Implementation:Microsoft Onnxruntime InferenceSession Init

Metadata

Field	Value
Implementation Name	InferenceSession_Init
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	2 of 6

Overview

API documentation for the onnxruntime.InferenceSession() constructor, which loads an ONNX model and prepares it for inference with specified execution providers.

API Signature

onnxruntime.InferenceSession(model_path, sess_options=None, providers=None)

Import

from onnxruntime import InferenceSession

Code Reference

Reference	Location
Import definition	onnxruntime/__init__.py:L82 (imported from capi.onnxruntime_inference_collection)
Usage example	docs/python/examples/plot_load_and_predict.py:L25

I/O Contract

Inputs

Parameter	Type	Required	Description
`model_path`	`str` or `bytes`	Yes	Path to an ONNX model file on disk, or a serialized ONNX model as bytes.
`sess_options`	`SessionOptions`	No	A configured SessionOptions object controlling optimization, profiling, and threading behavior.
`providers`	`list[str]`	No	Ordered list of execution provider names. Providers are tried in order for each operator.

Outputs

Output	Type	Description
return value	`InferenceSession`	A fully initialized session ready for inference via the `.run()` method.

Usage Example

import onnxruntime as rt

# Basic usage: load model with all available providers
sess = rt.InferenceSession("model.onnx", providers=rt.get_available_providers())

# With explicit options and provider selection
options = rt.SessionOptions()
sess = rt.InferenceSession(
    "model.onnx",
    options,
    providers=['CUDAExecutionProvider', 'CPUExecutionProvider']
)

From the source example at docs/python/examples/plot_load_and_predict.py:L25:

example1 = get_example("sigmoid.onnx")
sess = rt.InferenceSession(example1, providers=rt.get_available_providers())

From the profiling example at docs/python/examples/plot_profiling.py:L38:

sess = rt.InferenceSession(onnx_model_str, providers=rt.get_available_providers())

Related Pages

Page Connections

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

Principle

Implementation

Heuristic

Environment