Skip to content
Robotics

SemWAM-RoboTwin TwoStage v1.30

Stage2 Joint15k: чекпоинт робототехнической world-action модели

1 min read

Опубликован чекпоинт RainyBot/SemWAM-RoboTwin-TwoStage-v1.30-Stage2-Joint15k для задачи robotics. Это world-action модель (WAM) на PyTorch, построенная вокруг Wan-AI/Wan2.2-TI2V-5B: видео-DiT на 30 слоёв, отдельный action-DiT и вспомогательные depth/semantic-потоки. Сохранён optimizer step 15000.

Ключевые особенности

  • Архитектура rgb_dep_sem: RGB, depth и semantic-представления; action conditioning через predicted_depth_head_semantic.
  • Видео-DiT: 30 слоёв, hidden 3072, 24 головы, patch size 1×2×2; MoE на 4 эксперта с shared expert в слоях 21/24/27, routing per_sample и dense_masked.
  • Action-DiT: 30 слоёв, hidden 1024, 24 головы, 14-мерные действия и proprio.
  • Semantic-поток на DINOv3 ViT-B/16 (768 признаков) со SVAE-сжатием до 48 каналов; depth — через VDA/UniDepth v2 и dvae.
  • Данные: clean-сцены RoboTwin, три камеры (high + left/right wrist), 33 кадра, видео 384×320, 14D state/action.

Пример

Сценарий из карточки: на вход подаются три вида RoboTwin — cam_high и два wrist-вида, 14-мерное состояние и окно на 33 кадра. Модель должна выдать 14-мерные действия; глубина и семантика используются как вспомогательные модальности. Готового кода запуска в карточке нет — это checkpoint для собственного пайплайна.

ПодробнееСохранённый optimizer step: 15000, снимок от 2026-11-06.
Данные: flexpi_robotwin, clean-сцены, 33 кадра, видео 384×320, action/state 14D.
Бенчмарков и требований к VRAM в карточке нет; это узкоспециализированный чекпоинт для RoboTwin.
Discuss Telegram 0 20

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 14 GB · ready files16 GB
  • FP8 / INT8recommendedweights 6.8 GB8.0 GB
  • 4-bitweights 3.8 GB4.7 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 3080 · BF16 — from RTX 5080

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.

    # RainyBot/SemWAM-RoboTwin-TwoStage-v1.30-Stage2-Joint15k
    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 RainyBot/SemWAM-RoboTwin-TwoStage-v1.30-Stage2-Joint15k --local-dir ~/models/SemWAM-RoboTwin-TwoStage-v1.30-Stage2-Joint15k
    # 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 SemWAM-RoboTwin-TwoStage-v1.30-Stage2-Joint15k (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

RainyBot/SemWAM-RoboTwin-TwoStage-v1.30-Stage2-Joint15k
Author
RainyBot
Category
Robotics
Released
Posted

On NeyroLearn

Hugging Face

200

Ask AI about this model

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

Comments