zhiyuanhubj/LongRecipe

LongRecipe: Recipe for Efficient Long Context Generalization in Large Language Models

25
/ 100
Experimental

This project offers a method to enhance Large Language Models (LLMs) to understand and generate much longer text sequences, beyond their original design. It takes an existing LLM and your specific text data, then processes it through a multi-stage training 'recipe' to output a new, more capable LLM. This is for AI engineers or researchers who need to adapt LLMs for tasks requiring extensive context, such as analyzing long documents or complex conversations.

No commits in the last 6 months.

Use this if you need to train a Large Language Model to effectively process and generate extremely long text inputs and outputs without losing its core capabilities.

Not ideal if you are an end-user simply looking to use an off-the-shelf LLM, as this project requires significant technical expertise in model training and infrastructure.

LLM-fine-tuning long-context-AI natural-language-processing AI-model-training large-model-adaptation
No License Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 9 / 25
Maturity 8 / 25
Community 8 / 25

How are scores calculated?

Stars

79

Forks

4

Language

Python

License

Last pushed

Oct 16, 2024

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/transformers/zhiyuanhubj/LongRecipe"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.