Failing tests on `aarch64-darwin`

[drupol]

🐛 Bug

Some tests are failing on aarch64-darwin, e.g.:

>       self._compute_cavs_interpret(
            [["striped", "random"], ["ceo", "random"]],
            True,
            -1.0,  # acc is not defined, this field will not be asserted
            0.5000,
            8.185208066890937e-09,
            CustomClassifier_WO_Returning_Metrics(),
            processes=2,
        )

tests/concept/test_tcav.py:988: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
tests/concept/test_tcav.py:852: in _compute_cavs_interpret
    self.assertAlmostEqual(
E   AssertionError: 1.0 != 0.5 within 0.0001 delta (0.5 difference)
---

To Reproduce

Steps to reproduce the behavior:

1. Run the tests of version 0.7 on Darwin platform, aarch64 arch.
2. Find the failing tests at https://paste.glepage.com/upload/bear-pug-viper

Expected behavior

Environment

Describe the environment used for Captum

 - Captum / PyTorch Version (e.g., 1.0 / 0.4.0): 0.7.0
 - OS (e.g., Linux): Linux
 - How you installed Captum / PyTorch (`conda`, `pip`, source): source
 - Build command you used (if compiling from source):
 - Python version: 3.12.5
 - CUDA/cuDNN version: none
 - GPU models and configuration: none
 - Any other relevant information:

## Additional context

<!-- Add any other context about the problem here. -->