onnxruntime errors out due to ORT_ENABLE_EXTENDED optimization: Can't remove node Transpose as it still has output edges.

[coffezhou]

Expected behavior

onnxruntime can run the model when using the optimization ORT_ENABLE_EXTENDED.

Actual behavior

When using the optimization ORT_ENABLE_EXTENDED, onnxruntime crashes.

Traceback (most recent call last):
  File "/home/carla/Documents/test_onnxruntime/0322/test.py", line 15, in <module>
    ort_session = onnxruntime.InferenceSession(
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/carla/anaconda3/envs/onnxruntime/lib/python3.11/site-packages/onnxruntime/capi/onnxruntime_inference_collection.py", line 472, in __init__
    self._create_inference_session(providers, provider_options, disabled_optimizers)
  File "/home/carla/anaconda3/envs/onnxruntime/lib/python3.11/site-packages/onnxruntime/capi/onnxruntime_inference_collection.py", line 544, in _create_inference_session
    sess.initialize_session(providers, provider_options, disabled_optimizers)
onnxruntime.capi.onnxruntime_pybind11_state.RuntimeException: [ONNXRuntimeError] : 6 : RUNTIME_EXCEPTION : Exception during initialization: /onnxruntime_src/onnxruntime/core/graph/graph.cc:4037 bool onnxruntime::Graph::RemoveNode(onnxruntime::NodeIndex) node->GetOutputEdgesCount() == 0 was false. Can't remove node Transpose as it still has output edges.

To reproduce

Environment

OS: Ubuntu 20.04
onnxruntime: 1.22.0.dev20250320003

Steps to reproduce

This bug can be reproduced by the following code with the model in the attachment. For the model, it can be correctly checked by onnx.checker.check_model. However, if we replace ORT_ENABLE_EXTENDED with ORT_ENABLE_BASIC, every thing is ok.

from typing import Dict, List, Literal, Optional
import sys

import numpy as np
import onnx
import onnxruntime

model_path = "bugs/onnx_output5932/model.onnx"
onnx_model = onnx.load(model_path)
onnx.checker.check_model(onnx_model)

sess_options = onnxruntime.SessionOptions()              
sess_options.graph_optimization_level = onnxruntime.GraphOptimizationLevel.ORT_ENABLE_EXTENDED

ort_session = onnxruntime.InferenceSession(
                     onnx_model.SerializeToString(), sess_options, providers=["CPUExecutionProvider"]
                  )

model.zip

Urgency

No response

Platform

Linux

OS Version

Ubuntu 20.04

ONNX Runtime Installation

Built from Source

ONNX Runtime Version or Commit ID

1.22.0.dev20250320003

ONNX Runtime API

Python

Architecture

X64

Execution Provider

Default CPU

Execution Provider Library Version

No response

[github-actions]

This issue has been automatically marked as stale due to inactivity and will be closed in 30 days if no further activity occurs. If further support is needed, please provide an update and/or more details.