Skip to content
Robotics

PLWM

point-latent world model for robotics

2 min readupdated

PLWM (point-latent world model) is a research checkpoint for robotics that extends the Wan2.2-TI2V-5B architecture with a 3D point stream and a built-in action decoder. Trained on real-world, four-camera bimanual robot data, it predicts both visual transitions and manipulator actions. Under the hood, it combines the Wan2.2 video diffusion transformer with a frozen point autoencoder, making it a solid base for researchers and engineers working on robotics and world models.

Key features

  • Built on Wan2.2-TI2V-5B with an added 3D point-latent stream.
  • Integrated action decoder for bimanual end-effector actions running at 10 Hz.
  • Trained on real-world bimanual robot episodes recorded from four camera angles.
  • Includes the final training checkpoint from 8 H100 GPUs (step 10,000) along with the package code.

What you can do

  • Run a CPU model-loading smoke test using the included script.
  • Set up inference by linking public Wan2.2-TI2V-5B weights and the umt5-xxl text encoder.
  • Use the package codebase and Hydra configs to run your own world-model experiments.

Example

pip install torch hydra-core omegaconf einops imageio rich safetensors transformers
cd world_model && python load_test.py          # добавьте --ema для весов EMA
Read moreDetails
Built on Wan2.2-TI2V-5B and a frozen SE(3)-latent point AE (144 slots of 128 channels each).
Dataset: 46 training and 5 validation episodes of a bimanual robot captured at 10 Hz.
Optimization: DDP across 8x H100, global batch size 16, lr 1e-4, 8-bit AdamW, bf16.
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 19 GB · ready files21 GB
  • FP8 / INT8recommendedweights 9.4 GB11 GB
  • 4-bitweights 5.3 GB6.3 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 2080 Ti · BF16 — from RTX 4090

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.

    # 24jihoward/plwm-realworld
    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 24jihoward/plwm-realworld --local-dir ~/models/plwm-realworld
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to plwm-realworld
  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 plwm-realworld (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

24jihoward/plwm-realworld
Author
24jihoward
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