TreeEnsemble post_transform LOGISTIC does not work

[bartleusink]

Describe the issue

According to https://onnx.ai/onnx/operators/onnx_aionnxml_TreeEnsemble.html the following post_transform options are available for the TreeEnsemble node:

• NONE (0)
• SOFTMAX (1)
• LOGISTIC (2)
• SOFTMAX_ZERO (3)
• PROBIT (4)

However, when I create a simple model with one target the LOGISTIC transform is not used.

To reproduce

I've attached a minimal example of a TreeEnsemble with post_transform set to logistic.
tree_ensemble.onnx.zip

Code to reproduce the issue:

import onnxruntime as ort

session = ort.InferenceSession("tree_ensemble.onnx")
print(session.run(None, {"X": [[1, 0, 1, 0]]}))

This prints [array([[10.3366785]], dtype=float32)] while I would have expected a number between 0 and 1.

Urgency

No response

Platform

Mac

OS Version

26.1

ONNX Runtime Installation

Released Package

ONNX Runtime Version or Commit ID

1.23.2

ONNX Runtime API

Python

Architecture

ARM64

Execution Provider

Default CPU

Execution Provider Library Version

No response