openvinotoolkit / openvino

OpenVINO™ is an open-source toolkit for optimizing and deploying AI inference

https://docs.openvino.ai

Apache License 2.0

7.29k stars 2.27k forks source link

[Good First Issue]: Support aten::reflection_pad1d, aten::reflection_pad3d for pytorch models #23321

Open mvafin opened 8 months ago

mvafin commented 8 months ago

Context

OpenVINO component responsible for support of PyTorch models is called as PyTorch Frontend (PT FE). PT FE converts a model represented as TorchScript model to a model in OpenVINO opset. aten::reflection_pad2d is already supported, what needs to be done is to also support other sizes.

What needs to be done?

Implement conversion rule and/or transformation to support the new operation.
Implement operation tests in tests/layer_tests/pytorch_tests. Please consider different data types, but keep reasonable number of test cases

Example Pull Requests

18998

Resources

Contact points

@openvinotoolkit/openvino-pytorch-frontend-maintainers

Ticket

TBD

thinhngo-x commented 8 months ago

.take

github-actions[bot] commented 8 months ago

Thank you for looking into this issue! Please let us know if you have any questions or require any help.

thinhngo-x commented 8 months ago

Hello @mvafin, can you please show me where the aten::reflection_pad2dhas been implemented? Thanks.

mvafin commented 8 months ago

@thinhngo-x It is supported here: https://github.com/openvinotoolkit/openvino/blob/a78d9148cae10749d0c0e2d005012e366b73307d/src/frontends/pytorch/src/transforms/prim_list_construct_pad.cpp#L87

p-wysocki commented 7 months ago

Hello @thinhngo-x, are you still working on this?

duydl commented 6 months ago

.take

github-actions[bot] commented 6 months ago

Thank you for looking into this issue! Please let us know if you have any questions or require any help.

duydl commented 6 months ago

Hi, I have some question.

How pytorch/src/transforms/prim_list_construct_pad.cpp is related to pytorch/src/op/pad.cpp.
The test for ReflectionPad in test_pad.py passes (also torch.nn.ReflectionPad1d is the same as ReflectionPad2d and ReflectionPad3d i.e accepts 4-tuple and returns same output). https://github.com/openvinotoolkit/openvino/blob/68bdbbe0fba08386677071c83018be9d07f58916/tests/layer_tests/pytorch_tests/test_pad.py#L205 So it seems to me the relevant logic is already converted to openVINO ops.

Thank you.

jagadeeshmadinni commented 2 weeks ago

.take

github-actions[bot] commented 2 weeks ago

Thank you for looking into this issue! Please let us know if you have any questions or require any help.