.11/site-packages/torch/lib/libtorch_global_deps.so: cannot open shared object file: No such file or directory

[griff4692]

Thanks for flagging this.

We can unpin the cuda requirement and test internally with others.

dev20240723 was chosen because a bugfix was made by the PyTorch team that allowed our code to compile without error.

For now, feel free to try by installing the latest PyTorch nightly with your specific CUDA version.

We will do some testing internally and update the README.