Struct TotoModelBuilder 

pub struct TotoModelBuilder { /* private fields */ }

Fluent constructor for TotoModel: point it at a GGUF file, optionally layer on a parsed config.json and/or explicit InferConfig overrides, then call build.

use zsfm_toto::infer::TotoModel;

let model = TotoModel::builder("toto.gguf")
    .config_json(&serde_json::json!({ "d_model": 2048, "num_layers": 48 }))
    .with_compute_f64(true)
    .build()?;

Implementations

impl TotoModelBuilder

pub fn config(self, config: InferConfig) -> Self

Replace the whole config, e.g. one built by hand via InferConfig::default().with_d_model(...).with_num_layers(...).

pub fn config_json(self, v: &Value) -> Self

Populate the config from a parsed HuggingFace config.json.

pub fn with_compute_f64(self, v: bool) -> Self

Run the forward pass in F64 (double precision) to match PyTorch numerical accuracy. Uses ~2× memory. Default: F32.

pub fn build(self) -> Result<TotoModel>