Search Packages

dataflex

DataFlex is a data-centric training framework that enhances model performance by either selecting the most influential samples, optimizing their weights, or adjusting their mixing ratios.

transformers-domain-adaptation

:no_entry: [DEPRECATED] Adapt Transformer-based language models to new text domains

scar-tool

SCAR: An AI-powered tool for ranking and filtering instruction-answer pairs based on writing quality and style consistency

data-selection

Data Selection with Importance Resampling

gen-dedup

Code for Generative Deduplication For Socia Media Data Selection (Findings of EMNLP 2024)