Implementation:Microsoft Onnxruntime SessionOptions Init

Metadata

Field	Value
Implementation Name	SessionOptions_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	1 of 6

Overview

API documentation for the onnxruntime.SessionOptions() constructor, which creates a configuration object for tuning ONNX Runtime inference session behavior.

API Signature

onnxruntime.SessionOptions()

Import

from onnxruntime import SessionOptions, GraphOptimizationLevel

Code Reference

Reference	Location
Import definition	onnxruntime/__init__.py:L51
Usage example	docs/python/examples/plot_profiling.py:L49-51

I/O Contract

Inputs

Parameter	Type	Required	Description
(none)	--	--	The constructor takes no arguments.

Outputs

Output	Type	Description
return value	SessionOptions	A configured SessionOptions object with default settings.

Configurable Properties

Property	Type	Default	Description
enable_profiling	bool	False	Enables performance profiling for the session.
graph_optimization_level	GraphOptimizationLevel	ORT_ENABLE_BASIC	Controls which graph optimization passes are applied.
intra_op_num_threads	int	0 (auto)	Number of threads for intra-operator parallelism.
inter_op_num_threads	int	0 (auto)	Number of threads for inter-operator parallelism.
execution_mode	ExecutionMode	Sequential	Sequential or parallel operator execution.
enable_mem_pattern	bool	True	Enables memory pattern optimization.

Usage Example

import onnxruntime as rt

# Create a SessionOptions object
options = rt.SessionOptions()

# Enable profiling
options.enable_profiling = True

# Set graph optimization to maximum
options.graph_optimization_level = rt.GraphOptimizationLevel.ORT_ENABLE_ALL

# Pass options to InferenceSession
sess = rt.InferenceSession("model.onnx", options, providers=rt.get_available_providers())

The above pattern is derived from the profiling example at docs/python/examples/plot_profiling.py:L49-51:

options = rt.SessionOptions()
options.enable_profiling = True
sess_profile = rt.InferenceSession(onnx_model_str, options, providers=rt.get_available_providers())

Related Pages

• Principle:Microsoft_Onnxruntime_Session_Options_Configuration
• Implementation:Microsoft_Onnxruntime_InferenceSession_Init
• Environment:Microsoft_Onnxruntime_Python_Inference_Environment
• Heuristic:Microsoft_Onnxruntime_Graph_Optimization_Level_Selection
• Heuristic:Microsoft_Onnxruntime_Threading_Configuration_Tips