In [1]:
from sklearn_benchmarks.reporting.hp_match import HpMatchReporting
from sklearn_benchmarks.utils import default_results_directory
from pathlib import Path
import pandas as pd
pd.set_option('display.max_colwidth', None)
pd.set_option('display.max_columns', None)
pd.set_option('display.max_rows', None)

ONNX Runtime vs. scikit-learn¶

In [2]:
results_dir = default_results_directory()
In [3]:
# Parameters
results_dir = "./results/local/20220314T110428/"
In [4]:
results_dir = Path(results_dir)
In [5]:
reporting = HpMatchReporting(other_library="onnx", config="config.yml", log_scale=True, results_dir=results_dir)
reporting.make_report()

We assume here there is a perfect match between the hyperparameters of both librairies. For a given set of parameters and a given dataset, we compute the speed-up time scikit-learn / time onnx. For instance, a speed-up of 2 means that onnx is twice as fast as scikit-learn for a given set of parameters and a given dataset.

¶

onnx (1.11.0) vs. scikit-learn (1.0.2)

Speedup barplots ¶

All estimators share the following parameters: algorithm=brute.

Raw results ¶

predict

function n_samples_train n_samples n_features mean_duration_sklearn std_duration_sklearn iteration_throughput latency n_jobs n_neighbors accuracy_score_sklearn mean_duration_onnx std_duration_onnx accuracy_score_onnx speedup std_speedup sklearn_profiling onnx_profiling
0 predict 100000 1000 100 2.589 0.291 0.000 0.003 -1 1 0.676 17.875 0.043 0.676 0.145 0.145 Download Download
1 predict 100000 1 100 0.026 0.003 0.000 0.026 -1 1 0.000 0.334 0.003 0.000 0.076 0.076 Download Download
2 predict 100000 1000 100 3.098 0.034 0.000 0.003 -1 5 0.743 17.873 0.019 0.743 0.173 0.173 Download Download
3 predict 100000 1 100 0.026 0.002 0.000 0.026 -1 5 1.000 0.335 0.004 1.000 0.076 0.076 Download Download
4 predict 100000 1000 100 2.160 0.006 0.000 0.002 1 100 0.846 17.902 0.045 0.846 0.121 0.121 Download Download
5 predict 100000 1 100 0.022 0.000 0.000 0.022 1 100 1.000 0.336 0.004 1.000 0.067 0.067 Download Download
6 predict 100000 1000 100 2.976 0.066 0.000 0.003 -1 100 0.846 17.949 0.094 0.846 0.166 0.166 Download Download
7 predict 100000 1 100 0.026 0.001 0.000 0.026 -1 100 1.000 0.338 0.002 1.000 0.076 0.076 Download Download
8 predict 100000 1000 100 2.166 0.013 0.000 0.002 1 5 0.743 17.974 0.051 0.743 0.120 0.120 Download Download
9 predict 100000 1 100 0.022 0.000 0.000 0.022 1 5 1.000 0.334 0.005 1.000 0.067 0.067 Download Download
10 predict 100000 1000 100 1.297 0.007 0.001 0.001 1 1 0.676 17.935 0.035 0.676 0.072 0.072 Download Download
11 predict 100000 1 100 0.025 0.006 0.000 0.025 1 1 0.000 0.336 0.004 0.000 0.074 0.074 Download Download
12 predict 100000 1000 2 1.846 0.023 0.000 0.002 -1 1 0.845 4.291 0.007 0.845 0.430 0.430 Download Download
13 predict 100000 1 2 0.005 0.001 0.000 0.005 -1 1 1.000 0.259 0.005 1.000 0.018 0.018 Download Download
14 predict 100000 1000 2 2.791 0.065 0.000 0.003 -1 5 0.883 4.240 0.012 0.883 0.658 0.658 Download Download
15 predict 100000 1 2 0.007 0.003 0.000 0.007 -1 5 1.000 0.256 0.002 1.000 0.026 0.026 Download Download
16 predict 100000 1000 2 2.113 0.009 0.000 0.002 1 100 0.887 4.306 0.012 0.887 0.491 0.491 Download Download
17 predict 100000 1 2 0.003 0.000 0.000 0.003 1 100 1.000 0.256 0.003 1.000 0.012 0.012 Download Download
18 predict 100000 1000 2 2.798 0.037 0.000 0.003 -1 100 0.887 4.306 0.006 0.887 0.650 0.650 Download Download
19 predict 100000 1 2 0.007 0.004 0.000 0.007 -1 100 1.000 0.256 0.004 1.000 0.028 0.028 Download Download
20 predict 100000 1000 2 2.112 0.023 0.000 0.002 1 5 0.883 4.259 0.015 0.883 0.496 0.496 Download Download
21 predict 100000 1 2 0.003 0.000 0.000 0.003 1 5 1.000 0.253 0.003 1.000 0.013 0.013 Download Download
22 predict 100000 1000 2 1.199 0.016 0.000 0.001 1 1 0.845 4.326 0.017 0.845 0.277 0.277 Download Download
23 predict 100000 1 2 0.003 0.001 0.000 0.003 1 1 1.000 0.254 0.002 1.000 0.010 0.010 Download Download

¶

onnx (1.11.0) vs. scikit-learn (1.0.2)

Speedup barplots ¶

All estimators share the following parameters: learning_rate=0.01, n_iter_no_change=10.0, max_leaf_nodes=100.0, max_bins=255.0, min_samples_leaf=100.0, max_iter=300.0.

Raw results ¶

predict

function n_samples_train n_samples n_features mean_duration_sklearn std_duration_sklearn iteration_throughput latency accuracy_score_sklearn mean_duration_onnx std_duration_onnx accuracy_score_onnx speedup std_speedup sklearn_profiling onnx_profiling
0 predict 100000 1000 100 0.137 0.005 0.006 0.0 0.795 0.586 0.006 0.795 0.233 0.233 Download Download

Benchmark environment information¶

System¶
python 3.8.12 | packaged by conda-forge | (default, Jan 30 2022, 23:42:07) [GCC 9.4.0]
executable /usr/share/miniconda/envs/sklbench/bin/python
machine Linux-5.11.0-1028-azure-x86_64-with-glibc2.10

Dependencies¶

version
pip 22.0.4
setuptools 60.9.3
sklearn 1.0.2
numpy 1.22.3
scipy 1.8.0
Cython None
pandas 1.4.1
matplotlib 3.5.1
joblib 1.1.0
threadpoolctl 3.1.0

Threadpool¶

user_api internal_api prefix filepath version threading_layer architecture num_threads
0 blas openblas libopenblas /usr/share/miniconda/envs/sklbench/lib/libopenblasp-r0.3.18.so 0.3.18 pthreads SkylakeX 2
1 openmp openmp libgomp /usr/share/miniconda/envs/sklbench/lib/libgomp.so.1.0.0 None NaN NaN 2

CPU count¶
cpu_count 2
physical_cpu_count 2