Struct OnnxProviderConfig 

pub struct OnnxProviderConfig {
    pub model_path: PathBuf,
    pub tokenizer_path: PathBuf,
    pub dimension: u32,
    pub pooling: Pooling,
    pub normalize: bool,
    pub max_tokens: usize,
    pub intra_op_threads: Option<usize>,
}

What a deployment declares about its model.

Fields

model_path: PathBuf

tokenizer_path: PathBuf

dimension: u32

Vector width the model emits. Checked against the model’s own output on the first batch, not taken on trust.

pooling: Pooling

normalize: bool

L2-normalize the pooled vector. On for cosine similarity, which is what the gear’s index serves.

max_tokens: usize

Longest token sequence handed to the model; longer inputs are truncated.

intra_op_threads: Option<usize>

ONNX Runtime intra-op threads. None leaves the runtime’s default.

Implementations

impl OnnxProviderConfig

pub fn new( model_path: impl Into<PathBuf>, tokenizer_path: impl Into<PathBuf>, ) -> Self

The MiniLM-L6-v2 defaults ADR-0005 describes: 384 dimensions, mean pooling, L2-normalized.

Trait Implementations

impl Clone for OnnxProviderConfig

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for OnnxProviderConfig

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

Formats the value using the given formatter.