Refresh CUDA 12 migrator

[conda-forge-admin]

Recently the CUDA 11.8 migrator was closed out ( conda-forge/conda-forge-pinning-feedstock#5340 ). As part of that it was determined a small updated was needed to the CUDA 12 migrator ( conda-forge/conda-forge-pinning-feedstock#5340 (comment) ). While this largely works for most feedstocks without changes, those that have use_local: true will have issues with re-rendering until the CUDA 12 migrator they have is updated. This PR pulls over those updates to fix re-rendering on the feedstock. Please let us know if you have any other questions.

---

Hi! This is the friendly automated conda-forge-webservice.

I've started rerendering the recipe as instructed in #9.

If I find any needed changes to the recipe, I'll push them to this PR shortly. Thank you for waiting!

Here's a checklist to do before merging.

• Bump the build number if needed.

Fixes #9

[conda-forge-webservices]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[github-actions]

Hi! This is the friendly automated conda-forge-webservice.

I tried to rerender for you but ran into some issues. Please check the output logs of the latest rerendering GitHub actions workflow run for errors. You can also ping conda-forge/core for further assistance or try re-rendering locally.

This message was generated by GitHub actions workflow run https://github.com/conda-forge/torchmd-net-feedstock/actions/runs/7498737229.

[jakirkham]

@conda-forge-admin , please re-render

[jakirkham]

Seeing some test failures. This is taken from the no CUDA CI test failure. Though the CUDA 11.8 test failure shows the same issues

Details

tests/test_calculator.py::test_compare_forward_multiple FAILED

=================================== FAILURES ===================================
________________________ test_compare_forward_multiple _________________________

    def test_compare_forward_multiple():
        checkpoint = join(dirname(dirname(__file__)), "tests", "example.ckpt")
        z1, pos1, _ = create_example_batch(multiple_batches=False)
        z2, pos2, _ = create_example_batch(multiple_batches=False)
        calc = External(checkpoint, torch.stack([z1, z2], dim=0))
        model = load_model(checkpoint, derivative=True)
    
        e_calc, f_calc = calc.calculate(torch.cat([pos1, pos2], dim=0), None)
        e_pred, f_pred = model(
            torch.cat([z1, z2]),
            torch.cat([pos1, pos2], dim=0),
            torch.cat([torch.zeros(len(z1)), torch.ones(len(z2))]).long(),
        )
    
        assert_allclose(e_calc, e_pred)
>       assert_allclose(f_calc, f_pred.view(-1, len(z1), 3))

tests/test_calculator.py:81: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

actual = tensor([[[ -4.7301, -13.3530, -11.4289],
         [ 61.1712, -85.8726,  93.5858],
         [ -0.1567,   0.4098,  13.75...    [ -4.1669,  -2.3176, -13.8882],
         [ -4.4799,  -0.2830,   4.1498],
         [ -3.8376,  -3.0226,   2.2429]]])
expected = tensor([[[ -4.7301, -13.3530, -11.4289],
         [ 61.1711, -85.8726,  93.5857],
         [ -0.1567,   0.4098,  13.75...-13.8882],
         [ -4.4799,  -0.2830,   4.1498],
         [ -3.8376,  -3.0226,   2.2429]]], grad_fn=<ViewBackward0>)
rtol = 0.0001, atol = 1e-05, equal_nan = True, msg = ''

    def assert_allclose(
        actual: Any,
        expected: Any,
        rtol: Optional[float] = None,
        atol: Optional[float] = None,
        equal_nan: bool = True,
        msg: str = "",
    ) -> None:
        """
        .. warning::
    
           :func:`torch.testing.assert_allclose` is deprecated since ``1.12`` and will be removed in a future release.
           Please use :func:`torch.testing.assert_close` instead. You can find detailed upgrade instructions
           `here <https://github.com/pytorch/pytorch/issues/61844>`_.
        """
        warnings.warn(
            "`torch.testing.assert_allclose()` is deprecated since 1.12 and will be removed in a future release. "
            "Please use `torch.testing.assert_close()` instead. "
            "You can find detailed upgrade instructions in https://github.com/pytorch/pytorch/issues/61844.",
            FutureWarning,
            stacklevel=2,
        )
    
        if not isinstance(actual, torch.Tensor):
            actual = torch.tensor(actual)
        if not isinstance(expected, torch.Tensor):
            expected = torch.tensor(expected, dtype=actual.dtype)
    
        if rtol is None and atol is None:
            rtol, atol = default_tolerances(
                actual,
                expected,
                dtype_precisions={
                    torch.float16: (1e-3, 1e-3),
                    torch.float32: (1e-4, 1e-5),
                    torch.float64: (1e-5, 1e-8),
                },
            )
    
>       torch.testing.assert_close(
            actual,
            expected,
            rtol=rtol,
            atol=atol,
            equal_nan=equal_nan,
            check_device=True,
            check_dtype=False,
            check_stride=False,
            msg=msg or None,
        )
E       AssertionError: Tensor-likes are not close!
E       
E       Mismatched elements: 1 / 36 (2.8%)
E       Greatest absolute difference: 2.6702880859375e-05 at index (0, 2, 0) (up to 1e-05 allowed)
E       Greatest relative difference: 0.0001704368187347427 at index (0, 2, 0) (up to 0.0001 allowed)

../_test_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_plac/lib/python3.11/site-packages/torch/testing/_comparison.py:1562: AssertionError

[RaulPPelaez]

This particular test fails sometimes, its stochastic in nature. Rerunning usually fixes it.

[jakirkham]

Thank Raul! 🙏

Seems to be passing now 🎉