- Notifications
You must be signed in to change notification settings - Fork0
Archive of the Future Frame's open source Python package with inference code for foundation tabular models.
License
NotificationsYou must be signed in to change notification settings
edadaltocg/futureframe
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
Empowering Data Scientists with Foundation Models for Tabular Data
- This Python package allows you to interact with pre-trained foundation models for tabular data.
- Easily fine-tune them on your classification and regression use cases in a single line of code.
pip install futureframe
Use Future Frame to fine-tune a pre-trained foundation model on a classification task.
# Import standard librariesimportpandasaspdfromsklearn.model_selectionimporttrain_test_splitfromsklearn.metricsimportroc_auc_scoreimportfutureframeasff# Import datadataset_name="https://raw.githubusercontent.com/futureframeai/futureframe/main/tests/data/churn.csv"target_variable="Churn"df=pd.read_csv(dataset_name)X,y=df.drop(columns=[target_variable]),df[target_variable]X_train,X_test,y_train,y_test=train_test_split(X,y,test_size=0.3)# Fine-tune a pre-trained classifier with Future Framemodel=ff.models.cm2.CM2Classifier()model.finetune(X_train,y_train)y_pred=model.predict(X_test)auc=roc_auc_score(y_test,y_pred)print(f"AUC:{auc:0.2f}")
Model Name | Paper Title | Paper | GitHub |
---|---|---|---|
CM2 | Towards Cross-Table Masked Pretraining for Web Data Mining | Ye et al., 2024 | Link |
CARTE (soon!) | CARTE: Pretraining and Transfer for Tabular Learning | Kim et al., 2024 | Link |
TabText (soon!) | TabText: A Flexible and Contextual Approach to Tabular Data Representation | Carballo et al., 2023 | - |
TabPFN (soon!) | TabPFN: A Transformer That Solves Small Tabular Classification Problems in a Second | Hollmann et al., 2022 | Link |
TransTab (soon!) | Transtab: Learning Transferable Tabular Transformers Across Tables | Wang et al., 2022 | Link |
More models will be integrated into the library soon!
More to come!
- We are currently under heavy development.
- To report a bug, please write anissue.
About
Archive of the Future Frame's open source Python package with inference code for foundation tabular models.