Skip to content

[ONNX] Fix unsqueeze constant expr #9146

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
masahi merged 1 commit into from
Sep 29, 2021
Merged

[ONNX] Fix unsqueeze constant expr #9146

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
20 changes: 14 additions & 6 deletions python/tvm/relay/frontend/onnx.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1468,18 +1468,26 @@ class Unsqueeze(OnnxOpConverter):
"""Operator converter for Unsqueeze."""

@classmethod
def _impl_v1(cls, inputs, attr, params):
axes = sorted(attr["axes"])
def run_calculation(cls, tensor, axes):
axes = sorted(axes)
for axis in axes:
inputs[0] = _op.expand_dims(inputs[0], axis=axis, num_newaxis=1)
return inputs[0]
tensor = _op.expand_dims(tensor, axis=axis, num_newaxis=1)
return tensor

@classmethod
def _impl_v12(cls, inputs, attr, params):
def _impl_v1(cls, inputs, attr, params):
return cls.run_calculation(inputs[0], attr["axes"])

@classmethod
def _impl_v13(cls, inputs, attr, params):
if isinstance(inputs[1], _expr.Constant):
Copy link
Member

@masahi masahi Sep 29, 2021

Choose a reason for hiding this comment

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

I thought axes could also be an Expr involving only constants, but I hope such weird usage wouldn't happen.

constant_axes = list(inputs[1].data.numpy())
constant_axes = list(map(int, constant_axes))
return cls.run_calculation(inputs[0], constant_axes)

rank_input = len(infer_type(inputs[0]).checked_type.shape)
num_new_axis = int(infer_type(inputs[1]).checked_type.shape[0])
axes = relay.split(inputs[1], num_new_axis).astuple()

result = inputs[0]

# TODO (AndrewZhaoLuo): investigate performance issues with consecutive
Expand Down