# Isaac GR00T

## Docs

- [Dataset classes](https://mintlify.wiki/NVIDIA/Isaac-GR00T/api/data/dataset.md): Dataset classes for loading and processing VLA training data
- [Embodiment tags](https://mintlify.wiki/NVIDIA/Isaac-GR00T/api/data/embodiment-tags.md): Robot embodiment identifiers for cross-embodiment training
- [Processor classes](https://mintlify.wiki/NVIDIA/Isaac-GR00T/api/data/processor.md): Data processor interface for converting raw observations into model inputs
- [Data types](https://mintlify.wiki/NVIDIA/Isaac-GR00T/api/data/types.md): Core data types and structures for VLA training
- [Closed-loop evaluation](https://mintlify.wiki/NVIDIA/Isaac-GR00T/api/eval/closed-loop.md): Run policy rollouts in simulation environments to measure task success
- [Open-loop evaluation](https://mintlify.wiki/NVIDIA/Isaac-GR00T/api/eval/open-loop.md): Evaluate policy predictions against ground truth actions from recorded trajectories
- [Evaluation server](https://mintlify.wiki/NVIDIA/Isaac-GR00T/api/eval/server.md): Run a GR00T policy as a server for distributed inference and evaluation
- [Model configuration](https://mintlify.wiki/NVIDIA/Isaac-GR00T/api/model/config.md): Configuration classes for GR00T models including backbone and action head settings
- [GR00T model class](https://mintlify.wiki/NVIDIA/Isaac-GR00T/api/model/gr00t-model.md): Vision-Language-Action model with backbone and action head for robotic control
- [Gr00tPolicy](https://mintlify.wiki/NVIDIA/Isaac-GR00T/api/policy/gr00t-policy.md): Core policy class for GR00T model inference
- [Gr00tSimPolicyWrapper](https://mintlify.wiki/NVIDIA/Isaac-GR00T/api/policy/policy-wrapper.md): Policy wrapper for compatibility with GR00T simulation environments
- [PolicyClient and PolicyServer](https://mintlify.wiki/NVIDIA/Isaac-GR00T/api/policy/server-client.md): Server-client architecture for distributed GR00T policy inference
- [launch_finetune.py](https://mintlify.wiki/NVIDIA/Isaac-GR00T/api/training/launch-finetune.md): Script for fine-tuning Vision-Language-Action (VLA) models on custom datasets
- [Training configuration](https://mintlify.wiki/NVIDIA/Isaac-GR00T/api/training/training-config.md): Configuration classes for training GR00T models
- [Model architecture](https://mintlify.wiki/NVIDIA/Isaac-GR00T/architecture.md): Understanding the GR00T N1.6 vision-language-action architecture with VLM backbone and diffusion transformer action head
- [LeRobot v2 data format](https://mintlify.wiki/NVIDIA/Isaac-GR00T/concepts/data-format.md): GR00T-flavored LeRobot v2 dataset format and structure requirements
- [Embodiment tags](https://mintlify.wiki/NVIDIA/Isaac-GR00T/concepts/embodiment-tags.md): Explanation of embodiment tags and supported robots
- [Modality configuration system](https://mintlify.wiki/NVIDIA/Isaac-GR00T/concepts/modality-configs.md): Configure how your robot's data is loaded, processed, and interpreted
- [Model overview](https://mintlify.wiki/NVIDIA/Isaac-GR00T/concepts/model-overview.md): Overview of the GR00T model architecture and capabilities
- [Docker deployment](https://mintlify.wiki/NVIDIA/Isaac-GR00T/deployment/docker.md): Deploy GR00T in a containerized environment with CUDA support and pre-installed dependencies
- [Optimization techniques](https://mintlify.wiki/NVIDIA/Isaac-GR00T/deployment/optimization.md): Optimize GR00T inference performance with torch.compile and TensorRT acceleration
- [Server-client architecture](https://mintlify.wiki/NVIDIA/Isaac-GR00T/deployment/server-client.md): Deploy GR00T inference with server-client architecture for remote GPU execution
- [TensorRT optimization](https://mintlify.wiki/NVIDIA/Isaac-GR00T/deployment/tensorrt.md): Accelerate GR00T inference with TensorRT engine compilation for up to 2x faster performance
- [BEHAVIOR benchmark](https://mintlify.wiki/NVIDIA/Isaac-GR00T/examples/behavior.md): 50 household loco-manipulation tasks with Galaxea R1 Pro robot
- [DROID benchmark](https://mintlify.wiki/NVIDIA/Isaac-GR00T/examples/droid.md): Real-world manipulation with the DROID dataset
- [Unitree G1 locomanipulation](https://mintlify.wiki/NVIDIA/Isaac-GR00T/examples/g1-locomanipulation.md): Whole-body control tasks for Unitree G1 humanoid robot
- [LIBERO benchmark](https://mintlify.wiki/NVIDIA/Isaac-GR00T/examples/libero.md): Lifelong robot learning benchmark with multiple task suites
- [Benchmarks and examples](https://mintlify.wiki/NVIDIA/Isaac-GR00T/examples/overview.md): Overview of available benchmarks and example implementations for Isaac GR00T
- [PointNav benchmark](https://mintlify.wiki/NVIDIA/Isaac-GR00T/examples/pointnav.md): Point navigation tasks with COMPASS-generated datasets
- [RoboCasa benchmark](https://mintlify.wiki/NVIDIA/Isaac-GR00T/examples/robocasa.md): Large-scale kitchen simulation with 2,500+ assets and 100 manipulation tasks
- [SimplerEnv benchmark](https://mintlify.wiki/NVIDIA/Isaac-GR00T/examples/simplerenv.md): Framework for evaluating real-world robot manipulation policies in GPU-accelerated simulation
- [SO-100 benchmark](https://mintlify.wiki/NVIDIA/Isaac-GR00T/examples/so100.md): Fine-tuning and deployment for SO-100 robot arms
- [Hardware requirements](https://mintlify.wiki/NVIDIA/Isaac-GR00T/getting-started/hardware-requirements.md): Recommended hardware for training and deploying GR00T models
- [Installation](https://mintlify.wiki/NVIDIA/Isaac-GR00T/getting-started/installation.md): Install Isaac GR00T with uv or Docker
- [Quick start](https://mintlify.wiki/NVIDIA/Isaac-GR00T/getting-started/quickstart.md): Run inference with GR00T in minutes
- [Custom embodiment](https://mintlify.wiki/NVIDIA/Isaac-GR00T/guides/custom-embodiment.md): Adding support for new robot embodiments
- [Data preparation](https://mintlify.wiki/NVIDIA/Isaac-GR00T/guides/data-preparation.md): How to prepare robot demonstration data for GR00T training
- [Evaluation](https://mintlify.wiki/NVIDIA/Isaac-GR00T/guides/evaluation.md): Open-loop and closed-loop evaluation of GR00T models
- [Fine-tuning](https://mintlify.wiki/NVIDIA/Isaac-GR00T/guides/finetuning.md): Fine-tuning GR00T on custom datasets
- [Inference](https://mintlify.wiki/NVIDIA/Isaac-GR00T/guides/inference.md): Running inference with GR00T models
- [Policy API](https://mintlify.wiki/NVIDIA/Isaac-GR00T/guides/policy-api.md): Using the Gr00tPolicy API for inference
- [Introduction](https://mintlify.wiki/NVIDIA/Isaac-GR00T/introduction.md): NVIDIA Isaac GR00T N1.6 is an open vision-language-action model for generalized humanoid robot skills
- [Contributing](https://mintlify.wiki/NVIDIA/Isaac-GR00T/resources/contributing.md): How to contribute to the Isaac GR00T project
- [Frequently asked questions](https://mintlify.wiki/NVIDIA/Isaac-GR00T/resources/faq.md): Common questions about Isaac GR00T
- [Model checkpoints](https://mintlify.wiki/NVIDIA/Isaac-GR00T/resources/model-checkpoints.md): Available GR00T model checkpoints for base models and finetuned variants
- [Troubleshooting](https://mintlify.wiki/NVIDIA/Isaac-GR00T/resources/troubleshooting.md): Common issues and solutions for Isaac GR00T
- [What's new](https://mintlify.wiki/NVIDIA/Isaac-GR00T/whats-new.md): GR00T N1.6 introduces significant improvements in model architecture, training data, and code infrastructure compared to N1.5