MarkerWorker error

[DatBlindArcher]

Hello, I'm running OpenRAG on DockerDesktop for Windows 11.
Everything seems to be going fine untill the moment I want to upload a file.
It wants to start a MarkerWorker but it fails to launch.

(MarkerPool pid=638) INFO     | components.indexer.loaders.pdf_loaders.marker:process_pdf:184 - MarkerWorker allocated
(MarkerWorker pid=636) INFO     | components.indexer.loaders.pdf_loaders.marker:setup_mp:81 - Initializing MarkerWorker with 2 workers
(MarkerWorker pid=636) Traceback (most recent call last):
(MarkerWorker pid=636)   File "<string>", line 1, in <module>
(MarkerWorker pid=636)   File "/root/.local/share/uv/python/cpython-3.12.7-linux-x86_64-gnu/lib/python3.12/multiprocessing/spawn.py", line 122, in spawn_main
(MarkerWorker pid=636)     exitcode = _main(fd, parent_sentinel)
(MarkerWorker pid=636)                ^^^^^^^^^^^^^^^^^^^^^^^^^^
(MarkerWorker pid=636)   File "/root/.local/share/uv/python/cpython-3.12.7-linux-x86_64-gnu/lib/python3.12/multiprocessing/spawn.py", line 132, in _main
(MarkerWorker pid=636)     self = reduction.pickle.load(from_parent)
(MarkerWorker pid=636)            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(MarkerWorker pid=636)   File "/app/.venv/lib/python3.12/site-packages/torch/multiprocessing/reductions.py", line 181, in rebuild_cuda_tensor
(MarkerWorker pid=636)     storage = storage_cls._new_shared_cuda(
(MarkerWorker pid=636)               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(MarkerWorker pid=636)   File "/app/.venv/lib/python3.12/site-packages/torch/storage.py", line 1452, in _new_shared_cuda
(MarkerWorker pid=636)     return torch.UntypedStorage._new_shared_cuda(*args, **kwargs)
(MarkerWorker pid=636)            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
(MarkerWorker pid=636) RuntimeError: CUDA error: invalid resource handle
(MarkerWorker pid=636) CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
(MarkerWorker pid=636) For debugging consider passing CUDA_LAUNCH_BLOCKING=1
(MarkerWorker pid=636) Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.
(MarkerWorker pid=636)

Any idea what could be causing this?

[EnjoyBacon7]

I would initially blame trying to run OpenRAG on a windows machine, but for the sake of investigation, would you mind sharing your config?

Please note no deployment has ever been attempted on my side of things, I'm not familiar with Windows specific errors.