Skip to content
Other

GR00T N1.6-3B

робот-политика для 8 задач LIBERO

1 min read

Это inference-чекпоинт checkpoint-30000 из запуска lb_t8_gr00t16 — файнтюн nvidia/GR00T-N1.6-3B на восьми задачах LIBERO. Обучение шло на Isaac-GR00T n1.6.1-release с embodiment-тегом LIBERO_PANDA, action horizon 16 и без relative actions; в репозитории только файлы для инференса, без optimizer state.

Ключевые особенности

  • База — nvidia/GR00T-N1.6-3B; код Isaac-GR00T n1.6.1-release, examples/finetune.sh без изменений.
  • Embodiment-тег LIBERO_PANDA; modality.json взят из examples/LIBERO; action horizon 16, без relative actions.
  • 30 000 шагов, global batch 64 (32 на GPU × 2 H200), lr 1e-4, warm-up 5 %, weight decay 1e-5, state dropout 0.
  • Нормализация — собственная statistics.json датасета; у базовой модели нет статистик для libero_panda.
  • Восемь задач LIBERO: перенос миски, салатницы, соуса, открытие ящика, включение плиты и др.
  • Только inference-файлы checkpoint-30000 — без optimizer state.

Пример

python run_gr00t_server.py --model-path checkpoint-30000 --embodiment-tag LIBERO_PANDA --use-sim-policy-wrapper
python rollout_policy.py --n_action_steps 8 --max_episode_steps 720
ПодробнееОбучение: 30 000 шагов, global batch 64 (2×H200), lr 1e-4, warm-up 5 %.
Данные: 27 из ~30 человеческих демонстраций на задачу из hmkang/libero-t8-gr00t.
В карточке нет результатов бенчмарков; это inference-checkpoint без optimizer state.
Требования к VRAM не указаны; покрыты только 8 задач LIBERO с embodiment LIBERO_PANDA.
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 6.1 GB · ready files7.2 GB
  • FP8 / INT8recommendedweights 3.1 GB3.9 GB
  • 4-bitweights 1.7 GB2.4 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 · BF16 — from RTX 5060 Ti 8GB

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.

    # hmkang/gr00t16-libero-t8
    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 hmkang/gr00t16-libero-t8 --local-dir ~/models/gr00t16-libero-t8
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to gr00t16-libero-t8
  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 gr00t16-libero-t8 (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

hmkang/gr00t16-libero-t8
Author
hmkang
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

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

Comments