Implementation:Microsoft Onnxruntime InferenceSession Run

Metadata

Field	Value
Implementation Name	InferenceSession_Run
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	Python_Inference_Pipeline
Pair	5 of 6

Overview

API documentation for the session.run() method, which executes the forward pass on a loaded ONNX model and returns the requested output tensors as numpy arrays.

API Signature

session.run(output_names: list[str] | None, input_feed: dict[str, numpy.ndarray]) -> list[numpy.ndarray]

Import

No additional import is needed beyond the InferenceSession and numpy (for input preparation).

Code Reference

Reference	Location
Usage example	docs/python/examples/plot_load_and_predict.py:L54

I/O Contract

Inputs

Parameter	Type	Required	Description
`output_names`	`list[str]` or `None`	Yes	List of output tensor names to retrieve. Pass `None` to retrieve all outputs.
`input_feed`	`dict[str, numpy.ndarray]`	Yes	Dictionary mapping input tensor names to numpy arrays containing the input data.

Outputs

Output	Type	Description
return value	`list[numpy.ndarray]`	A list of numpy arrays, one per requested output, in the order specified by `output_names`.

Usage Example

# Run with specific output
res = sess.run([output_name], {input_name: x})

# Run with all outputs
res = sess.run(None, {input_name: x})

From the source example at docs/python/examples/plot_load_and_predict.py:L54:

res = sess.run([output_name], {input_name: x})
print(res)

From the profiling example at docs/python/examples/plot_profiling.py:L42:

res = sess.run(None, {input_name: x})

Complete Inference Pipeline

import numpy
import onnxruntime as rt

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

# Inspect metadata
input_name = sess.get_inputs()[0].name
output_name = sess.get_outputs()[0].name

# Prepare input
x = numpy.random.random((3, 4, 5)).astype(numpy.float32)

# Run inference
res = sess.run([output_name], {input_name: x})
print(res[0])

Related Pages

Page Connections

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

Principle

Implementation

Heuristic

Environment