Struct EncoderModel 

pub struct EncoderModel<'a> {
    pub graph: &'a Graph,
    pub metadata: ModelMetadata,
    pub weights: Option<&'a WeightStore>,
}

An IR Graph bundled with the model-level metadata and live weight bytes needed to encode a complete ONNX ModelProto.

Construct with Model::new and refine via Model::with_metadata / Model::with_weights. A WeightStore is required only when the graph has External-backed initializers (inline initializers carry their own bytes).

Fields

graph: &'a Graph

The graph to encode.

metadata: ModelMetadata

Model-level metadata (see ModelMetadata).

weights: Option<&'a WeightStore>

Live weight store backing any WeightRef::External initializers.

Implementations

impl<'a> Model<'a>

pub fn new(graph: &'a Graph) -> Model<'a>

A model over graph with default metadata and no external weight store.

Note on defaulted metadata: the load path does not preserve model-level metadata (it keeps only opset_imports on the Graph), so a model built this way stamps DEFAULT_IR_VERSION and empty producer / version / metadata_props fields. A rewrite path that must reproduce the original model faithfully should capture the source metadata and pass it via Model::with_metadata rather than relying on these defaults.

pub fn with_metadata(self, metadata: ModelMetadata) -> Model<'a>

Attach model-level metadata.

pub fn with_weights(self, weights: &'a WeightStore) -> Model<'a>

Attach the live WeightStore backing external initializers.