Function export_weights 

pub fn export_weights<P, B, M>(
    path: P,
    module: &M,
    model_name: &str,
    input_shape: &[i64],
) -> Result<()>
where
    P: AsRef<Path>,
    B: Backend,
    M: Module<B>,

Export a module’s weights as an ONNX model file.

This creates a “weight-only” ONNX model: the initializer tensors contain the model’s learned parameters, and the graph describes a simple sequential pass from input to output.

Arguments

• path: output file path (typically .onnx)
• module: the trained module to export
• model_name: name for the ONNX graph
• input_shape: shape of the model’s input tensor

Example

let model = Linear::new(784, 10, true, DType::F32, &dev)?;
export_weights("model.onnx", &model, "classifier", &[1, 784])?;