module `external.onnxruntime_perf`¶

Short summary¶

module pymlbenchmark.external.onnxruntime_perf

Implements a benchmark about performance for onnxruntime

Classes¶

class	truncated documentation
`OnnxRuntimeBenchPerfTest`	Specific test to compare computing time predictions with scikit-learn and onnxruntime. See example …

Methods¶

method	truncated documentation
`__init__`
`_get_random_dataset`	Returns a random datasets.
`_init`	Finalizes the init.
`data`	Generates random features.
`extract_model_info_onnx`	Populates member `self.onnx_info` with additional information on the ONNX graph.
`extract_model_info_skl`	Populates member `self.skl_info` with additional information on the model such as the number of node for …
`model_info`	Returns additional informations about a model.
`validate`	Checks that methods predict and predict_proba returns the same results for both scikit-learn and …

Documentation¶

Implements a benchmark about performance for onnxruntime

source on GitHub

class pymlbenchmark.external.onnxruntime_perf.OnnxRuntimeBenchPerfTest(estimator, dim=None, N_fit=100000, runtimes=('python_compiled', 'onnxruntime1'), onnx_options=None, dtype=<class 'numpy.float32'>, **opts)¶

Bases: BenchPerfTest

Specific test to compare computing time predictions with scikit-learn and onnxruntime. See example Benchmark of onnxruntime on LogisticRegression. The class requires the following modules to be installed: onnx, onnxruntime, skl2onnx, mlprodict.

source on GitHub

Parameters:

estimator – estimator class
dim – number of features
N_fit – number of observations to fit an estimator
runtimes – runtimes to test for class OnnxInference
opts – training settings
onnx_options – ONNX conversion options
dtype – dtype (float32 or float64)

source on GitHub

__init__(estimator, dim=None, N_fit=100000, runtimes=('python_compiled', 'onnxruntime1'), onnx_options=None, dtype=<class 'numpy.float32'>, **opts)¶