airpack_scripts.tf2.run_benchmarks

Utility script to benchmark the data rate that a neural network will support.

Module Contents

airpack_scripts.tf2.run_benchmarks.benchmark(onnx_file)

Convert an onnx file to an optimized plan file and benchmark the maximum data rate throughput.

See also

This is a wrapper for airpack.deploy.trt.onnx2plan() and airpack.deploy.trt.plan_bench() that is setup with default values for the airpack.tf2.model.default_network(). If non-default parameters are needed for your customization, it is better practice to call directly the airpack.deploy.trt.onnx2plan() and airpack.tf2.model.default_network() instead of this demonstration wrapper.

Parameters

onnx_file (Union[str, os.PathLike]) – Trained neural network saved as an onnx file

Returns

Return type

None