experimental_experiment.model_run

experimental_experiment.model_run.create_feeds(sess: onnxruntime.InferenceSession, batch_size: int = 1) → Dict[str, Any]

Creates random feeds for a model.

Parameters:

• sess – onnxruntime session

• batch_size – batch_size

Returns:

feeds

experimental_experiment.model_run.create_tensor(shape: Tuple[int, ...], dtype: int, batch_size: int = 1) → ndarray

Creates a random tensor.

Parameters:

• shape – shape

• dtype – onnx type

• batch_size – batch_size

Returns:

numpy array

experimental_experiment.model_run.model_run(model: str | ModelProto, repeat: int = 10, warmup: int = 5, batch_size: int = 1, processor: str = 'CPU', verbose: int = 0, validate: str | ModelProto | None = None) → Dict[str, Any]

Loads a model with onnxruntime and measures the inference time.

Parameters:

• model – model to run

• warmup – number of iterations to run before measuring

• repeat – number of iterations to run to measure

• batch_size – batch size of the inputs

• processor – processor to run

• verbose – verbosity

Returns:

metrics