fix: avoid removing from tuple in _get_to_kwargs

[orangeH25]

This PR fixes a bug in _get_to_kwargs where args is a tuple but the code attempted to call args.remove(arg), leading to:

    def _get_to_kwargs(self, *args, **kwargs):
        # `torch._C._nn._parse_to` can't handle `layout` argument
        for arg in args:
            if isinstance(arg, torch.layout):
>               args.remove(arg)
                ^^^^^^^^^^^
E               AttributeError: 'tuple' object has no attribute 'remove'

torchao/utils.py:722: AttributeError

Temporary test case in test/test_utils.py::TestTorchAOBaseTensor(unittest.TestCase):

    def test__get_to_kwargs_with_layout(self):
        MyClass = TorchAOBaseTensor()
        import torch
        MyClass._get_to_kwargs(torch.strided,device="cuda")

Temporary test command:

pytest --count=1 test/test_utils.py::TestTorchAOBaseTensor::test__get_to_kwargs_with_layout

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/3018

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit dc8d607 with merge base 58c3064 ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[orangeH25]

Hey @jerryzh168 @huydhn , Please help to review this one,thank you.

[jerryzh168]

@orangeH25 thanks! can you add the test to test/test_utils.py?

[orangeH25]

@orangeH25 thanks! can you add the test to test/test_utils.py?

Got it, I'll add the test

[orangeH25]

Hey @jerryzh168 , the test case has been added. Please review it again. Thanks!

[jerryzh168]

this is not the intended use for TorchAOBaseTensor, maybe you can add a test item in

ao/test/test_utils.py

Line 108 in 18dbe87

def _test_default_impls_helper(self, lp_tensor, lp_tensor_for_copy):

since _get_to_kwargs is also a default impl

[jerryzh168]

we can probably expand the op test for to:

ao/test/test_utils.py

Lines 148 to 152 in 18dbe87

	original_device = lp_tensor.device
	lp_tensor = lp_tensor.to("cuda")
	self.assertEqual(lp_tensor.device.type, "cuda")
	lp_tensor = lp_tensor.to(original_device)
	self.assertEqual(lp_tensor.device, original_device)

and also add an explicit test for _get_to_kwargs as well

[orangeH25]

Thanks, I'll modify the test

[jerryzh168]

btw, testing the higher level call to to op directly might be better here, lp_tensor.to(...)

[jerryzh168]

I'll merge for now and follow up to change the call to to