Implementation:Microsoft Onnxruntime InferenceSession Run For Prediction

Metadata

Field	Value
Implementation Name	InferenceSession_Run_For_Prediction
Repository	Microsoft_Onnxruntime
Source Repository	https://github.com/microsoft/onnxruntime
Type	API Doc
Language	Python
Domain	ML_Inference, Model_Optimization
Last Updated	2026-02-10
Workflow	Train_Convert_Predict
Pair	5 of 5

Overview

API documentation for using onnxruntime.InferenceSession.run() to perform optimized production inference on a validated ONNX model, retrieving both class labels and probability outputs.

API Signature

session.run([output_name], {input_name: X_test.astype(numpy.float32)})[0]

Import

import onnxruntime as rt
import numpy

Code Reference

Reference	Location
Prediction and probability retrieval	docs/python/examples/plot_train_convert_predict.py:L79-101

I/O Contract

Inputs

Parameter	Type	Required	Description
`InferenceSession`	`rt.InferenceSession`	Yes	A loaded ONNX inference session with the validated model.
Test data	`numpy.ndarray` (float32)	Yes	Input feature data cast to `numpy.float32` matching the model's input specification.

Outputs

Output	Type	Description
Predictions (labels)	`numpy.ndarray`	Predicted class labels as an integer array.
Probabilities	`numpy.ndarray`	Probability distributions over classes for each sample.

Usage Example

import onnxruntime as rt
import numpy

# Load validated ONNX model
sess = rt.InferenceSession("logreg_iris.onnx", providers=rt.get_available_providers())

# Get input/output tensor names
input_name = sess.get_inputs()[0].name
label_name = sess.get_outputs()[0].name
prob_name = sess.get_outputs()[1].name

# Get predicted labels
pred = sess.run([label_name], {input_name: X_test.astype(numpy.float32)})[0]

# Get probability distributions
probs = sess.run([prob_name], {input_name: X_test.astype(numpy.float32)})[0]

From the source at docs/python/examples/plot_train_convert_predict.py:L79-101:

input_name = sess.get_inputs()[0].name
label_name = sess.get_outputs()[0].name

pred_onx = sess.run([label_name], {input_name: X_test.astype(numpy.float32)})[0]

prob_name = sess.get_outputs()[1].name
prob_rt = sess.run([prob_name], {input_name: X_test.astype(numpy.float32)})[0]

Single-Sample Inference (Web Service Pattern)

def sess_predict(x):
    return sess.run([label_name], {input_name: x.astype(numpy.float32)})[0]

def sess_predict_proba(x):
    return sess.run([prob_name], {input_name: x.astype(numpy.float32)})[0]

# Process one sample at a time
for i in range(X_test.shape[0]):
    sample = X_test[i:i+1]
    prediction = sess_predict(sample)
    probability = sess_predict_proba(sample)

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment