Skip to content

Setup and System Requirements

This guide covers the setup requirements for post-training Cosmos Reason 2 for intelligent transportation.

System Requirements

Minimum Hardware Requirements

  • GPU: 8xA100, H100, or later GPUs (recommended minimum).
  • Memory: Sufficient VRAM for model training and inference. We recommend at least 80GB per GPU.
  • Storage: Adequate space for model weights and datasets.

Software Requirements

The setup requires the Cosmos Reason 2 and Cosmos-RL repositories and Cosmos Reason 2-8B model to be properly installed and configured.

Installation

Cosmos Reason 2 Setup

To set up Cosmos Reason 2 repository and model, follow the detailed installation and post-training environment setup instructions at:

Cosmos Reason 2 Post-Training Installation Guide Cosmos-RL Installation Documentation

Verification

After completing the installation, verify the setup by running the inference examples provided in the Cosmos Reason 2 repository on your custom traffic video and question 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 post-train Cosmos Reason 2 for intelligent transportation.