Lightwood

Lightwood is an AutoML framework that enables you to generate and customize machine learning pipelines declarative syntax called JSON-AI.

Our goal is to make the data science/machine learning (DS/ML) life cycle easier by allowing users to focus on what they want to do their data without needing to write repetitive boilerplate code around machine learning and data preparation. Instead, we enable you to focus on the parts of a model that are truly unique and custom.

Lightwood works with a variety of data types such as numbers, dates, categories, tags, text, arrays and various multimedia formats. These data types can be combined together to solve complex problems. We also support a time-series mode for problems that have between-row dependencies.

Our JSON-AI syntax allows users to change any and all parts of the models Lightwood automatically generates. The syntax outlines the specifics details in each step of the modeling pipeline. Users may override default values (for example, changing the type of a column) or alternatively, entirely replace steps with their own methods (ex: use a random forest model for a predictor). Lightwood creates a "JSON-AI" object from this syntax which can then be used to automatically generate python code to represent your pipeline.

For details on how to generate JSON-AI syntax and how Lightwood works, check out the Lightwood Philosophy.

Lightwood Philosophy

Lightwood abstracts the ML pipeline into 3 core steps:

(1) Pre-processing and data cleaning
(2) Feature engineering
(3) Model building and training

i) Pre-processing and cleaning

For each column in your dataset, Lightwood will identify the suspected data type (numeric, categorical, etc.) via a brief statistical analysis. From this, it will generate a JSON-AI syntax.

If the user keeps default behavior, Lightwood will perform a brief pre-processing approach to clean each column according to its identified data type. From there, it will split the data into train/dev/test splits.

The cleaner and splitter objects respectively refer to the pre-processing and the data splitting functions.

ii) Feature Engineering

Data can be converted into features via "encoders". Encoders represent the rules for transforming pre-processed data into a numerical representations that a model can be used.

Encoders can be rule-based or learned. A rule-based encoder transforms data per a specific set of instructions (ex: normalized numerical data) whereas a learned encoder produces a representation of the data after training (ex: a "[CLS]" token in a language model).

Encoders are assigned to each column of data based on the data type; users can override this assignment either at the column-based level or at the data-type based level. Encoders inherit from the BaseEncoder class.

iii) Model Building and Training

We call a predictive model that intakes encoded feature data and outputs a prediction for the target of interest a mixer model. Users can either use Lightwood's default mixers or create their own approaches inherited from the BaseMixer class.

We predominantly use PyTorch based approaches, but can support other models.

Usage

We invite you to check out our documentation for specific guidelines and tutorials! Please stay tuned for updates and changes.

Quick use cases

Lightwood works with pandas.DataFrames. Once a DataFrame is loaded, defined a "ProblemDefinition" via a dictionary. The only thing a user needs to specify is the name of the column to predict (via the key target).

Create a JSON-AI syntax from the command json_ai_from_problem. Lightwood can then use this object to automatically generate python code filling in the steps of the ML pipeline via code_from_json_ai.

You can make a Predictor object, instantiated with that code via predictor_from_code.

To train a Predictor end-to-end, starting with unprocessed data, users can use the predictor.learn() command with the data.

import pandas as pd
from lightwood.api.high_level import (
    ProblemDefinition,
    json_ai_from_problem,
    code_from_json_ai,
    predictor_from_code,
)

if __name__ == '__main__':
    # Load a pandas dataset
    df = pd.read_csv("https://raw.githubusercontent.com/mindsdb/benchmarks/main/benchmarks/datasets/hdi/data.csv"
    )

    # Define the prediction task by naming the target column
    pdef = ProblemDefinition.from_dict(
        {
            "target": "Development Index",  # column you want to predict
        }
    )

    # Generate JSON-AI code to model the problem
    json_ai = json_ai_from_problem(df, problem_definition=pdef)

    # OPTIONAL - see the JSON-AI syntax
    # print(json_ai.to_json())

    # Generate python code
    code = code_from_json_ai(json_ai)

    # OPTIONAL - see generated code
    # print(code)

    # Create a predictor from python code
    predictor = predictor_from_code(code)

    # Train a model end-to-end from raw data to a finalized predictor
    predictor.learn(df)

    # Make the train/test splits and show predictions for a few examples
    test_df = predictor.split(predictor.preprocess(df))["test"]
    preds = predictor.predict(test_df).iloc[:10]
    print(preds)

BYOM: Bring your own models

Lightwood supports user architectures/approaches so long as you follow the abstractions provided within each step.

Our tutorials provide specific use cases for how to introduce customization into your pipeline. Check out "custom cleaner", "custom splitter", "custom explainer", and "custom mixer". Stay tuned for further updates.

Installation

You can install Lightwood as follows:

pip3 install lightwood

Note: depending on your environment, you might have to use pip instead of pip3 in the above command.

However, we recommend creating a python virtual environment.

Setting up a dev environment

• Python version should be in the range >=3.8, < 3.11
• Clone lightwood
• cd lightwood && pip install -r requirements.txt && pip install -r requirements_image.txt
• Add it to your python path (e.g. by adding export PYTHONPATH='/where/you/cloned/lightwood':$PYTHONPATH as a newline at the end of your ~/.bashrc file)
• Check that the unittests are passing by going into the directory where you cloned lightwood and running: python -m unittest discover tests

If python default to python2.x on your environment use python3 and pip3 instead

Currently, the preferred environment for working with lightwood is visual studio code, a very popular python IDE. However, any IDE should work. While we don't have guides for those, please feel free to use the following section as a template for VSCode, or to contribute your own tips and tricks to set up other IDEs.

Setting up a VSCode environment

• Install and enable setting sync using github account (if you use multiple machines)
• Install pylance (for types) and make sure to disable pyright
• Go to Python > Lint: Enabled and disable everything but flake8
• Set python.linting.flake8Path to the full path to flake8 (which flake8)
• Set Python › Formatting: Provider to autopep8
• Add --global-config=<path_to>/lightwood/.flake8 and --experimental to Python › Formatting: Autopep8 Args
• Install live share and live share whiteboard

Contribute to Lightwood

We love to receive contributions from the community and hear your opinions! We want to make contributing to Lightwood as easy as it can be.

Being part of the core Lightwood team is possible to anyone who is motivated and wants to be part of that journey!

Please continue reading this guide if you are interested in helping democratize machine learning.

How can you help us?

• Report a bug
• Improve documentation
• Solve an issue
• Propose new features
• Discuss feature implementations
• Submit a bug fix
• Test Lightwood with your own data and let us know how it went!

Code contributions

In general, we follow the "fork-and-pull" git workflow. Here are the steps:

1. Fork the Lightwood repository
2. Checkout the staging branch, which is the development version that gets released weekly (there can be exceptions, but make sure to ask and confirm with us).
3. Make changes and commit them
4. Make sure that the CI tests pass. You can run the test suite locally with flake8 . to check style and python -m unittest discover tests to run the automated tests. This doesn't guarantee it will pass remotely since we run on multiple envs, but should work in most cases.
5. Push your local branch to your fork
6. Submit a pull request from your repo to the staging branch of mindsdb/lightwood so that we can review your changes. Be sure to merge the latest from staging before making a pull request!

Note: You will need to sign a CLI agreement for the code since lightwood is under a GPL license.

Feature and Bug reports

We use GitHub issues to track bugs and features. Report them by opening a new issue and fill out all of the required inputs.

Code review process

Pull request (PR) reviews are done on a regular basis. If your PR does not address a previous issue, please make an issue first.

If your change has a chance to affecting performance we will run our private benchmark suite to validate it.

Please, make sure you respond to our feedback/questions.

Community

If you have additional questions or you want to chat with MindsDB core team, you can join our community: .

To get updates on Lightwood and MindsDB’s latest announcements, releases, and events, sign up for our Monthly Community Newsletter.

Join our mission of democratizing machine learning and allowing developers to become data scientists!

Contributor Code of Conduct

Please note that this project is released with a Contributor Code of Conduct. By participating in this project, you agree to abide by its terms.

Current contributors

License

• Lightwood License