Skip to content

feat(atenlib): arange with overloads #285

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

Merged
merged 9 commits into from
Jan 9, 2023
Merged

feat(atenlib): arange with overloads #285

Changes from 5 commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
7cfae57
feat(atenlib): arange with overloads
justinchuby Jan 6, 2023
7e24cf9
def
justinchuby Jan 6, 2023
3962f7a
args
justinchuby Jan 6, 2023
be2d43f
Merge branch 'main' into justinchu/arange
justinchuby Jan 6, 2023
1d598d4
Update core.py
justinchuby Jan 6, 2023
20bbca1
Merge branch 'main' into justinchu/arange
justinchuby Jan 7, 2023
a10443d
Test
justinchuby Jan 7, 2023
8074704
fix typing
justinchuby Jan 8, 2023
0fba26f
typing
justinchuby Jan 8, 2023
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
31 changes: 29 additions & 2 deletions onnxscript/function_libs/torch_aten/ops/core.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -240,10 +240,37 @@ def aten_any(self: TensorType) -> TensorType:
raise NotImplementedError()


def aten_arange(end: float) -> TensorType:
@torch_op("aten::arange")
def aten_arange(end: TReal, dtype: int = -1) -> TReal:
# arange(Scalar end, *, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor

raise NotImplementedError()
if dtype != -1:
end = op.Cast(end, to=dtype)

return op.Range(0, end, 1)


@torch_op("aten::arange", overload=True)
def aten_arange_start(start: TReal, end: TReal, dtype: int = -1) -> TReal:
# arange.start(Scalar start, Scalar end, *, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor

if dtype != -1:
Copy link
Contributor

Choose a reason for hiding this comment

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

Since TReal has limited dtypes, what if the given dtype doesn't exist in the scope of TReal?

Should we set a default value it to dtype in case the given value of dtype is not in the scope of TReal?

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

Good point. I am considering if I should cast after Range

Copy link
Contributor

Choose a reason for hiding this comment

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

Good point. I am considering if I should cast after Range

If we pass start/end/step with an unexpected dtype into op.Range, will it work successfully?

Copy link
Collaborator Author

@justinchuby justinchuby Jan 8, 2023 •
edited
Loading

Choose a reason for hiding this comment

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

Updated. It will work if dtype is specified. Otherwise they need to be casted before entering the function.

start = op.Cast(start, to=dtype)
end = op.Cast(end, to=dtype)

return op.Range(start, end, 1)


@torch_op("aten::arange", overload=True)
def aten_arange_start_step(start: TReal, end: TReal, step: TReal, dtype: int = -1) -> TReal:
# arange.start_step(Scalar start, Scalar end, Scalar step=1, *, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor

if dtype != -1:
start = op.Cast(start, to=dtype)
end = op.Cast(end, to=dtype)
step = op.Cast(step, to=dtype)

return op.Range(start, end, step)


def aten_arccos(self: TensorType) -> TensorType:
Expand Down