Skip to content
Other

Zero-WAM Franka

три файнтюна для реальной руки Franka

1 min readupdated

Репозиторий с чекпоинтами трёх прогонов на реальной Franka с камерой на голове и на запястье. Все запуски используют одни данные: четыре задачи и 601 эпизод телеоперации. Внутри — полный transformer Zero-WAM, head fine-tuning и LingBot-VA.

Ключевые особенности

  • icl_4tasks_676_1102: обучается весь transformer Zero-WAM, старт с zero-wam-pretrain, около 20 ГБ на чекпоинт.
  • headft_pretrain_4tasks_1107: обучаются только дельты 24 голов video-expert и 48 голов action-expert, transformer заморожен; 0.45 ГБ.
  • lingbot_4tasks_1106: полный transformer LingBot-VA, 10.2 ГБ, старт с lingbot-va-base.
  • В каждом прогоне два чекпоинта — step 2000 и step 4000, плюс metrics.jsonl и SHA256SUMS.
  • Видео человеческих демонстраций в репозиторий не входят.

Пример

Чекпоинт icl_4tasks_676_1102 не обучался на одной инструкции, поэтому в карточке его советуют запускать в ICL-режиме с человеческой демонстрацией задачи. Для LingBot-VA достаточно языковой инструкции.

Подробнее601 эпизод: strawberry_plate — 100, two_plushies — 200, cube_drawer — 101, red_cube_on_blue — 200.
4000 шагов оптимизатора по 16 сэмплов, 8×A100-80GB, около 106 проходов по данным.
Loss на train у icl-прогона: video 0.217→0.092, action 0.0149→0.0032; сравнивать чекпоинты нужно на роботе.
Код обучения и сервинга для headft и LingBot-VA ещё не выложен; VAE, text encoder и tokenizer берутся из zero-wam-pretrain.
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 19 GB · ready files21 GB
  • FP8 / INT8recommendedweights 9.5 GB11 GB
  • 4-bitweights 5.4 GB6.4 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 2080 Ti · BF16 — from RTX 4090

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.

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

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

About the model

zhuyuze/zero-wam-franka-ckpts
Author
zhuyuze
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