Open ListonQH opened 2 months ago
Thank you for reporting this issue. It seems to be caused by fusing transformation. @itikhono please take a look.
We're currently working on this issue. It requires some investigations and changes in Convolution transformations.
The issue has been fixed https://github.com/openvinotoolkit/openvino/pull/25876 @ListonQH , please verify
OpenVINO Version
openvino - 2024.2.0
Operating System
Ubuntu 20.04 (LTS)
Device used for inference
CPU
Framework
PyTorch
Model used
No response
Issue description
Describe
When I try conver an ONNX to OV-IR, got error: Argument shapes are inconsistent.
ONNX as the picture:
More
I also run it by onnxruntime and get target result:
Env
Step-by-step reproduction
when I try convert onnx to OV-IR:
Error:
class TestModel(torch.nn.Module): def init(self, *args, *kwargs) -> None: super().init(args, **kwargs) self.constan1 = torch.rand(size=(1, 4, 18, 18))
m = TestModel() m.eval() torch.onnx.export(m, torch.rand(1, 3, 20, 20),f='test.onnx', input_names=['input'], output_names=['output'], do_constant_folding=False)