abaheti95/LoL-RL
Advantage Leftover Lunch Reinforcement Learning (A-LoL RL): Improving Language Models with Advantage-based Offline Policy Gradients
This project offers a method for fine-tuning large language models using existing language data and custom reward functions. It takes your pre-existing language models and a dataset, applies a technique called Advantage Leftover Lunch Reinforcement Learning, and outputs an improved language model. This is useful for AI researchers and machine learning engineers who need to refine language models for specific tasks like generating helpful assistant responses or improving common sense.
No commits in the last 6 months.
Use this if you need to fine-tune a language model efficiently for a specific, measurable goal using an arbitrary reward function and an existing dataset.
Not ideal if you prefer to use online reinforcement learning methods or don't have a clear, quantifiable reward function for your language model's output.
Stars
26
Forks
8
Language
Python
License
MIT
Category
Last pushed
Sep 10, 2024
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/transformers/abaheti95/LoL-RL"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.