Installation Guide

This guide will help you install and set up InfinityStar on your system. Follow these steps carefully to ensure proper installation and configuration.

System Requirements

Before installing InfinityStar, ensure your system meets the following requirements:

• Python 3.8 or higher
• PyTorch 2.5.1 or higher (required for FlexAttention support)
• CUDA-capable GPU with sufficient memory (recommended for inference)
• At least 40GB of free disk space for model checkpoints
• Sufficient RAM for model loading and inference

Step 1: Install PyTorch

InfinityStar uses FlexAttention to speed up training, which requires PyTorch version 2.5.1 or higher. Install PyTorch with CUDA support if you have a compatible GPU:

For CPU-only installation (not recommended for inference):

Verify your PyTorch installation by checking the version:

Ensure the version is 2.5.1 or higher.

Step 2: Install Required Packages

Clone the InfinityStar repository and navigate to the project directory:

Install all required Python packages using pip:

This will install all necessary dependencies including transformers, datasets, and other required libraries.

Step 3: Download Model Checkpoints

Download the appropriate model checkpoints based on your needs:

• 720p Model: For high-quality 5-second video generation at 720p resolution. This model is optimized for text-to-video tasks.
• 480p Model: For variable-length video generation (5 or 10 seconds) at 480p resolution. This model works well for image-to-video and video-to-video tasks.

Model checkpoints are available from the official repository. The full model size is approximately 35 gigabytes, so ensure you have sufficient storage space and bandwidth.

Place the downloaded checkpoints in the appropriate directory as specified in the repository structure.

Step 4: Verify Installation

After installation, verify that all components are working correctly:

If all commands execute without errors, your installation is complete.

Step 5: Prepare for Inference

Before running inference, ensure you have:

• Downloaded the appropriate model checkpoint
• Placed the checkpoint in the correct directory
• Prepared your input (text description, image, or video)
• Ensured sufficient GPU memory is available

Running Inference

For 720p video generation, use the following command:

For 480p variable-length video generation, use:

You can specify parameters such as text prompts, image paths, or video paths in the script or through command-line arguments, depending on the implementation.

Training Setup

If you plan to train or fine-tune the model, additional setup is required:

• Organize your training data according to the structure specified in data/README.md
• Extract features from your data using the provided feature extraction scripts
• Configure training parameters in the training scripts
• Ensure you have sufficient computational resources for training

For detailed training instructions, refer to the data/README.md file in the repository.

Troubleshooting

If you encounter issues during installation:

• Ensure PyTorch version is 2.5.1 or higher
• Check that all dependencies are installed correctly
• Verify CUDA compatibility if using GPU
• Check available disk space for model checkpoints
• Review error messages for specific dependency issues

Additional Resources

For more information, refer to:

• Official repository documentation
• Training scripts and data organization guides
• Inference script documentation
• Research paper for technical details