GR00T N1.5 Fine-tuned on SO-100 Dataset

This model is a fine-tuned version of nvidia/GR00T-N1.5-3B on a SO-100 robot arm dataset with single camera setup.

Model Details

  • Base Model: nvidia/GR00T-N1.5-3B
  • Fine-tuned on: SO-100 table cleanup dataset
  • Camera Setup: Single camera (webcam)
  • Embodiment: SO-100 robot arm
  • Training Steps: 1000 (demo configuration)

Usage

from gr00t.model.gr00t_n1 import GR00T_N1_5

# Load the fine-tuned model
model = GR00T_N1_5.from_pretrained("HelloCephalopod/gr00t-n1.5-so100-finetuned")

# Use for inference on SO-100 robot

Training Details

  • Training Data: SO-100 pick and place block task
  • Training Steps: 1000
  • Batch Size: 8
  • Learning Rate: 1e-4
  • Hardware: Google Colab GPU

Intended Use

This model is intended for research and educational purposes with SO-101 robot arms performing table cleanup tasks.

Limitations

  • Trained on a small dataset (demo purposes)
  • Single camera setup only
  • Specific to SO-100 embodiment

Citation

If you use this model, please cite the original GR00T paper and acknowledge the fine-tuning work.

Downloads last month
4
Safetensors
Model size
2.72B params
Tensor type
F32
·
BF16
·
Video Preview
loading

Model tree for HelloCephalopod/gr00t-n1.5-so100-finetuned

Finetuned
(3)
this model