[BUG] Attempting to compile/download foundation models for use on CPU.

[CompRhys]

Describe the bug
Following https://www.nequip.net/models/mir-group/NequIP-OAM-L:0.1 but attempting to compile to run on CPU leads to following errors.

To Reproduce

(torch-sim-atomistic) ➜  torch-sim git:(nequip-framework) ✗ nequip-compile \
  nequip.net:mir-group/NequIP-OAM-L:0.1 \
  mir-group__NequIP-OAM-L__0.1.nequip.pt2 \
  --mode aotinductor \
  --device cpu \
  --target ase
[2025-10-03 17:30:50,601][nequip.scripts.compile][INFO] - [rank: 0] Compiling for device: cpu
[2025-10-03 17:30:50,602][nequip.model.saved_models.load_utils][INFO] - [rank: 0] Fetching mir-group/NequIP-OAM-L:0.1 from nequip.net...
Downloading from Zenodo: 100%|██████████████████████████████████████████████████████| 78.5M/78.5M [00:27<00:00, 2.90MB/s]
[2025-10-03 17:31:18,461][nequip.model.saved_models.load_utils][INFO] - [rank: 0] Download complete, loading model...
[2025-10-03 17:31:18,461][nequip.model.saved_models.package][INFO] - [rank: 0] Loading model from package file: /var/folders/nl/sh8rg5qn11b4f5g6lgn35m380000gn/T/tmpf3a9vlz8.nequip.zip ...
Traceback (most recent call last):
  File "/Users/radical-rhys/Radical/torch-sim/.venv/bin/nequip-compile", line 10, in <module>
    sys.exit(main())
             ^^^^^^
  File "/Users/radical-rhys/Radical/torch-sim/.venv/lib/python3.12/site-packages/nequip/scripts/compile.py", line 186, in main
    model = load_saved_model(
            ^^^^^^^^^^^^^^^^^
  File "/Users/radical-rhys/Radical/torch-sim/.venv/lib/python3.12/site-packages/nequip/model/saved_models/load_utils.py", line 120, in load_saved_model
    model = ModelFromPackage(actual_input_path, compile_mode=compile_mode)
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/radical-rhys/Radical/torch-sim/.venv/lib/python3.12/site-packages/nequip/model/saved_models/package.py", line 132, in ModelFromPackage
    model = imp.load_pickle(
            ^^^^^^^^^^^^^^^^
  File "/Users/radical-rhys/Radical/torch-sim/.venv/lib/python3.12/site-packages/torch/package/package_importer.py", line 303, in load_pickle
    result = unpickler.load()
             ^^^^^^^^^^^^^^^^
  File "/opt/homebrew/Caskroom/miniconda/base/lib/python3.12/pickle.py", line 1256, in load
    dispatch[key[0]](self)
  File "/opt/homebrew/Caskroom/miniconda/base/lib/python3.12/pickle.py", line 1760, in load_build
    setstate(state)
  File "<torch_package_0>.e3nn/util/codegen/_mixin.py", line 121, in __setstate__
  File "/Users/radical-rhys/Radical/torch-sim/.venv/lib/python3.12/site-packages/torch/storage.py", line 530, in _load_from_bytes
    return torch.load(io.BytesIO(b), weights_only=False)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/radical-rhys/Radical/torch-sim/.venv/lib/python3.12/site-packages/torch/serialization.py", line 1549, in load
    return _legacy_load(
           ^^^^^^^^^^^^^
  File "/Users/radical-rhys/Radical/torch-sim/.venv/lib/python3.12/site-packages/torch/serialization.py", line 1807, in _legacy_load
    result = unpickler.load()
             ^^^^^^^^^^^^^^^^
  File "/Users/radical-rhys/Radical/torch-sim/.venv/lib/python3.12/site-packages/torch/serialization.py", line 1742, in persistent_load
    obj = restore_location(obj, location)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/radical-rhys/Radical/torch-sim/.venv/lib/python3.12/site-packages/torch/serialization.py", line 698, in default_restore_location
    result = fn(storage, location)
             ^^^^^^^^^^^^^^^^^^^^^
  File "/Users/radical-rhys/Radical/torch-sim/.venv/lib/python3.12/site-packages/torch/serialization.py", line 636, in _deserialize
    device = _validate_device(location, backend_name)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/radical-rhys/Radical/torch-sim/.venv/lib/python3.12/site-packages/torch/serialization.py", line 605, in _validate_device
    raise RuntimeError(
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.

Expected behavior
The tensors should be saved from CPU such that they can be loaded to a CPU device. I expect this is the problem here.

Environment (please complete the following information):

• OS: MacOS
• python version: 3.12
• python environment (commands are given for python interpreter):
  • nequip version 0.15
  • e3nn version 0.5.7
  • pytorch version 2.7.1
• (if relevant) GPU support with CUDA
  • cuda Version according to nvcc (nvcc --version)
  • cuda version according to PyTorch (import torch; torch.version.cuda)

Additional context
Add any other context about the problem here.