Struct Session

pub struct Session { /* private fields */ }

An inference session represents a model that is loaded and ready to perform inference on the GPU.

Examples

Basic usage:

let mut session = Session::from_path("path/to/model.onnx").await.unwrap();

Implementations

impl Session

pub async fn from_path<P: AsRef<Path>>(path: P) -> Result<Session, SessionError>

pub async fn from_path_with_config<P: AsRef<Path>>( path: P, config: &SessionConfig, ) -> Result<Session, SessionError>

pub async fn from_bytes(bytes: &[u8]) -> Result<Session, SessionError>

Read an ONNX model from bytes and create a session, using default session config.

pub async fn from_bytes_with_config( bytes: &[u8], config: &SessionConfig, ) -> Result<Session, SessionError>

Read an ONNX model from bytes and create a session with the specified session config.

pub async fn from_model_with_config( model: ModelProto, config: &SessionConfig, ) -> Result<Session, SessionError>

Create a session using the provided onnx::ModelProto and session config.

pub async fn from_model(model: ModelProto) -> Result<Session, SessionError>

Create a Session given an ONNX model, using default configuration.

pub async fn run<'a>( &self, inputs: &HashMap<String, InputTensor<'a>>, ) -> Result<HashMap<String, OutputTensor>, SessionError>

Perform inference given the inputs provided and return all the outputs the model was compiled to return.