Skip to content
Robotics

GR00T N1.7 MDRIFT Clean1106v1

готовый .nqrr-бандл для NPU Qualcomm

1 min read

Это .nqrr-бандл политики GR00T N1.7 от geonmin-kim для запуска на NPU Qualcomm IQ-9075 (QCS9075) через Nota QNN Robotics Runtime. Внутри — HTP-контекстные бинарники, golden-тензоры и host glue; крупные члены упакованы в model.safetensors. Размер — 5.31 GiB, параметров около 5.37 млрд.

Ключевые особенности

  • Формат nqrr/3: контекстные бинарники, golden-тензоры и host glue в одном бандле; крупные члены лежат в model.safetensors.
  • Целевое железо — Qualcomm IQ-9075 (QCS9075), Hexagon v73; рантайм NQRR, QAIRT 2.47.0.260601.
  • Контексты dit_step_0, dit_step_1, llm_0 и vision; весовой буфер каждого должен быть меньше 2 GiB.
  • Фиксированные параметры: seq_len 64, 2 вида по 64 патча, 16 vision tokens, DiT 1 шаг, action horizon 40 и max_action_dim 132.
  • Есть golden для device gate: кадр 0 датасета и seed 0; verify сверяет хэши всех членов.

Пример

Проверка и извлечение бандла:

python -m nqrr.core.bundle_cli verify GROOT-N17-MDRIFT-CLEAN1106V1-TGT128-S64.nqrr
python -m nqrr.core.bundle_cli fingerprint GROOT-N17-MDRIFT-CLEAN1106V1-TGT128-S64.nqrr
python -m nqrr.core.bundle_cli extract GROOT-N17-MDRIFT-CLEAN1106V1-TGT128-S64.nqrr <dir>
ПодробнееРазмер — 5,696,770,457 B (5.31 GiB); формат nqrr/3, QAIRT 2.47.0.260601.
Цель: soc_id 77, dsp_arch v73, VTCM 8 MB. Лимит 2 GiB на контекст; llm_0 занимает 94.0%.
seq_len 64, vision 2 вида × 64 патча, DiT 1 шаг, action horizon 40.
Golden собран на кадре 0 датасета с seed 0; без seed его не воспроизвести.
Discuss Telegram 0

Will it run?

Runs in U8

Fits without quantisation.

  • U8weights 5.3 GB · ready files6.3 GB
  • FP8 / INT8recommendedweights 5.0 GB6.0 GB
  • 4-bitweights 2.8 GB3.6 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 5060 Ti 8GB · U8 — 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.

    # geonmin-kim/nqrr-GR00T-Mdrift_crop128_cft-Dclean1106v1-frommergedv4-step3000
    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 geonmin-kim/nqrr-GR00T-Mdrift_crop128_cft-Dclean1106v1-frommergedv4-step3000 --local-dir ~/models/nqrr-GR00T-Mdrift_crop128_cft-Dclean1106v1-frommergedv4-step3000
    # 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 nqrr-GR00T-Mdrift_crop128_cft-Dclean1106v1-frommergedv4-step3000 (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

geonmin-kim/nqrr-GR00T-Mdrift_crop128_cft-Dclean1106v1-frommergedv4-step3000
Author
geonmin-kim
Category
Robotics
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