Lightning-AI / Lightning-AI/litgpt

Adding DoRA (Weight-Decomposed Low-Rank Adaptation) to improve LoRA

Open
#934 20 comments 6 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

enhancement fine-tuning
Dominant language
Python
Stars
13.7k
Forks
1.5k
Avg merge
15h 37m
Merged PRs (30d)
1

Description

The new DoRA ([https://arxiv.org/abs/2402.09353](https://arxiv.org/abs/2402.09353)) method is a super promising improvement of LoRA.

Screenshot 2024-02-18 at 1 41 03 PM

If there's interest, we can add it to Lit-GPT. Perhaps as an arg setting for the existing LoRA scripts.

The authors didn't release the code yet, but I have a working from-scratch implementation here: https://github.com/rasbt/dora-from-scratch

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reading the linked DoRA paper and the rasbt/dora-from-scratch implementation, then inspect Lit-GPT's existing LoRA scripts. The issue suggests exposing DoRA as an argument, but does not name files or tests; done would mean a working DoRA option integrated with the existing LoRA workflow.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
ai, machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.