How can I output every node's output shape when infer with onnx models with lots of if branches

[DamonsJ]

Describe the issue

I want to print every layers' output shape
but this model
https://github.com/snakers4/silero-vad/blob/master/src/silero_vad/data/silero_vad.onnx
has so many if branches

it's mean you can only print layers' output when you really infer this model

how can I do this?

To reproduce

input = torch.ones(1,512)
state = torch.ones(2,1,128)
sr = torch.tensor(16000, dtype=torch.int64)
sess = ort.InferenceSession("silero_vad.onnx")
outputs = sess.run(None, {"input": input.numpy(),
"state": state.numpy(),
"sr": sr.numpy()})

Urgency

No response

Platform

Mac

OS Version

any

ONNX Runtime Installation

Released Package

ONNX Runtime Version or Commit ID

any

ONNX Runtime API

Python

Architecture

X64

Execution Provider

Default CPU

Execution Provider Library Version

No response

[tianleiwu]

One way to debug it to build onnxruntime package from source and adding --cmake_extra_defines onnxruntime_DEBUG_NODE_INPUTS_OUTPUTS=1 in build command line.

Then you should be able to see shape info of each input/output of each node.

See https://onnxruntime.ai/docs/build/inferencing.html