thuml/LogME

Code release for "LogME: Practical Assessment of Pre-trained Models for Transfer Learning" (ICML 2021) and Ranking and Tuning Pre-trained Models: A New Paradigm for Exploiting Model Hubs (JMLR 2022)

/ 100

Emerging

This tool helps machine learning engineers and researchers quickly assess which pre-trained deep learning models will perform best on a new dataset or task, without needing extensive fine-tuning. You provide feature representations and corresponding labels from your dataset, and it outputs a 'LogME score' that predicts transfer learning performance. This allows you to efficiently select the most compatible pre-trained model for your specific classification or regression problem.

211 stars. No commits in the last 6 months.

Use this if you need to choose the best pre-trained model from a diverse collection (like a model hub) for a new task, and you want to avoid time-consuming hyper-parameter tuning for each candidate model.

Not ideal if you are looking for a tool to perform the actual fine-tuning of pre-trained models, as this focuses solely on ranking and selection.

transfer-learning model-selection deep-learning-applications applied-ai machine-learning-research

Stale 6m No Package No Dependents

Maintenance 0 / 25

Adoption 10 / 25

Maturity 16 / 25

Community 13 / 25

How are scores calculated?

Stars

211

Forks

Language

Python

License

MIT

Higher-rated alternatives

limix-ldm-ai/LimiX

LimiX: Unleashing Structured-Data Modeling Capability for Generalist Intelligence...

tatsu-lab/stanford_alpaca

Code and documentation to train Stanford's Alpaca models, and generate the data.

google-research/plur

PLUR (Programming-Language Understanding and Repair) is a collection of source code datasets...

YalaLab/pillar-finetune

Finetuning framework for Pillar medical imaging models.

joisino/reeval-wmd

Code for "Re-evaluating Word Mover’s Distance" (ICML 2022)

Explore ML Frameworks

All categories Trending ML Framework directory Insights