No project description provided
Project description
Scikit-learn Smithy
CLI to forge scikit-learn compatible estimator templates with ease.
Why
Writing a scikit-learn compatible estimators might be harder than expected.
While everyone knows about the fit
and predict
, there are other behaviours, methods and attributes that scikit-learn might be expecting from your estimator. These depend on:
- The type of estimator you're writing.
- The signature of the estimator.
- The signature of the
.fit(...)
method.
This tool aims to help you with that by asking you a few questions about your estimator, and then generating the boilerplate code for you, so that you can focus on the core implementation of the estimator, and not on the nitty-gritty details of the scikit-learn API.
Once the core logic is implemented, the estimator should be ready to test against the somewhat official parametrize_with_checks
pytest compatible decorator:
from sklearn.utils.estimator_checks import parametrize_with_checks
@parametrize_with_checks([YourAwesomeRegressor, MoreAwesomeClassifier, EvenMoreAwesomeTransformer])
def test_sklearn_compatible_estimator(estimator, check):
check(estimator)
Installation
As of today, this is work in progress ๐ง and not published to pypi, but it is possible to install it directly from github
python -m pip install git+https://github.com/FBruzzesi/sklearn-smithy.git
This will make the smith
command available in your terminal.
Commands
The smith
entrypoint offers two commands:
smith --help
Usage: smith [OPTIONS] COMMAND [ARGS]...
Awesome CLI to generate scikit-learn estimator boilerplate code
...
โญโ Commands โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฎ
โ forge Asks a list of questions to generate a shiny new estimator โจ โ
โ version Display library version. โ
โฐโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฏ
and as you can already guess, the forge
command is the one that will generate the boilerplate code for you.
smith forge --help
Asks a list of questions to generate a shiny new estimator โจ
Depending on the **estimator type** the additional information could be required:
* if the estimator is linear (classifier or regression)
* if the estimator has a `predict_proba` method (classifier or outlier detector)
* is the estimator has a `decision_function` method (classifier only)
Finally, the following two questions will be prompt:
* if the estimator should have tags (To know more about tags, check the dedicated
[scikit-learn documentation](https://scikit-learn.org/dev/developers/develop.html#estimator-tags))
* in which file the class should be saved (default is `f'{name.lower()}.py'`)
โญโ Options โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฎ
โ * --name TEXT name the estimator. [default: None] [required] โ
โ * --estimator-type [classifier|outlier|regressor|transformer] Estimator type. [default: None] [required] โ
โ --required-params TEXT List of required parameters (comma-separated). โ
โ --other-params TEXT List of optional parameters (comma-separated). โ
โ --support-sample-weight --no-support-sample-weight Whether or not `.fit()` does support `sample_weight`. [default: no-support-sample-weight] โ
โ --help Show this message and exit. โ
โฐโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฏ
Origin story
The idea for this tool originated from scikit-lego #660:
So the story goes as the following:
- The CI/CD fails for scikit-learn==1.5rc1 because of a change in the
check_estimator
internals- In the scikit-learn issue I got a better picture of how to run test for compatible components
- In particular, in rolling your own estimator suggests to use
parametrize_with_checks
, and of course I thought "that is a great idea to avoid dealing manually with each test"- Say no more, I enter a rabbit hole to refactor all our tests - which would be fine
- Except that these tests failures helped me figure out a few missing parts in the codebase
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for sklearn_smithy-0.0.2-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 2e6b0af2adcb9ee9f80f2eb27a754f5b79fcb05f831926b2a8c99c0a8f7cb0ac |
|
MD5 | 6863142e92afa312e7d5f368b1f83b9f |
|
BLAKE2b-256 | 2f5c995c731d68c25a5a6e5d3999a63d732b1aef262c7815999cfaaae8945c60 |