stablerandom

stablerandom provides a stable and repeatable implementation of the NumPy random number generator (numpy.random). With this package, you can generate the same sequence of random numbers across different platforms and Python environments, ensuring reproducibility in scientific computing, machine learning, and unit testing.

stablerandom can decorate any function or method and provide a call-stack scoped seeded random generator. stablerandom can also be used as a Context. It is thread safe and supports nested scopes.

Example decorator

Using the @stablerandom decorator to get a stable output for numpy.random.triangular

import numpy.random
from stablerandom import stablerandom

@stablerandom
def random_triangular(samples):
   return numpy.random.triangular(1, 5, 10, samples)

print(random_triangular(3))
>>> [5.1017943  2.64970963 8.6996913 ]
print(random_triangular(3))
>>> [5.1017943  2.64970963 8.6996913 ]

Example context

Using the with stablerandom context to get a stable output:

import numpy.random
from stablerandom import stablerandom

with stablerandom:
    print(numpy.random.triangular(1, 5, 10, 3))
    print(numpy.random.triangular(1, 5, 10, 3))
>>> [5.1017943  2.64970963 8.6996913 ]
>>> [3.30538708 2.39427919 3.57653224]

with stablerandom:
    print(numpy.random.triangular(1, 5, 10, 3))
>>> [5.1017943  2.64970963 8.6996913 ]

Installing

$ pip install stablerandom

The source code is currently hosted on GitHub at https://github.com/GistLabs/stablerandom and published in PyPI at https://pypi.org/project/stablerandom/

The versioning scheme currently used is {major}.{minor}.{auto build number} from git rev-list --count HEAD.

We recommend picking a version like:

stablerandom = "^0.5"

Dependencies

This library has been tested with NumPy back to version 1.22

Community guidelines

We welcome contributions and questions. Please head over to github and send us pull requests or create issues!