Struct OrtSessionConfig 

pub struct OrtSessionConfig {
    pub intra_threads: Option<usize>,
    pub inter_threads: Option<usize>,
    pub parallel_execution: Option<bool>,
    pub optimization_level: Option<OrtGraphOptimizationLevel>,
    pub execution_providers: Option<Vec<OrtExecutionProvider>>,
    pub enable_mem_pattern: Option<bool>,
    pub log_severity_level: Option<i32>,
    pub log_verbosity_level: Option<i32>,
    pub session_config_entries: Option<HashMap<String, String>>,
}

Configuration for ONNX Runtime sessions.

This struct contains various configuration options for ONNX Runtime sessions, including threading, memory management, and optimization settings.

Fields

intra_threads: Option<usize>

Number of threads used to parallelize execution within nodes

inter_threads: Option<usize>

Number of threads used to parallelize execution across nodes

parallel_execution: Option<bool>

Enable parallel execution mode

optimization_level: Option<OrtGraphOptimizationLevel>

Graph optimization level

execution_providers: Option<Vec<OrtExecutionProvider>>

Execution providers in order of preference

enable_mem_pattern: Option<bool>

Enable memory pattern optimization

log_severity_level: Option<i32>

Log severity level (0=Verbose, 1=Info, 2=Warning, 3=Error, 4=Fatal)

log_verbosity_level: Option<i32>

Log verbosity level

session_config_entries: Option<HashMap<String, String>>

Session configuration entries (key-value pairs)

Implementations

impl OrtSessionConfig

pub fn new() -> OrtSessionConfig

Creates a new OrtSessionConfig with default values.

pub fn with_intra_threads(self, threads: usize) -> OrtSessionConfig

Sets the number of intra-op threads.

pub fn with_inter_threads(self, threads: usize) -> OrtSessionConfig

Sets the number of inter-op threads.

pub fn with_parallel_execution(self, enabled: bool) -> OrtSessionConfig

Enables or disables parallel execution.

pub fn with_optimization_level( self, level: OrtGraphOptimizationLevel, ) -> OrtSessionConfig

Sets the graph optimization level.

pub fn with_execution_providers( self, providers: Vec<OrtExecutionProvider>, ) -> OrtSessionConfig

Sets the execution providers, in order of preference.

pub fn add_execution_provider( self, provider: OrtExecutionProvider, ) -> OrtSessionConfig

Appends a single execution provider.

pub fn with_memory_pattern(self, enable: bool) -> OrtSessionConfig

Enables or disables memory pattern optimization.

pub fn with_log_severity_level(self, level: i32) -> OrtSessionConfig

Sets the log severity level (0=Verbose, 1=Info, 2=Warning, 3=Error, 4=Fatal).

pub fn with_log_verbosity_level(self, level: i32) -> OrtSessionConfig

Sets the log verbosity level.

pub fn add_config_entry<K, V>(self, key: K, value: V) -> OrtSessionConfig

where K: Into<String>, V: Into<String>,

Adds a session configuration entry.

pub fn with_coreml_config(self, config: OrtCoreMLConfig) -> OrtSessionConfig

Sets advanced options for any CoreML execution provider in this session.

pub fn get_intra_threads(&self) -> usize

Effective intra-op thread count, defaulting to available parallelism.

pub fn get_inter_threads(&self) -> usize

Effective inter-op thread count, defaulting to 1.

pub fn get_optimization_level(&self) -> OrtGraphOptimizationLevel

Effective graph optimization level, defaulting to OrtGraphOptimizationLevel::default().

pub fn get_execution_providers(&self) -> Vec<OrtExecutionProvider>

Configured execution providers, defaulting to CPU.

pub fn has_accelerator_provider(&self) -> bool

Returns whether an explicitly configured hardware accelerator is present.

execution_providers is a preference-ordered list: ONNX Runtime lets each provider claim graph nodes in list order, and CPU can claim almost any node, so a CPU entry listed first effectively runs the session on CPU regardless of what accelerators follow it. Only the first provider therefore determines whether this is an accelerated configuration. No provider configuration, an empty provider list, and a CPU-first list (including CPU alone) all use CPU-oriented pipeline defaults; an accelerator listed first with a CPU fallback after it (CUDA, TensorRT, DirectML, OpenVINO, CoreML, or WebGPU) counts as accelerated.

Trait Implementations

impl Clone for OrtSessionConfig

fn clone(&self) -> OrtSessionConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for OrtSessionConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.

impl Default for OrtSessionConfig

fn default() -> OrtSessionConfig

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for OrtSessionConfig

fn deserialize<__D>( __deserializer: __D, ) -> Result<OrtSessionConfig, <__D as Deserializer<'de>>::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for OrtSessionConfig

fn serialize<__S>( &self, __serializer: __S, ) -> Result<<__S as Serializer>::Ok, <__S as Serializer>::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.