Skip to content

Remove extra defensive checks on aten ops in annotator #21058

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

Open
JakeStevens wants to merge 1 commit into
base: main
Choose a base branch
from
+4 −19
Open

Remove extra defensive checks on aten ops in annotator #21058

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
23 changes: 4 additions & 19 deletions backends/arm/quantizer/quantization_annotator.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -603,6 +603,10 @@ def _get_fixed_qparams_qspec(
torch.ops.aten.split_copy.Tensor,
torch.ops.aten.transpose.int,
torch.ops.aten.transpose_copy.int,
torch.ops.aten.moveaxis.int,
torch.ops.aten.moveaxis.intlist,
torch.ops.aten.movedim.int,
torch.ops.aten.movedim.intlist,
torch.ops.aten.t_copy.default,
torch.ops.aten.tile.default,
torch.ops.aten.flip.default,
Expand All @@ -628,25 +632,6 @@ def _get_fixed_qparams_qspec(
torch.ops.aten.detach_copy.default,
}

# Dimname has been removed from upstream PyTorch, but there may be a window
# where developers in this backend are using a mainline build of this backend
# with an older version of PyTorch.
# TODO: remove this once the build has time to be propagated and majority of
# dev expected to be unimpacted
_transpose_dimname = getattr(torch.ops.aten.transpose, "Dimname", None)
if _transpose_dimname is not None:
_one_to_one_shared_input_qspec.add(_transpose_dimname)

for _op in (
getattr(torch.ops.aten.moveaxis, "int", None),
getattr(torch.ops.aten.moveaxis, "intlist", None),
getattr(torch.ops.aten.movedim, "int", None),
getattr(torch.ops.aten.movedim, "intlist", None),
):
if _op is not None:
_one_to_one_shared_input_qspec.add(_op)


_one_to_one_shared_input_or_input_act_qspec: set[OpOverload] = {
torch.ops.aten.alias.default,
torch.ops.aten.clone.default,
Expand Down
Loading