[PyTorch] : implement support for replicated{1,2,3} pad #28271
Conversation
Signed-off-by: 11happy <[email protected]>
|
humble ping! |
|
build_jenkins |
Signed-off-by: 11happy <[email protected]>
Signed-off-by: 11happy <[email protected]>
|
build_jenkins |
Signed-off-by: 11happy <[email protected]>
|
I have fixed the code style , ran clang-format. |
|
build_jenkins |
Signed-off-by: 11happy <[email protected]>
|
humble ping! |
|
build_jenkins |
Signed-off-by: 11happy <[email protected]>
|
build_jenkins |
| class TestReplicatePad1D(PytorchLayerTest): | ||
| def _prepare_input(self, ndim=4, dtype="float32"): | ||
| import numpy as np | ||
| input_5d_shape = [5,9,1,1,2,4] |
There was a problem hiding this comment.
please remove input_5d_shape and ndim from here and add normal input_shape to @pytest.mark.parametrize.
The same comment for other tests
There was a problem hiding this comment.
done,
please suggest the input_shapes to keep, currently kept random from my end,
thank you
Signed-off-by: 11happy <[email protected]>
Signed-off-by: 11happy <[email protected]>
|
I think in reflection pad as well we can change |
Yes, please do the same for reflection_pad, and constant_padnd as it is similar. You would need to add the entries in op_table.cpp for TS table. |
Signed-off-by: 11happy <[email protected]>
Signed-off-by: 11happy <[email protected]>
Signed-off-by: 11happy <[email protected]>
Signed-off-by: 11happy <[email protected]>
src/frontends/pytorch/src/op/pad.cpp
Outdated
| num_inputs_check(context, 2, 2); | ||
| auto data = context.get_input(0); | ||
| auto paddings = context.const_input<std::vector<int64_t>>(1); | ||
| Output<Node> pad_value = context.mark_node(v0::Constant::create(element::f32, Shape{}, {0})); | ||
| return translate_pad_common(context, data, paddings, pad_value, "reflect"); | ||
| } | ||
|
|
||
| OutputVector translate_replicate_pad_nd(const NodeContext & context) { |
There was a problem hiding this comment.
| OutputVector translate_replicate_pad_nd(const NodeContext & context) { | |
| OutputVector translate_replicate_pad_nd(const NodeContext& context) { |
| {"aten.replicate_pad1d.default", op::translate_replicate_pad_nd}, | ||
| {"aten.replicate_pad2d.default", op::translate_replicate_pad_nd}, | ||
| {"aten.replicate_pad3d.default", op::translate_replicate_pad_nd}, |
There was a problem hiding this comment.
Tests for torch.export fail. I can see that names are wrong, but tests fail on torch side, please check them
There was a problem hiding this comment.
I am not sure but If I am trying to open the logs
its just loading & loading not showing the logs. How may I acces the failed logs.
Thank you
There was a problem hiding this comment.
I have corrected the names. Thank you
Signed-off-by: 11happy <[email protected]>
Overview:
This pull request fixes #23322.
Testing:
CC: