Skip to content
Robotics

AgileX pour FT IDM

native-трансформер LingBot-VA для четырёх робозадач

1 min readupdated

Это полный native-трансформер LingBot-VA от MarshCurrant после action-specific supervised fine-tuning pour-чекпоинта. Несмотря на имя, адаптация шла на четырёх задачах: pour, pick, opendraw и stack3bowls. Это не LoRA-адаптер: веса не нужно накладывать на исходный трансформер.

Ключевые особенности

  • Чекпоинт step 950 выбран по минимальному fixed validation action loss 0.0159488310 в 1000-шаговом прогоне.
  • Обучены только action_embedder, action_proj_out и action time-conditioning модули; shared transformer blocks, video-модули, VAE и T5 заморожены.
  • Изменены 10 action-specific тензоров, остальные 829 — без изменений; это native transformer export, а не LeRobot PreTrainedPolicy.
  • Нужно использовать включённый normalization.json и сохранять порядок камер и action-каналов; статистики из deployment/test не пересчитывать.
  • В релизе нет LeRobot processors, frozen VAE, T5 encoder и tokenizer — нужен совместимый LingBot-VA и те же frozen-компоненты.

Пример

from huggingface_hub import snapshot_download

model_root = snapshot_download(
    repo_id="MarshCurrant/agilex_pour_ft_idm",
    local_dir="checkpoints/agilex_pour_ft_idm",
)
transformer_path = model_root + "/transformer"
ПодробнееВалидационный action loss: 0.0159488310 (шаг 950 из 1000).
Параметров: 5 088 872 670 всего, 67 058 718 обучаемых; shared-блоки, video-модули, VAE и T5 заморожены.
Нормализация q01/q99 считалась по 500 эпизодам, включая validation; для независимой проверки нужны новые сессии.
Это не заявка о real-robot success rate: fixed validation loss не равен closed-loop успеху.
Перед железом проверяйте лимиты суставов, gripper, масштабирование действий и аварийный стоп.
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.

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

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

About the model

MarshCurrant/agilex_pour_ft_idm
Author
MarshCurrant
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