Struct Interpreter 

pub struct Interpreter { /* private fields */ }

A model interpreter that holds a loaded neural network model.

The interpreter is the main entry point for MNN inference. It manages the model and can create multiple sessions for concurrent inference.

Thread Safety

The interpreter is thread-safe and can be shared across threads. Multiple sessions can be created from a single interpreter.

Example

use mnn_rs::{Interpreter, ScheduleConfig, BackendType};

// Load a model
let interpreter = Interpreter::from_file("model.mnn")?;

// Create a session with configuration
let config = ScheduleConfig::new()
    .backend(BackendType::CPU)
    .num_threads(4);

let mut session = interpreter.create_session(config)?;

// Get input tensor and fill with data
let input = session.get_input(None)?;
// ... fill input with data ...

// Run inference
session.run()?;

// Get output
let output = session.get_output(None)?;

Implementations

impl Interpreter

pub fn from_file<P: AsRef<Path>>(path: P) -> MnnResult<Self>

Create a new interpreter from a model file.

Arguments

• path - Path to the MNN model file

Returns

A new interpreter on success, or an error if the model cannot be loaded.

Example

use mnn_rs::Interpreter;

let interpreter = Interpreter::from_file("model.mnn")?;

pub fn from_bytes(data: &[u8]) -> MnnResult<Self>

Create a new interpreter from in-memory model data.

This is useful for embedding models in the binary or loading models from non-filesystem sources.

Arguments

• data - The model data as bytes

Returns

A new interpreter on success, or an error if the model cannot be loaded.

Example

use mnn_rs::Interpreter;

let model_data = include_bytes!("model.mnn");
let interpreter = Interpreter::from_bytes(model_data)?;

pub fn create_session(&self, config: ScheduleConfig) -> MnnResult<Session>

Create a new inference session.

Sessions hold the runtime state for inference and can be created with different backend configurations.

Arguments

• config - Schedule configuration specifying backend and settings

Returns

A new session on success, or an error if session creation fails.

pub fn model_path(&self) -> Option<&str>

Get the model path (if loaded from file).

pub fn business_code(&self) -> String

Get the business code (model identifier).

Returns

The business code string.

pub fn uuid(&self) -> String

Get the model UUID.

Returns

The UUID string.

pub fn inner(&self) -> *mut MNNInterpreter

Get the raw pointer to the underlying MNN Interpreter.

Safety

The returned pointer is owned by this Interpreter and must not be freed.

pub fn set_session_mode(&self, mode: SessionMode)

Set session mode.

Arguments

• mode - The session mode to set

pub fn set_cache_file<P: AsRef<Path>>(&self, path: P, key_size: usize)

Set cache file for optimization.

Arguments

• path - Path to the cache file
• key_size - Key size for cache lookup (default: 128)

pub fn update_cache(&self, session: &Session) -> MnnResult<()>

Update cache from a session.

Arguments

• session - The session to update cache from

Returns

Ok(()) on success, or an error on failure.

pub fn set_external_file<P: AsRef<Path>>(&self, path: P, flag: usize)

Set external file for model.

Arguments

• path - Path to the external file
• flag - Flag for external file processing

pub fn get_input_names(&self, session: &Session) -> Vec<String>

Get input tensor names for a session.

Arguments

• session - The session to get input names from

Returns

A vector of input tensor names.

pub fn get_output_names(&self, session: &Session) -> Vec<String>

Get output tensor names for a session.

Arguments

• session - The session to get output names from

Returns

A vector of output tensor names.

pub fn resize_tensor(&self, tensor: &mut Tensor, shape: &[i32])

Resize a tensor with new shape.

Arguments

• tensor - The tensor to resize
• shape - New shape

pub fn resize_session(&self, session: &mut Session)

Resize a session after tensor resizing.

Arguments

• session - The session to resize

pub fn get_session_flops(&self, session: &Session) -> f32

Get session FLOPS count.

Arguments

• session - The session to query

Returns

FLOPS count in millions.

pub fn get_session_op_count(&self, session: &Session) -> i32

Get session operator count.

Arguments

• session - The session to query

Returns

Operator count (approximate, based on FLOPS).

Trait Implementations

impl Debug for Interpreter

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

Formats the value using the given formatter.

impl Drop for Interpreter

fn drop(&mut self)

Executes the destructor for this type.

impl Send for Interpreter

impl Sync for Interpreter