---
license: apache-2.0
datasets:
- anonymised-submission/toys2box
language:
- en
base_model:
- nvidia/GR00T-N1.5-3B
pipeline_tag: robotics
tags:
- Robotics
- gr00t_n1_5
---
# Model Card

[NVIDIA Isaac GR00T N1.5](https://arxiv.org/abs/2503.14734) is an open foundation model for generalized humanoid robot reasoning and skills. 
This cross-embodiment model takes multimodal input, including language and images, to perform manipulation tasks in diverse environments.

GR00T N1.5 is trained on an expansive humanoid dataset, consisting of real captured data, synthetic data generated using the components of NVIDIA Isaac GR00T Blueprint (examples of neural-generated trajectories), and internet-scale video data. 
It is adaptable through post-training for specific embodiments, tasks and environments.

## How to Get Started with the Model

For a complete walkthrough, see the [training guide](https://github.com/NVIDIA/Isaac-GR00T).
Below is the short version on how to train and run inference/eval:

### Evaluate the policy/run inference

Run the newly trained model

```bash
python scripts/inference_service.py --server \
    --model-path <MODEL_PATH> \
    --embodiment-tag new_embodiment
    --data-config <DATA_CONFIG>
```

Run the offline evaluation script

```bash
python scripts/eval_policy.py --plot \
    --dataset-path <DATASET_PATH> \
    --embodiment-tag new_embodiment \
    --data-config <DATA_CONFIG>
```