cannot load mixedbread-ai model with optimization enabled

[arnej27959]

Describe the issue

Trying to load the FP16 version of mixedbread-ai/mxbai-embed-large-v1 fails in version 1.26.0 with this error:

2026-06-18 11:33:11.071936790 [E:onnxruntime:, inference_session.cc:2742 operator()] Exception during initialization: /onnxruntime_src/onnxruntime/core/graph/graph_utils.cc:29 int onnxruntime::graph_utils::GetIndexFromName(const onnxruntime::Node&, const std::string&, bool) itr != node_args.end() was false. Attempting to get index by a name which does not exist:InsertedPrecisionFreeCast_/encoder/layer.23/output/LayerNorm/Constant_output_0for node: /embeddings/LayerNorm/Mul/SimplifiedLayerNormFusion/

it works with ORT_DISABLE_ALL, and it also works fine in older versions like onnxruntime 1.23.2 which we used until now.

To reproduce

Download
https://huggingface.co/mixedbread-ai/mxbai-embed-large-v1/resolve/main/onnx/model_fp16.onnx

and try loading it into onnxruntime, for example like this:

python3 -m venv ~/onnxruntime-env
source ~/onnxruntime-env/bin/activate
pip install --upgrade pip
pip install onnx onnxruntime==1.26.0

python3 << 'EOF'

import onnxruntime as ort
print("onnxruntime", ort.__version__)

print("LOAD 1")
options = ort.SessionOptions()
options.graph_optimization_level = ort.GraphOptimizationLevel.ORT_DISABLE_ALL
session = ort.InferenceSession("model_fp16.onnx", sess_options=options, providers=["CPUExecutionProvider"])
print("LOAD 1 OK")

print("LOAD 2")
options = ort.SessionOptions()
options.graph_optimization_level = ort.GraphOptimizationLevel.ORT_ENABLE_ALL
session = ort.InferenceSession("model_fp16.onnx", sess_options=options, providers=["CPUExecutionProvider"])
print("LOAD 2 OK")

EOF

Urgency

No response

Platform

Linux

OS Version

AlmaLinux 10

ONNX Runtime Installation

Released Package

ONNX Runtime Version or Commit ID

1.26.0

ONNX Runtime API

Python

Architecture

X64

Execution Provider

Default CPU

Execution Provider Library Version

No response

[arnej27959]

I can confirm that recompiling 1.26.0 adding #29196 as a patch fixes this problem for me.