- Notifications
You must be signed in to change notification settings - Fork96
LAMA - automatic model creation framework
License
sberbank-ai-lab/LightAutoML
Folders and files
| Name | Name | Last commit message | Last commit date | |
|---|---|---|---|---|
Repository files navigation
LightAutoML (LAMA) is an AutoML framework by Sber AI Lab.
It provides automatic model creation for the following tasks:
- binary classification
- multiclass classification
- regression
Current version of the package handles datasets that have independent samples in each row. I.e.each row is an object with its specific features and target.Multitable datasets and sequences are a work in progress :)
Note: we useAutoWoE library to automatically create interpretable models.
Authors:Alexander Ryzhkov,Anton Vakhrushev,Dmitry Simakov, Vasilii Bunakov, Rinchin Damdinov, Pavel Shvets, Alexander Kirilin.
Documentation of LightAutoML is availablehere, you can alsogenerate it.
Full GPU pipeline for LightAutoML currently available for developers testing (still in progress). The code and tutorialsavailable here
- Installation LightAutoML from PyPI
- Quick tour
- Resources
- Contributing to LightAutoML
- License
- For developers
- Support and feature requests
To install LAMA framework on your machine from PyPI, execute following commands:
# Install base functionality:pip install -U lightautoml# For partial installation use corresponding option.# Extra dependecies: [nlp, cv, report]# Or you can use 'all' to install everythingpip install -U lightautoml[nlp]
Additionaly, run following commands to enable pdf report generation:
# MacOSbrew install cairo pango gdk-pixbuf libffi# Debian / Ubuntusudo apt-get install build-essential libcairo2 libpango-1.0-0 libpangocairo-1.0-0 libgdk-pixbuf2.0-0 libffi-dev shared-mime-info# Fedorasudo yum install redhat-rpm-config libffi-devel cairo pango gdk-pixbuf2# Windows# follow this tutorial https://weasyprint.readthedocs.io/en/stable/install.html#windows
Let's solve the popular Kaggle Titanic competition below. There are two main ways to solve machine learning problems using LightAutoML:
- Use ready preset for tabular data:
importpandasaspdfromsklearn.metricsimportf1_scorefromlightautoml.automl.presets.tabular_presetsimportTabularAutoMLfromlightautoml.tasksimportTaskdf_train=pd.read_csv('../input/titanic/train.csv')df_test=pd.read_csv('../input/titanic/test.csv')automl=TabularAutoML(task=Task(name='binary',metric=lambday_true,y_pred:f1_score(y_true, (y_pred>0.5)*1)))oof_pred=automl.fit_predict(df_train,roles= {'target':'Survived','drop': ['PassengerId']})test_pred=automl.predict(df_test)pd.DataFrame({'PassengerId':df_test.PassengerId,'Survived': (test_pred.data[:,0]>0.5)*1}).to_csv('submit.csv',index=False)
LighAutoML framework has a lot of ready-to-use parts and extensive customization options, to learn more check out theresources section.
- Tabular Playground Series April 2021 competition solution
- Titanic competition solution (80% accuracy)
- Titanic12-code-lines competition solution (78% accuracy)
- House prices competition solution
- Natural Language Processing with Disaster Tweets solution
- Tabular Playground Series March 2021 competition solution
- Tabular Playground Series February 2021 competition solution
- Interpretable WhiteBox solution
- Custom ML pipeline elements inside existing ones
Google Colab tutorials andother examples:
Tutorial_1_basics.ipynb- get started with LightAutoML on tabular data.
Tutorial_2_WhiteBox_AutoWoE.ipynbTutorial_3_sql_data_source.ipynbTutorial_4_NLP_Interpretation.ipynbTutorial_5_uplift.ipynbTutorial_6_custom_pipeline.ipynbTutorial_7_ICE_and_PDP_interpretation.ipynb
Note 1: for production you have no need to use profiler (which increase work time and memory consomption), so please do not turn it on - it is in off state by default
Note 2: to take a look at this report after the run, please comment last line of demo with report deletion command.
LightAutoML crash courses:
Video guides:
- (Russian)LightAutoML webinar for Sberloga community (Alexander Ryzhkov,Dmitry Simakov)
- (Russian)LightAutoML hands-on tutorial in Kaggle Kernels (Alexander Ryzhkov)
- (English)Automated Machine Learning with LightAutoML: theory and practice (Alexander Ryzhkov)
- (English)LightAutoML framework general overview, benchmarks and advantages for business (Alexander Ryzhkov)
- (English)LightAutoML practical guide - ML pipeline presets overview (Dmitry Simakov)
Papers:
- Anton Vakhrushev, Alexander Ryzhkov, Dmitry Simakov, Rinchin Damdinov, Maxim Savchenko, Alexander Tuzhilin"LightAutoML: AutoML Solution for a Large Financial Services Ecosystem". arXiv:2109.01528, 2021.
Articles about LightAutoML:
If you are interested in contributing to LightAutoML, please read theContributing Guide to get started.
This project is licensed under the Apache License, Version 2.0. SeeLICENSE file for more details.
First of all you need to installgit andpoetry.
# Load LAMA source codegit clone https://github.com/sberbank-ai-lab/LightAutoML.gitcd LightAutoML/# !!!Choose only one item!!!# 1. Global installation: Don't create virtual environmentpoetry config virtualenvs.createfalse --local# 2. Recommended: Create virtual environment inside your project directorypoetry config virtualenvs.in-projecttrue# For more information read poetry docs# Install LAMApoetry lockpoetry install
importpandasaspdfromsklearn.metricsimportf1_scorefromlightautoml.automl.presets.tabular_presetsimportTabularAutoMLfromlightautoml.tasksimportTaskdf_train=pd.read_csv('../input/titanic/train.csv')df_test=pd.read_csv('../input/titanic/test.csv')# define that machine learning problem is binary classificationtask=Task("binary")reader=PandasToPandasReader(task,cv=N_FOLDS,random_state=RANDOM_STATE)# create a feature selectormodel0=BoostLGBM(default_params={'learning_rate':0.05,'num_leaves':64,'seed':42,'num_threads':N_THREADS})pipe0=LGBSimpleFeatures()mbie=ModelBasedImportanceEstimator()selector=ImportanceCutoffSelector(pipe0,model0,mbie,cutoff=0)# build first level pipeline for AutoMLpipe=LGBSimpleFeatures()# stop after 20 iterations or after 30 secondsparams_tuner1=OptunaTuner(n_trials=20,timeout=30)model1=BoostLGBM(default_params={'learning_rate':0.05,'num_leaves':128,'seed':1,'num_threads':N_THREADS})model2=BoostLGBM(default_params={'learning_rate':0.025,'num_leaves':64,'seed':2,'num_threads':N_THREADS})pipeline_lvl1=MLPipeline([ (model1,params_tuner1),model2],pre_selection=selector,features_pipeline=pipe,post_selection=None)# build second level pipeline for AutoMLpipe1=LGBSimpleFeatures()model=BoostLGBM(default_params={'learning_rate':0.05,'num_leaves':64,'max_bin':1024,'seed':3,'num_threads':N_THREADS},freeze_defaults=True)pipeline_lvl2=MLPipeline([model],pre_selection=None,features_pipeline=pipe1,post_selection=None)# build AutoML pipelineautoml=AutoML(reader, [ [pipeline_lvl1], [pipeline_lvl2],],skip_conn=False)# train AutoML and get predictionsoof_pred=automl.fit_predict(df_train,roles= {'target':'Survived','drop': ['PassengerId']})test_pred=automl.predict(df_test)pd.DataFrame({'PassengerId':df_test.PassengerId,'Survived': (test_pred.data[:,0]>0.5)*1}).to_csv('submit.csv',index=False)
Seek prompt advice atSlack community orTelegram group.
Open bug reports and feature requests on GitHubissues.
About
LAMA - automatic model creation framework