Sign up to save tools and stay up to date with the latest in AI
bg
bg
1

GitHub - ictnlp/LLaMA-Omni: LLaMA-Omni is a low-latency and high-quality end-to-end speech interaction model built upon Llama-3.1-8B-Instruct, aiming to achieve speech capabilities at the GPT-4o level.

Sep 19, 2024 - news.bensbites.com
LLaMA-Omni is a speech-language model developed by Qingkai Fang, Shoutao Guo, Yan Zhou, Zhengrui Ma, Shaolei Zhang, and Yang Feng. It is built on Llama-3.1-8B-Instruct and is designed for low-latency, high-quality speech interactions. The model can simultaneously generate both text and speech responses based on speech instructions. It was trained in less than three days using only four GPUs and can achieve a latency as low as 226ms.

The authors provide detailed instructions for installing and running the model, including how to clone the repository, install necessary packages, and download the required models. They also provide a demo and instructions for local inference. The code is released under the Apache-2.0 License, but as it is built on Llama 3.1, it must comply with the Llama 3.1 License. The authors encourage anyone who finds their work useful to cite their paper, "LLaMA-Omni: Seamless Speech Interaction with Large Language Models."

Key takeaways:

  • LLaMA-Omni is a speech-language model that supports low-latency and high-quality speech interactions, simultaneously generating both text and speech responses based on speech instructions.
  • The model is built on Llama-3.1-8B-Instruct, ensuring high-quality responses and it supports low-latency speech interaction with a latency as low as 226ms.
  • It is capable of simultaneous generation of both text and speech responses and was trained in less than 3 days using just 4 GPUs.
  • The model can be installed and used locally, with detailed instructions provided for installation, quick start, and local inference.
View Full Article

Comments (0)

Be the first to comment!