network_onnx (macula_tweann v0.18.1)

ONNX export for neural networks.

Exports networks to ONNX (Open Neural Network Exchange) format for cross-platform inference in Python, JavaScript, C++, mobile, and edge devices.

Export a trained network: {ok, OnnxBinary} = network_onnx:to_onnx(Network) file:write_file("model.onnx", OnnxBinary)

Load in Python: import onnxruntime as ort session = ort.InferenceSession("model.onnx") outputs = session.run(None, {"input": inputs})

Load in JavaScript: const session = await ort.InferenceSession.create("model.onnx") const results = await session.run({input: tensor})

Summary

to_onnx(Network)

Export network to ONNX format.

Export network to ONNX format with options.

-spec to_onnx(network_evaluator:network()) -> {ok, binary()} | {error, term()}.

Export network to ONNX format.

-spec to_onnx(network_evaluator:network(), map()) -> {ok, binary()} | {error, term()}.

Export network to ONNX format with options.

Options: model_name - Name of the model (default: "macula_network") producer - Producer name (default: "macula-tweann")