Skip to content
Robotics

Bagel-Robotics-1

чекпоинт 1000 для роботизированной политики

2 min read

Bagel-Robotics-1 — одобренное публичное зеркало точного checkpoint 1000 из bageldotcom/Bagel-Robotics-1, использованного в кампании оценки. Это compact-aware VPP2-артефакты на компонентах Wan2.1, а не Transformers AutoModel: видео-эксперт содержит 12 468 409 664 параметра. Релиз включает веса и код инференса через адаптер XPolicyLab.

Ключевые особенности

  • Публичное зеркало checkpoint 1000: файлы можно скачать без токена.
  • Видео-эксперт на 23,95% меньше зафиксированного родителя: 12 468 409 664 параметра.
  • Learned masks преобразованы в меньшие FFN и cross-attention слои; self-attention heads и action expert сохранены.
  • Для инференса не нужны teacher или learned-mask модуль.
  • Адаптер XPolicyLab: policy_name=bagel_robotics_1, ckpt_name=cp1000, действия ARX X5 end-effector, один environment на процесс политики.
  • Настройки инференса: BF16; 32 предсказанных действия и 16 исполняемых; 10 шагов Euler; shift 1; blending 90% новых / 10% предыдущих.

Пример

hf download GnanaPraveen/Bagel-Robotics-1-Public \
  --revision 2651631c5d08819088475e8778676f5cf6d9afcf \
  --include video.pt action.pt compact_architecture.json dataset_stats.json COMPLETE.json validation.json \
  --local-dir /absolute/path/to/checkpoints/cp1000
Подробнее Чекпоинт-тензоры занимают ~29,0 ГБ; ещё ~16,7 ГБ — отдельные VAE/conditioning/tokenizer assets.
Квалификация: 96 ГБ RTX PRO 6000 Blackwell, Python 3.10.21, PyTorch 2.11.0+cu130; меньший объём GPU не проверялся.
Сокращение параметров не заявлено как измеренное ускорение инференса.
Полная multi-seed валидация для официального лидерборда отдельно; исторические оценки не являются официальными результатами.
В релизе нет optimizer/scheduler/dataset state — точное продолжение обучения из весов не поддерживается. Код обучения, прунинга и экспорта обещан к 20 ноября 2026.
Discuss Telegram 0

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.

    # GnanaPraveen/Bagel-Robotics-1-Public
    # ===== 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 GnanaPraveen/Bagel-Robotics-1-Public --local-dir ~/models/Bagel-Robotics-1-Public
    # 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 Bagel-Robotics-1-Public (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

GnanaPraveen/Bagel-Robotics-1-Public
Author
GnanaPraveen
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