Skip to content
Robotics

LingBot-VA LB43 s11750

полный файнтюн VLA-модели для RoboTwin 2.0

2 min readupdated

LingBot-VA LB43 — чекпойнт step 11750 полного файнтюна трансформера robbyant/lingbot-va-base на чистых демонстрациях 43 задач RoboTwin 2.0. Размер модели — 5,09 млрд параметров; VAE и текстовый энкодер заморожены и скопированы из базовой модели без изменений. Обучение шло только на 43 «виденных» задачах, 7 невиденных задач в тренировке не участвовали.

Ключевые особенности

  • Полный файнтюн только трансформера: 10,2 ГБ в bf16 safetensors, конфиг совпадает с базовой моделью.
  • Данные: 2150 эпизодов — по 50 демонстраций на каждую из 43 задач, без рандомизированных демо.
  • Рецепт LingBot-VA: AdamW, lr 1e-5, grad clip 2.0, text dropout 0.1, эффективный batch 16 на 8×H100 с FSDP2.
  • Чекпойнт step 11750 — около 87 эпох; полный прогон доходил до 18800 шагов, но здесь опубликован не последний чекпойнт.
  • В репозитории есть конфиг обучения, метрики по шагам и все эпизоды оценки.

Пример

Сценарий из карточки — оценка на шести невиденных задачах RoboTwin 2.0 по 50 эпизодов: move_stapler_pad, cup, scale, stamp_seal, bread и stack_blocks_three. Для move_stapler_pad и stamp_seal использованы критерии успеха Zero-WAM, поэтому их результаты не сравнимы напрямую со стандартным RoboTwin.

ПодробнееНа невиденных задачах step 11750 решает 94 из 300 эпизодов (31,3%); последний чекпойнт прогона step 18800 — 109 из 300 (36,3%).
По задачам: stapler 27/50, cup 27/50, scale 7/50, seal 8/50, bread 25/50, stack 0/50.
Все эпизоды stack_blocks_three упёрлись в лимит 640 действий; open_microwave не оценивался.
Лицензия Apache-2.0, как у базовой модели.
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 9.5 GB · ready files11 GB
  • FP8 / INT8recommendedweights 4.7 GB5.7 GB
  • 4-bitweights 2.7 GB3.4 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 2080 Ti

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.

    # kenzhengjk/lingbot-va-lb43-s11750
    # ===== 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 kenzhengjk/lingbot-va-lb43-s11750 --local-dir ~/models/lingbot-va-lb43-s11750
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to lingbot-va-lb43-s11750
  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 lingbot-va-base (quants, finetunes, LoRAs) will show up under the bell.

Fine-tunes 12All on Hugging Face
LoRAs and adapters 4All on Hugging Face

About the model

kenzhengjk/lingbot-va-lb43-s11750
Author
kenzhengjk
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