Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

MPS: isin_mps_friendly can support 0D tensors #34538

Merged
merged 5 commits into from
Nov 4, 2024
Merged

MPS: isin_mps_friendly can support 0D tensors #34538

Show file tree
Hide file tree
Changes from 4 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
4391007
apply fix
gante Oct 31, 2024
301f20c
tested
gante Oct 31, 2024
06d7e18
make fixup
gante Oct 31, 2024
dc38c62
Merge branch 'main' into is_in_fix
gante Oct 31, 2024
d641307
Merge branch 'main' into is_in_fix
gante Nov 4, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 4 additions & 1 deletion src/transformers/pytorch_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -314,14 +314,17 @@ def isin_mps_friendly(elements: torch.Tensor, test_elements: torch.Tensor | int)

Args:
elements (`torch.Tensor`): Input elements
test_elements (`torch.Tensor`): The elements to check against.
test_elements (`torch.Tensor` or `int`): The elements to check against.

Returns:
`torch.Tensor`: A boolean tensor of the same shape as `elements` that is True for `elements` in `test_elements`
and False otherwise
"""

if elements.device.type == "mps" and not is_torch_greater_or_equal_than_2_4:
test_elements = torch.tensor(test_elements)
if test_elements.ndim == 0:
test_elements = test_elements.unsqueeze(0)
Comment on lines +325 to +327
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

we ought to open the issue on pytorch / link the tracker !

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

On torch >=2.4 this is fixed, i.e. torch.isin is supported with mps

(in other words, this fix only affects older versions of torch) 🤗

return elements.tile(test_elements.shape[0], 1).eq(test_elements.unsqueeze(1)).sum(dim=0).bool().squeeze()
else:
# Note: don't use named arguments in `torch.isin`, see https://github.com/pytorch/pytorch/issues/126045
Expand Down
7 changes: 6 additions & 1 deletion tests/utils/test_modeling_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1711,7 +1711,12 @@ def test_isin_mps_friendly(self):
torch.isin(random_ids, random_test_integer), isin_mps_friendly(random_ids, random_test_integer)
)
)
# We can match against an tensor of integers
# We can match against an 0D tensor
random_test_tensor = torch.randint(0, 100, (1,)).squeeze()
self.assertTrue(
torch.equal(torch.isin(random_ids, random_test_tensor), isin_mps_friendly(random_ids, random_test_tensor))
)
# We can match against an 1D tensor (with many items)
random_test_tensor = torch.randint(0, 100, (10,))
self.assertTrue(
torch.equal(torch.isin(random_ids, random_test_tensor), isin_mps_friendly(random_ids, random_test_tensor))
Expand Down
Loading