Skip to content
Robotics

ACT ALOHA TransferCube

политика для симулятора, обученная на MacBook

2 min readupdated

Это ACT-политика для задачи AlohaTransferCube-v0 в симуляции, обученная с LeRobot 0.7.0 на датасете из 50 человеческих демонстраций. Модель с примерно 50 млн параметров обучена с нуля на Apple M4 MacBook Pro через MPS, без NVIDIA GPU: 80k шагов, batch size 8, около 10 часов. Гиперпараметры — дефолтные для LeRobot, без тюнинга.

Ключевые особенности

  • Обучение с нуля на ноутбуке Apple M4: MPS вместо CUDA, 80k шагов и batch size 8.
  • Архитектура ACT: ResNet-18 backbone с ImageNet init, chunk size 100, 100 action steps, VAE включён, KL weight 10, AdamW и learning rate 1e-5.
  • На 200 эпизодах с --seed=1000 — 75,5% успеха (151/200), 95% CI 69,1–80,9.
  • У официального чекпоинта 81,0% (162/200), 95% CI 75,0–85,8; интервалы пересекаются, на такой выборке модели статистически неразличимы.
  • На 50k шагах успех ниже: 64,0% для seed 1000 и 62,0% для seed 1001. Loss не предсказывал успех, а 50 эпизодов не позволяют ранжировать соседние чекпоинты.
  • Только симуляция, одна задача и один набор seed; успех считается по определению gym-aloha: куб поднят правой рукой и передан в левый захват.

Пример

pip install 'lerobot[aloha]==0.7.0'
lerobot-eval --policy.path=ChadSaglam/act_aloha_sim_transfer_cube --env.type=aloha --env.task=AlohaTransferCube-v0 --eval.n_episodes=200 --seed=1000 --policy.device=cuda
ПодробнееБенчмарк: 75,5% против 81,0% у официального ACT-чекпоинта; 95% доверительные интервалы Уилсона пересекаются.
На 50k шагах — 64,0% и 62,0% для двух seed; 200 эпизодов, seed 1000.
Обучение на M4: 2,2 steps/s; 50k шагов за 6 ч 22 мин, ещё 30k — за 3 ч 47 мин.
Ограничения: только симуляция, одна задача, один набор seed; чекпоинт в формате processor pipeline LeRobot 0.7.0.
Discuss Telegram 0 18

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.

    # ChadSaglam/act_aloha_sim_transfer_cube
    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 ChadSaglam/act_aloha_sim_transfer_cube --local-dir ~/models/act_aloha_sim_transfer_cube
    # 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 act_aloha_sim_transfer_cube (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

ChadSaglam/act_aloha_sim_transfer_cube
Author
ChadSaglam
Category
Robotics
Released
Posted

On NeyroLearn

Hugging Face

180

Ask AI about this model

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

Comments