Skip to content
Robotics

pi05-droid-fp16-ov

OpenVINO package for the π0.5 robotics model

2 min readupdated

This package is a converted version of the lerobot/pi05_droid model (the general-purpose π0.5 model by Physical Intelligence, fine-tuned for DROID Franka robot arms) into the OpenVINO IR format with fp16 weights. The package is optimized to run inference on Intel central (CPU) and graphics (GPU) processors using the Cascadia ov-policy engine. The build also includes the necessary normalization statistics and the PaliGemma tokenizer converted to the OpenVINO format. This model will be useful for robotics developers and engineers testing robot arms on Intel hardware.

Key features

  • Converted from PyTorch to OpenVINO IR with fp16 weights for Intel processors and graphics cards.
  • Normalization statistics added, trimmed to 8 real dimensions (7 joints and gripper).
  • One of the camera slots (right_wrist_0_rgb) is left empty in accordance with the logic of the original DROID policy.
  • The PaliGemma tokenizer is also fully converted to the OpenVINO format.

What you can do

  • Run the robotics policy on Intel hardware via the Cascadia engine.
  • Test visual data processing from cameras (base_0_rgb and left_wrist_0_rgb) and action sequence generation for the robot arm.
  • Compare the inference performance of OpenVINO on CPU and GPU with the original PyTorch version.

Example

Running the model for inference via the Cascadia engine:

cascadia run ./pi05-droid-fp16-ov --engine ov-policy --device GPU
Read moreDetails
• Weight precision: fp16 with a float32 graph.
• Input data: two 224x224 RGB cameras, an 8-value state, and a text prompt-instruction.
• Output data: a chunk of 15 actions (first 8 used).
• License: subject to the Gemma Terms of Use and Gemma Prohibited Use Policy.
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 5.8 GB · ready files6.9 GB
  • FP8 / INT8recommendedweights 2.9 GB3.7 GB
  • 4-bitweights 1.6 GB2.3 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 2060 · BF16 — from RTX 5060 Ti 8GB

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.

    # pawanpaudel93/pi05-droid-fp16-ov
    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 pawanpaudel93/pi05-droid-fp16-ov --local-dir ~/models/pi05-droid-fp16-ov
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to pi05-droid-fp16-ov
  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_droid (quants, finetunes, LoRAs) will show up under the bell.

Fine-tunes 1All on Hugging Face
LoRAs and adapters 1All on Hugging Face

About the model

pawanpaudel93/pi05-droid-fp16-ov
Author
pawanpaudel93
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