Skip to content
Robotics

PANDA-HSP/groot17-panda-ee-new-bread-all-noft

robotic arm control model without a force sensor

2 min readupdated

This model is a fine-tune of the 3.14-billion-parameter nvidia/GR00T-N1.7-3B base model, built to control the Panda EE robotic arm. Its main distinction is that it was trained to operate without the Robotiq FT 300 force-torque sensor typically mounted between the manipulator flange and the gripper. During data preparation, all poses in the states and actions were offset by 37.5 millimeters along the Z-axis so the gripper fingertips still reach the exact same positions as in the original demonstrations. It is useful for researchers and engineers working with the lerobot framework and Panda robots running a sensorless setup.

Key features

  • Base model: Built on NVIDIA GR00T-N1.7-3B using the lerobot library (version 0.6.2).
  • Geometry adaptation: Accounts for removing the 37.5 mm sensor by offsetting coordinates along the flange Z-axis.
  • Rotation representation: Uses a rotation vector (rotvec) with the ee_rotation_frame = "gripper_down" frame.
  • Multitask training: Trained on a combined dataset of 230 episodes (new_bread_1, 2, and 3) covering bread manipulation tasks.

What you can do

  • Move bread slices: Apply the model to robotic tasks involving transferring slices of bread from a cutting board onto a plate.
  • Assemble and disassemble sandwiches: Use it to grasp a second slice of bread and position it accurately on top of another.
  • Deploy on hardware: Run control pipelines on a real or simulated Panda EE arm without an FT300 sensor installed.

Example

During deployment and inference, make sure to use the same rotation representation and gripper frame configuration (ee_rotation_frame = "gripper_down"), and account for the missing physical force sensor in the manipulator's hardware setup.

Read moreDetails
• Architecture: Based on nvidia/GR00T-N1.7-3B (3.14B parameters)
• Training: 40,000 steps, batch size 8, lr 1e-4 (cosine scheduler)
• Dataset: 230 episodes (multitask new_bread_all_noft)
• Validation: 5% held-out episodes (eval_split=0.05)
Discuss Telegram 0 32

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # PANDA-HSP/groot17-panda-ee-new-bread-all-noft
    # ===== Tokens: paste yours between the quotes =====
    export HF_TOKEN=""        # huggingface.co/settings/tokens, for gated models
    # ==========================================
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download PANDA-HSP/groot17-panda-ee-new-bread-all-noft --local-dir ~/models/groot17-panda-ee-new-bread-all-noft
    # The model is in ~/models. How to run it is in the model card (link above).
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of groot17-panda-ee-new-bread-all-noft (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

PANDA-HSP/groot17-panda-ee-new-bread-all-noft
Author
PANDA-HSP
Category
Robotics
Released
Posted

On NeyroLearn

Hugging Face

320

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments