<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> LM-Nav provides a framework for robotic navigation powered by large pre-trained models encompassing language, vision, and action. It targets researchers and engineers in robotics and AI, enabling robots to interpret and act upon natural language commands for navigation tasks.

How It Works

The system integrates large language models (e.g., GPT-3), vision-language models (e.g., CLIP), and a custom graph search algorithm. This allows for flexible, language-guided control, where natural language queries are translated into navigation actions through a pipeline involving model inference and graph traversal.

Quick Start & Requirements

• Installation: pip install .
• Prerequisites: Python 3.7.13, GPU, CUDA 10.2.
• API Keys: Requires OpenAI API key for GPT-3 and GooseAI API key for open-source model ablations.
• Execution: Run experiments via jupyter_experiment.ipynb or colab_experiment.ipynb. Ablation studies are in ablation_text_to_landmark.ipynb.
• Resources: Setup involves installing dependencies and configuring API keys.

Highlighted Details

• Leverages GPT-3, CLIP, and a custom graph search for navigation.
• Includes notebooks for full pipeline execution and language processing ablation studies.
• Features cached OpenAI API calls for sample queries.

Maintenance & Community

No specific details on maintenance, community channels, or contributors are present in the provided README.

Licensing & Compatibility

The README does not specify a software license. Compatibility for commercial use or closed-source linking is undetermined.

Limitations & Caveats

The README does not detail specific limitations, unsupported platforms, or known bugs. The project appears to be research code, and its stability or production readiness is not explicitly stated.