Skip to content
Robotics

π₀.₅ stack yellow on red

чекпойнт 8k для робота-манипулятора

1 min read

Это fine-tune π₀.₅ от arkojit1 на базе lerobot/pi05_base: политика для Franka, обученная складывать жёлтый блок на красный. Датасет — 112 эпизодов Haply-телеоперации, 4-D state и 4-D end-effector delta action; модель на 3,62 млрд параметров. Опубликован чекпойнт шага 8 000 из 20 000 (10,4 эпохи).

Ключевые особенности

  • Дообучение π₀.₅ на 112 эпизодах Franka с Haply-телеоперацией; одна задача — сложить жёлтый блок на красный.
  • Обучен только action expert; SigLIP и Gemma-2B заморожены.
  • Eval loss 0.1105 на 12 отложенных эпизодах, train loss 0.094; это не success rate.
  • 8×MI300X, global batch 64, 8k из 20k шагов; лучший eval 0.1085 был на 4k.
  • На входе три камеры: observation.images.wrist, scene0, scene1 — подавать под именами датасета.

Пример

from lerobot.policies.pi05.modeling_pi05 import PI05Policy
policy = PI05Policy.from_pretrained("arkojit1/pi05_stack_yellow_on_red_e112_bs64_8k_20k")
ПодробнееОбучение: 8×MI300X, global batch 64, AdamW, LR 2.5e-5, bf16; запланировано 20k шагов.
На этом чекпойнте eval loss 0.1105; лучший eval 0.1085 на 4k, разброс ≈±0.01.
Eval loss — flow-matching loss на 12 holdout-эпизодах, не success rate и не сравним с другими ревизиями датасета.
Для загрузки нужны LeRobot 0.4.4 и HF-форк transformers для π₀ (ветка fix/lerobot_openpi).
Discuss Telegram 0 43

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.

    # arkojit1/pi05_stack_yellow_on_red_e112_bs64_8k_20k
    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 arkojit1/pi05_stack_yellow_on_red_e112_bs64_8k_20k --local-dir ~/models/pi05_stack_yellow_on_red_e112_bs64_8k_20k
    # 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 pi05_stack_yellow_on_red_e112_bs64_8k_20k (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

arkojit1/pi05_stack_yellow_on_red_e112_bs64_8k_20k
Author
arkojit1
Category
Robotics
Released
Posted

On NeyroLearn

Hugging Face

430

Ask AI about this model

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

Comments