Guide: How to Run Stable Beluga on a Computer with 32GB RAM

Stable Beluga is a modern AI-powered tool for image generation. In this guide, we will show you how to run Stable Beluga on a computer with 32GB RAM, step by step.

Prerequisites

Before starting the installation, make sure your system meets the following requirements:

Operating System: Windows 10/11 or Linux (Ubuntu 20.04 LTS recommended)
Processor: Intel i7 or newer / AMD Ryzen 7 or newer
RAM: 32GB
Graphics Card: NVIDIA RTX 2060 or newer (with at least 8GB GPU memory)
Disk Space: 50GB of free space

Installing Dependencies

On Windows

Install graphics card drivers:
- Download the latest drivers for your graphics card from the official NVIDIA website.
- Follow the manufacturer's instructions to complete the installation.
Install Python:
- Download and install Python 3.8 or later from the official python.org website.
- Make sure to check the "Add Python to PATH" option during installation.
Install Git:
- Download and install Git from the official git-scm.com website.

On Linux (Ubuntu 20.04 LTS)

Update the system:
```
sudo apt update && sudo apt upgrade -y
```

Install NVIDIA drivers:

sudo apt install nvidia-driver-470
sudo reboot

Install Python and Git:

sudo apt install python3 python3-pip git

Cloning the Stable Beluga Repository

Open the terminal (or command prompt in Windows).

Run the following command:

git clone https://github.com/stability-ai/stable-beluga.git
cd stable-beluga

Installing Python Dependencies

Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

Install the required packages:
```
pip install -r requirements.txt
```

Configuring Stable Beluga

Copy the configuration file:
```
cp config.example.yaml config.yaml
```

Edit the config.yaml file in a text editor, e.g., VS Code:

model:
  name: "stable-beluga"
  path: "models/stable-beluga.safetensors"
device: "cuda"  # Use "cpu" if you don't have a graphics card
precision: "fp16"  # You can change to "fp32" if you have enough memory

Downloading the Model

Download the Stable Beluga model from the official repository:
```
wget https://example.com/models/stable-beluga.safetensors -P models/
```
(Replace the URL with the current model's address.)

Running Stable Beluga

Run the main script:
```
python main.py
```
If you encounter memory issues, try reducing the batch size in the config.yaml file:
```
batch_size: 1  # Default is 4, reduce if needed
```

Usage Examples

Generating an Image

from stable_beluga import StableBeluga

model = StableBeluga.from_config("config.yaml")
prompt = "Astronaut riding a horse on Mars"
image = model.generate(prompt)
image.save("output.png")

Generating a Series of Images

prompts = [
    "Sunset over the ocean",
    "Forest in autumn",
    "City skyline at night"
]

for prompt in prompts:
    image = model.generate(prompt)
    image.save(f"{prompt.replace(' ', '_')}.png")

Memory Optimization

If you encounter memory issues, try the following solutions:

Reduce the batch size:
```
batch_size: 1
```
Use fp32 precision:
```
precision: "fp32"
```
Disable gradients:
```
enable_gradients: false
```

Troubleshooting

GPU Memory Error

If you receive a GPU memory error, try:

Reducing the output resolution:

output_size: [512, 512]  # Reduce to [256, 256] if needed

Using a smaller model:
```
model:
  name: "stable-beluga-small"
```

Dependency Error

If you encounter dependency issues, try:

Updating pip:
```
pip install --upgrade pip
```

Installing packages manually:

pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu113

Summary

In this guide, we have shown you how to run Stable Beluga on a computer with 32GB RAM. Remember that memory optimization may be necessary depending on your hardware. With these steps, you should be able to generate high-quality images using Stable Beluga.

If you have additional questions or encounter issues, visit the official Stable Beluga forum or GitHub repository.