module external.onnxruntime_perf_list
¶
Short summary¶
module pymlbenchmark.external.onnxruntime_perf_list
Returns predefined tests.
Functions¶
function |
truncated documentation |
---|---|
Returns a list of benchmarks for binary classifier. It compares onnxruntime predictions against scikit-learn. … |
|
Returns a list of benchmarks for binary classifier. It compares onnxruntime predictions against scikit-learn. … |
|
Runs a benchmark for onnxruntime. |
Documentation¶
Returns predefined tests.
- pymlbenchmark.external.onnxruntime_perf_list.onnxruntime_perf_binary_classifiers(bincl=None, N_fit=100000)¶
Returns a list of benchmarks for binary classifier. It compares onnxruntime predictions against scikit-learn.
- Parameters:
bincl – test class to use, by default, it is
OnnxRuntimeBenchPerfTestBinaryClassification
N_fit – number of rows needed to train a model
- pymlbenchmark.external.onnxruntime_perf_list.onnxruntime_perf_regressors(regcl=None, N_fit=100000)¶
Returns a list of benchmarks for binary classifier. It compares onnxruntime predictions against scikit-learn.
- Parameters:
regcl – test class to use, by default, it is
OnnxRuntimeBenchPerfTestRegression
N_fit – number of rows needed to train a model
- pymlbenchmark.external.onnxruntime_perf_list.run_onnxruntime_test(folder, name, repeat=100, verbose=True, stop_if_error=True, validate=True, N=None, dim=None, N_fit=100000, fLOG=None, kwbefore=None)¶
Runs a benchmark for onnxruntime.
- Parameters:
folder – where to dump the results
name – name of the test (one in the list returned by
onnxruntime_perf_binary_classifiers
)repeat – number of times to repeat predictions
verbose – print progress with tqdm
stop_if_error – by default, it stops when method validate fails, if False, the function stores the exception
validate – validate the outputs against the baseline
N – overwrites N parameter
dim – overwrites dims parameter
N_fit – number of rows needed to train a model
kwbefore – additional arguments before training
fLOG – logging function
- Returns:
two dataframes, one for the results, the other one for the context (see
machine_information
)