Skip to content
Robotics

GR00T N1.7 for robotic arm: model for block-moving task

2 min read

A new operating policy for robotics, trained on NVIDIA's GR00T N1.7 architecture. The model is built to let a robot perform precise actions based on camera feeds and manipulator state. Trained using the LeRobot library, it specializes in the basic task of picking up and moving items. Useful for researchers and engineers working with physical robotic arms.

Key features

  • GR00T N1.7 foundation: built on NVIDIA's advanced architecture for humanoid robots and robotics tasks.
  • Multimodal input: processes data from two cameras (head and table) along with the current robot state.
  • Narrow specialization: the model is trained on the task of moving a block into a box.
  • LeRobot integration: fully supported by the LeRobot ecosystem for quick launch and fine-tuning.

What you can do

  • Run on a robot: apply the ready-made policy to perform the block-in-box task using a rollout script.
  • Use the dataset: explore the training demonstrations and trajectories from the associated cube_task_v1 dataset.
  • Fine-tune the model: use the collected episodes to adapt the robot to new scenarios and environments.

Example

Running the ready-made policy on your robot using the LeRobot command-line utility:

lerobot-rollout \
  --strategy.type=base \
  --robot.type=so_follower \
  --robot.port= \
  --robot.cameras="{ head: {type: opencv, index_or_path: 0, width: 640, height: 480, fps: 30}, table: {type: opencv, index_or_path: 1, width: 640, height: 480, fps: 30}}" \
  --policy.path=ButteredWaffle/cube_task_v1_groot_n17 \
  --task="Put the cube in the box" \
  --duration=60
Read moreDetails
License: apache-2.0
Robot type: so_follower (cameras: head, table)
Training data: 75 episodes, 37,058 frames at 30 FPS
Training parameters: 7,000 steps, batch size 4, adamw optimizer (lr=0.0001)
Discuss Telegram 0 18

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 12 GB · ready files13 GB
  • BF16weights 5.9 GB6.9 GB
  • FP8 / INT8recommendedweights 2.9 GB3.7 GB
  • 4-bitweights 1.6 GB2.3 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · F32 — from RTX 5080

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

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.

    # ButteredWaffle/cube_task_v1_groot_n17
    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 ButteredWaffle/cube_task_v1_groot_n17 --local-dir ~/models/cube_task_v1_groot_n17
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to cube_task_v1_groot_n17
  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 cube_task_v1_groot_n17 (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

ButteredWaffle/cube_task_v1_groot_n17
Author
ButteredWaffle
Category
Robotics
Released
Posted

On NeyroLearn

Hugging Face

180

Ask AI about this model

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

Comments