Installation problem with new Poetry version

[hiraksarkar]

I think the error I am getting is because of a specification change in Poetry's version

Poetry (version 2.0.1)

Error I am getting is,

(torch_topo) ➜  pytorch-topological git:(main) /n/fs/ragr-data/users/hirak/soft/poetry/bin/poetry install                        

The Poetry configuration is invalid:
  - project must contain ['name'] properties

I reported the issue python-poetry/poetry#9987 here, maybe a quick PR will fix.

[Pseudomanifold]

Sorry to hear that—can you try updating pyproject.toml with the name property and see whether that works? Seems to be what the other issue is suggesting.

[Roman223]

It seems in order to make this work one need to make this:

1. change [tool.project] to [project], since the former is unknown key for pip parser.
2. change authors to:

authors = [
    {name = "Bastian Rieck", email = "bastian@rieck.me"},
]

Because this value has to be like here.
3. Upgrade POT to ^0.9.0 becuase of numpy error:

ModuleNotFoundError: No module named 'numpy.distutils'

The last thing has to be clear to be sure that this is a fix for problem: does POT upgrade preserve logic of pytorch-topological?

---

UPD: unfortunately, it seems there are a few errors. I have a traceback after running of tests:

tests/test_alpha_complex.py:22 (TestAlphaComplexBatchHandling.test_processing)
self = <tests.test_alpha_complex.TestAlphaComplexBatchHandling object at 0x7941d0a1ef60>

    def test_processing(self):
        for (x, y) in self.loader:
>           pers_info = self.ac(x)

test_alpha_complex.py:25: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
<HIDE>/venv/lib/python3.12/site-packages/torch/nn/modules/module.py:1736: in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
<HIDE>/venv/lib/python3.12/site-packages/torch/nn/modules/module.py:1747: in _call_impl
    return forward_call(*args, **kwargs)
../torch_topological/nn/alpha_complex.py:89: in forward
    return batch_handler(x, self._forward)
../torch_topological/nn/data.py:294: in batch_handler
    return [handler_fn(torch.as_tensor(x_), **kwargs) for x_ in x]
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

self = AlphaComplex()
x = tensor([[ 0.3596,  0.4990,  0.7885],
        [-0.3597, -0.3509, -0.8645],
        [-0.3849,  0.2649, -0.8841],
       ...690, -0.4866,  0.0901],
        [ 0.5751, -0.6836, -0.4494],
        [ 0.8362,  0.4342, -0.3351]], dtype=torch.float64)

    def _forward(self, x):
>       alpha_complex = gudhi.alpha_complex.AlphaComplex(
            x.cpu().detach(),
            precision='fast',
        )
E       AttributeError: module 'gudhi' has no attribute 'alpha_complex'. Did you mean: 'AlphaComplex'?

../torch_topological/nn/alpha_complex.py:92: AttributeError

I've managed to fix this. All tests have passed. So I really hope that these tests have true coverage...

[Pseudomanifold]

Thanks! It seems like I should update the pyproject.toml? Could you open a PR for this?

[Roman223]

@Pseudomanifold , done!