Skip to content

Setup and System Requirements

This guide covers the setup requirements for running Cosmos Predict 2 for traffic anomaly generation using LoRA post-training.

System Requirements

Minimum Hardware Requirements

  • GPU: Single node with 8 GPUs (A100, H100, or later recommended)
  • Memory: Sufficient VRAM for model inference and training (50GB+ recommended)
  • Storage: Adequate space for model weights, datasets, and checkpoints

Software Requirements

The setup requires the Cosmos Predict 2 repository and model to be properly installed and configured.

Installation

Cosmos Predict 2 Setup

To set up Cosmos Predict 2 repository and model, follow the detailed installation and inference setup instructions at:

Cosmos Predict 2 Installation Guide

Verification

After completing the installation, verify the setup by running the inference examples provided in the Cosmos Predict 2 repository to ensure the model is working correctly before proceeding with the post-training pipeline.

Next Steps

Once the setup is complete, proceed to the post-training tutorial to learn how to use Cosmos Predict 2 for traffic anomaly generation with LoRA adaptation.