Introducing Llama 3 VerusGPT – Open-Source Training Data and Domain-Expert LLM for Verus & Other Uses
June 21 2024 - 2:37PM
The Verus community is excited to announce the launch of Llama 3
VerusGPT, a fully open-source domain-expert Language Model (LLM)
designed to answer questions about the Verus Protocol. This AI has
been trained to specialize in Verus, aiming to educate people about
Verus in the short term and develop AI participants for Verus'
decentralized blockchain economy in the long term.
Helping Users & Builders
Llama 3 VerusGPT is the first LLM dedicated to helping users
learn about cryptocurrency and the unique features of the Verus
Protocol. This AI tool is designed for both newcomers and builders,
making the Verus ecosystem more accessible. VerusGPT can answer any
questions related to the Verus Protocol and its advanced
capabilities.
To try VerusGPT or use it as an example for your own projects,
join the Verus Discord, and visit the #verus-ai channel. As a
model that can run locally or on a server, you can also access the
model, files, and a tutorial on how all of this was done, which can
serve as a step by step tutorial or starting point for anyone
working to train their own large language models on their company
or community information. With its permissive license, others are
welcome to also integrate this data into their own LLM projects.
You can access the model files, data pipelines, and step by step
instructions on HuggingFace at this
link: https://huggingface.co/VerusCommunity/Llama-3-VerusGPT
The dataset generation code and datasets used to train Llama 3
VerusGPT are publicly available. You can find them here.
Open-Source for Everyone
This open-sourcing provides a powerful use case and demonstrates
how domain-expert LLMs can be rapidly developed. By sharing these
resources, Verus invites other companies, projects, and communities
to learn how to train their own specialized AI models.
If you want to collaborate on this open-source AI
effort, join the Verus Discord.
Dataset Generation and Training
Llama 3 VerusGPT is built with Meta's Llama 3, trained on the
base Llama 3 8b model. It uses a combination of generalized
assistant-style data and special instruct tuning data focused on
the Verus Protocol. Find the data set generation code
here.
Data has been used
from: docs.verus.io, wiki.verus.io, Medium, API
documentation & the Verus Vision paper.
The goal behind this project is to produce a domain-expert AI
capable of answering basic to intermediate questions about the
Verus Protocol, community, and project. Additionally, it
understands the Verus community ethos and mission, effectively
communicating this to the users.
An AI-Enhanced Future
VerusGPT paves the way for integrating Verus' self-sovereign and
private identity protocols (VerusID) with LLMs, creating
next-generation social applications. These applications will offer
conversational experiences free from the biases of any particular
company.
Decentralized and unbiased social media platforms can create
more transparent, equitable, and user-centric experiences, giving
individuals full control over their content and interactions.
Join the Movement
The global Verus community invites everyone to participate in
this open-source movement to create LLMs for various communities
and organizations.
We also encourage you to start interacting with VerusGPT
on Discord, where you can ask any questions about the Verus
Protocol and receive expert answers.
X: https://x.com/veruscoinTelegram:
https://t.me/veruscommunityYoutube:
https://www.youtube.com/@VerusCoinCommunityReddit:
https://www.reddit.com/r/VerusCoin/Medium:
https://medium.com/veruscoin
Rahul K
PR Manageer
Verus
press at veruscoin.io