neurtu

Simple performance measurement tool

neurtu is a Python package providing a common interface for multi-metric benchmarks (including time and memory measurements). It can can be used to estimate time and space complexity of algorithms, while pandas integration allows quick analysis and visualization of the results.

Setting the number of threads at runtime in OpenBlas, and MKL is also supported on Linux and MacOS.

neurtu means “to measure / evaluate” in Basque language.

Installation

neurtu requires Python 2.7 or 3.4+, it can be installed with,

pip install neurtu

pandas is an optional (but highly recommended) dependency.

Note

the above command will install memory_profiler, shutil (to measure memory use) and tqdm (to make progress bars) mostly for convinience. However, neurtu does not have any hard depedencies, it you don’t need these functionalites, you can install it with pip install --no-deps neurtu

Quickstart

To illustrate neurtu usage, will will benchmark array sorting in numpy. First, we will generator of cases,

import numpy as np
import neurtu

def cases()
    rng = np.random.RandomState(42)

    for N in [1000, 10000, 100000]:
        X = rng.rand(N)
        tags = {'N' : N}
        yield neurtu.delayed(X, tags=tags).sort()

that yields a sequence of delayed calculations, each tagged with the parameters defining individual runs.

We can evaluate the run time with,

>>> df = neurtu.timeit(cases())
>>> print(df)
        wall_time
N
1000     0.000014
10000    0.000134
100000   0.001474

which will internally use timeit module with a sufficient number of evaluation to work around the timer precision limitations (similarly to IPython’s %timeit). It will also display a progress bar for long running benchmarks, and return the results as a pandas.DataFrame (if pandas is installed).

By default, all evaluations are run with repeat=1. If more statistical confidence is required, this value can be increased,

>>> neurtu.timeit(cases(), repeat=3)
       wall_time
            mean       max       std
N
1000    0.000012  0.000014  0.000002
10000   0.000116  0.000149  0.000029
100000  0.001323  0.001714  0.000339

In this case we will get a frame with a pandas.MultiIndex for columns, where the first level represents the metric name (wall_time) and the second – the aggregation method. By default neurtu.timeit is called with aggregate=['mean', 'max', 'std'] methods, as supported by the pandas aggregation API. To disable, aggregation and obtains timings for individual runs, use aggregate=False. See neurtu.timeit documentation for more details.

To evaluate the peak memory usage, one can use the neurtu.memit function with the same API,

>>> neurtu.memit(cases(), repeat=3)
        peak_memory
               mean  max  std
N
10000           0.0  0.0  0.0
100000          0.0  0.0  0.0
1000000         0.0  0.0  0.0

More generally neurtu.Benchmark supports a wide number of evaluation metrics,

>>> bench = neurtu.Benchmark(wall_time=True, cpu_time=True, peak_memory=True)
>>> bench(cases())
         cpu_time  peak_memory  wall_time
N
10000    0.000100          0.0   0.000142
100000   0.001149          0.0   0.001680
1000000  0.013677          0.0   0.018347

including [psutil process metrics](https://psutil.readthedocs.io/en/latest/#psutil.Process).

For more information see the Examples.

Examples

The following examples illustrate neurtu usage

Time complexity of numpy.sort

Time complexity of numpy.sort

In this example we will look into the time complexity of numpy.sort()

import numpy as np
from neurtu import timeit, delayed


rng = np.random.RandomState(42)


df = timeit(delayed(np.sort, tags={'N': N, 'kind': kind})(rng.rand(N), kind=kind)
            for N in np.logspace(2, 5, num=5).astype('int')
            for kind in ["quicksort", "mergesort", "heapsort"])

print(df.to_string())

Out:

                  wall_time
N      kind
100    quicksort   0.000005
       mergesort   0.000006
       heapsort    0.000006
562    quicksort   0.000011
       mergesort   0.000017
       heapsort    0.000037
3162   quicksort   0.000170
       mergesort   0.000199
       heapsort    0.000296
17782  quicksort   0.001223
       mergesort   0.001389
       heapsort    0.002012
100000 quicksort   0.007667
       mergesort   0.009005
       heapsort    0.014728

we can use the pandas plotting API (that requires matplotlib)

ax = df.wall_time.unstack().plot(marker='o')
ax.set_xscale('log')
ax.set_yscale('log')
ax.set_ylabel('Wall time (s)')
ax.set_title('Time complexity of numpy.sort')

LogisticRegression scaling in scikit-learn

LogisticRegression scaling in scikit-learn

In this example we will look into the time and space complexity of sklearn.linear_model.LogisticRegression

from collections import OrderedDict

import numpy as np
from sklearn.linear_model import LogisticRegression
from neurtu import Benchmark, delayed


rng = np.random.RandomState(42)

n_samples, n_features = 50000, 100


X = rng.rand(n_samples, n_features)
y = rng.randint(2, size=(n_samples))


def benchmark_cases():
    for N in np.logspace(np.log10(100), np.log10(n_samples), 5).astype('int'):
        for solver in ['newton-cg', 'lbfgs', 'liblinear', 'sag', 'saga']:
            tags = OrderedDict(N=N, solver=solver)
            model = delayed(LogisticRegression, tags=tags)(
                                solver=solver, random_state=rng)

            yield model.fit(X[:N], y[:N])


bench = Benchmark(wall_time=True, peak_memory=True)
df = bench(benchmark_cases())

print(df.tail())

Out:

  0%|          | 0/50 [00:00<?, ?it/s]
  4%|4         | 2/50 [00:00<00:04, 11.74it/s]
  6%|6         | 3/50 [00:00<00:07,  6.39it/s]
  8%|8         | 4/50 [00:00<00:08,  5.56it/s]
 10%|#         | 5/50 [00:00<00:08,  5.11it/s]/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)

 14%|#4        | 7/50 [00:01<00:08,  5.08it/s]
 16%|#6        | 8/50 [00:01<00:07,  5.84it/s]/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)

 18%|#8        | 9/50 [00:01<00:08,  4.85it/s]/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)

 20%|##        | 10/50 [00:01<00:07,  5.54it/s]
 22%|##2       | 11/50 [00:02<00:08,  4.77it/s]
 24%|##4       | 12/50 [00:02<00:06,  5.54it/s]
 26%|##6       | 13/50 [00:02<00:07,  4.97it/s]
 28%|##8       | 14/50 [00:02<00:06,  5.64it/s]
 30%|###       | 15/50 [00:02<00:07,  4.95it/s]/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)

 34%|###4      | 17/50 [00:03<00:06,  5.06it/s]/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)

 36%|###6      | 18/50 [00:03<00:05,  5.39it/s]/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)

 38%|###8      | 19/50 [00:03<00:06,  4.67it/s]/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)
/home/docs/checkouts/readthedocs.org/user_builds/neurtu/envs/stable/lib/python3.7/site-packages/sklearn/linear_model/sag.py:337: ConvergenceWarning: The max_iter was reached which means the coef_ did not converge
  "the coef_ did not converge", ConvergenceWarning)

 40%|####      | 20/50 [00:03<00:06,  4.94it/s]
 42%|####2     | 21/50 [00:04<00:06,  4.21it/s]
 44%|####4     | 22/50 [00:04<00:05,  4.70it/s]
 46%|####6     | 23/50 [00:04<00:05,  4.63it/s]
 48%|####8     | 24/50 [00:04<00:05,  5.03it/s]
 50%|#####     | 25/50 [00:05<00:05,  4.60it/s]
 52%|#####2    | 26/50 [00:05<00:04,  5.16it/s]
 54%|#####4    | 27/50 [00:05<00:05,  4.02it/s]
 56%|#####6    | 28/50 [00:05<00:05,  3.93it/s]
 58%|#####8    | 29/50 [00:06<00:06,  3.48it/s]
 60%|######    | 30/50 [00:06<00:05,  3.38it/s]
 62%|######2   | 31/50 [00:06<00:05,  3.55it/s]
 64%|######4   | 32/50 [00:07<00:06,  2.76it/s]
 66%|######6   | 33/50 [00:07<00:05,  2.88it/s]
 68%|######8   | 34/50 [00:07<00:05,  3.05it/s]
 70%|#######   | 35/50 [00:08<00:04,  3.02it/s]
 72%|#######2  | 36/50 [00:08<00:04,  3.41it/s]
 74%|#######4  | 37/50 [00:08<00:04,  2.66it/s]
 76%|#######6  | 38/50 [00:09<00:05,  2.35it/s]
 78%|#######8  | 39/50 [00:09<00:04,  2.46it/s]
 80%|########  | 40/50 [00:10<00:04,  2.47it/s]
 82%|########2 | 41/50 [00:12<00:08,  1.03it/s]
 84%|########4 | 42/50 [00:14<00:10,  1.34s/it]
 86%|########6 | 43/50 [00:15<00:07,  1.07s/it]
 88%|########8 | 44/50 [00:15<00:04,  1.21it/s]
 90%|######### | 45/50 [00:16<00:03,  1.29it/s]
 92%|#########2| 46/50 [00:16<00:02,  1.36it/s]
 94%|#########3| 47/50 [00:19<00:04,  1.38s/it]
 96%|#########6| 48/50 [00:22<00:03,  1.90s/it]
 98%|#########8| 49/50 [00:24<00:01,  1.92s/it]
100%|##########| 50/50 [00:26<00:00,  1.89s/it]

                 wall_time  peak_memory
N     solver
49999 newton-cg   2.245351    75.835938
      lbfgs       0.189091     0.003906
      liblinear   0.596465    79.875000
      sag         2.844747     0.007812
      saga        1.905433     0.000000

The above section will run in approximately 1min, a progress bar will be displayed.

We can use the pandas plotting API (that requires matplotlib) to visualize the results,

ax = df.wall_time.unstack().plot(marker='o')
ax.set_xscale('log')
ax.set_yscale('log')
ax.set_ylabel('Wall time (s)')
ax.set_title('Run time scaling for LogisticRegression.fit')

The solver with the best scalability in this example is “lbfgs”.

Similarly the memory scaling is represented below,

ax = df.peak_memory.unstack().plot(marker='o')
ax.set_xscale('log')
ax.set_yscale('log')
ax.set_ylabel('Peak memory (MB)')
ax.set_title('Peak memory usage for LogisticRegression.fit')

Peak memory usage for “liblinear” and “newton-cg” appear to be significant above 10000 samples, while the other solvers use less memory than the detection threshold. Note that these benchmarks do not account for the memory used by X and y arrays.

API Reference

neurtu.timeit(obj[, timer, number, repeat, …])	A benchmark decorator
neurtu.memit(obj[, repeat, aggregate, …])	Measure the memory use.
neurtu.Benchmark([wall_time, cpu_time, …])	Benchmark calculations
neurtu.delayed(obj[, tags, env])	Delayed object evaluation

Release notes

Version 0.3

July 21, 2019

API changes

• Functions to set the number of BLAS threads at runtime were removed in favour of using threadpoolctl.

Enhancements

• Add get_args and get_kwargs to Delayed object.
• Better progress bars in Jupyter notebooks with the tqdm.auto backend.

Bug fixes

• Fix progress bar rendering when repeat>1.
• Fix warnings due to collection.abc.

Version 0.2

August 28, 2018

New features

• Runtime detection of the BLAS used by numpy #14
• Ability to set the number of threads in OpenBlas and MKL BLAS at runtime on Linux. #15.

Enhancements

• Better test coverage
• Documentation improvements
• In depth refactoring of the benchmarking code

API changes

• The API of timeit, memit, Benchmark changed significantly with respect to v0.1

Version 0.1

March 4, 2018

First release, with support for,

• wall time, cpu time and peak memory measurements
• parametric benchmarks using delayed objects