Forward modelling causes NumPy FileNotFoundError

[AlexanderJuestel]

This is more like a discussion than a bug report.

Since the old modification methods such as modify_orientations for probabilistic modeling do not work in GemPy 3 anymore, I have just set up a for loop that creates completely new models with new input data sets (the easiest way for me now instead of digging too deep into the new GemPy API). However, after about 400 iterations, I am getting the following error, which I could reproduce on another machine in a fresh environment.

It is attributed to some NumPy Error. I have no idea if NumPy is saving some temporary files or so...

Happy to discuss some solutions here.

---> 35 sol = gp.compute_model(geo_model)
     37 list_vertices1.append(geo_model.transform.apply_inverse(geo_model.solutions.raw_arrays.vertices[0]))
     38 list_vertices2.append(geo_model.transform.apply_inverse(geo_model.solutions.raw_arrays.vertices[1]))

File ~\Anaconda3\envs\gempy3\Lib\site-packages\gempy\API\compute_API.py:35, in compute_model(gempy_model, engine_config)
     32 match engine_config.backend:
     33     case AvailableBackends.numpy | AvailableBackends.PYTORCH:
---> 35         BackendTensor.change_backend_gempy(
     36             engine_backend=engine_config.backend,
     37             use_gpu=engine_config.use_gpu,
     38             dtype=engine_config.dtype
     39         )
     41         # TODO: To decide what to do with this.
     42         interpolation_input = gempy_model.interpolation_input_copy

File ~\Anaconda3\envs\gempy3\Lib\site-packages\gempy_engine\core\backend_tensor.py:45, in BackendTensor.change_backend_gempy(cls, engine_backend, use_gpu, dtype)
     43 @classmethod
     44 def change_backend_gempy(cls, engine_backend: AvailableBackends, use_gpu: bool = True, dtype: Optional[str] = None):
---> 45     cls._change_backend(engine_backend, pykeops_enabled=DEFAULT_PYKEOPS, use_gpu=use_gpu, dtype=dtype)

File ~\Anaconda3\envs\gempy3\Lib\site-packages\gempy_engine\core\backend_tensor.py:65, in BackendTensor._change_backend(cls, engine_backend, pykeops_enabled, use_gpu, dtype)
     63     spec = find_spec('numpy')
     64     tfnp = module_from_spec(spec)
---> 65     spec.loader.exec_module(tfnp)
     67 cls._set_active_backend_pointers(engine_backend, tfnp)
     68 cls.tensor_types = Union[tfnp.ndarray]  # tensor Types with respect the backend

File <frozen importlib._bootstrap_external>:940, in exec_module(self, module)

File <frozen importlib._bootstrap>:241, in _call_with_frames_removed(f, *args, **kwds)

File ~\Anaconda3\envs\gempy3\Lib\site-packages\numpy\__init__.py:112
    108     if os.path.isdir(libs_dir):
    109         os.add_dll_directory(libs_dir)
--> 112 _delvewheel_patch_1_5_2()
    113 del _delvewheel_patch_1_5_2
    114 # end delvewheel patch

File ~\Anaconda3\envs\gempy3\Lib\site-packages\numpy\__init__.py:109, in _delvewheel_patch_1_5_2()
    107 libs_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), os.pardir, 'numpy.libs'))
    108 if os.path.isdir(libs_dir):
--> 109     os.add_dll_directory(libs_dir)

File <frozen os>:1118, in add_dll_directory(path)

FileNotFoundError: [WinError 206] Der Dateiname oder die Erweiterung ist zu lang: 'C:\\Users\\ale93371\\Anaconda3\\envs\\gempy3\\Lib\\site-packages\\numpy.libs'

[javoha]

Hi,
no idea where this is coming from but have you tried allwoing long paths on windows, like suggested here:
https://stackoverflow.com/questions/72352528/how-to-fix-winerror-206-the-filename-or-extension-is-too-long-error

[AlexanderJuestel]

Yes, was already enabled