Skip to content
Robotics

few-shot-run-2

SFT-чекпоинт MolmoAct2 для LIBERO

1 min read

Это SFT-чекпоинт MolmoAct2 (5.6 млрд параметров) от Riasok на шаге 5000, обученный локально на 10 демонстрациях на задачу для 40 задач LIBERO — всего 400 демонстраций и 66 788 кадров. Базовая модель — allenai/MolmoAct2, библиотека LeRobot. Это стартовый SFT-чекпоинт для нового Spatial PPO-эксперимента на 150 обновлений.

Ключевые особенности

  • Обучение: 10 демонстраций на задачу, 40 задач LIBERO, 400 демонстраций и 66 788 кадров.
  • VLM LoRA rank 64 / alpha 16; action expert также обучался.
  • LIBERO Spatial success: 402/500 (80.4%).
  • В репозитории — веса политики LeRobot, пре/постпроцессоры и тензоры нормализации; нужна training-enabled реализация MolmoAct2.
  • Отдельная модель от Riasok/few-shot-run-1 и старого PPO Riasok/PPO-run1.

Пример

from huggingface_hub import snapshot_download

local_dir = snapshot_download("Riasok/few-shot-run-2")
# далее загрузите local_dir через MolmoAct2Policy.from_pretrained
# и сохранённые процессоры
ПодробнееLIBERO Spatial: 402/500 (80.4%); оценка только Spatial, хотя SFT шёл на всех четырёх наборах LIBERO.
Eval: seeds 1000+, 50 settling steps, horizon 280, 10 flow steps, 256px agent/wrist RGB, FP32 inference.
Обучение: BF16, global batch 16 (4 на GPU), VLM LoRA rank 64 / alpha 16.
Официальный full-data референс для LIBERO — allenai/MolmoAct2-LIBERO от AllenAI.
Discuss Telegram 0 31

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.

    # Riasok/few-shot-run-2
    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 Riasok/few-shot-run-2 --local-dir ~/models/few-shot-run-2
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to few-shot-run-2
  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 few-shot-run-2 (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

Riasok/few-shot-run-2
Author
Riasok
Category
Robotics
Released
Posted

On NeyroLearn

Hugging Face

310

Ask AI about this model

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

Comments

Robotics

FlowPilot

FlowPilot