Skip to content
Robotics

task1task2_subtask_rel

multi-task robotic arm control model

2 min readupdated

Built on the pi0.5 architecture (a versioned robot control model from Physical Intelligence / LeRobot), this multi-task model is trained for three manipulation tasks: inserting a board into a module, grasping a cable, and moving the module to the center of the table. It uses a vision-language backbone (VLM) to predict intermediate subtasks before generating control actions. Combining several datasets, the model processes feeds from three cameras alongside sensor state data (a 24-dimensional vector), making it a solid fit for robotics researchers and developers working with the LeRobot framework.

Key features

  • Supports three different robotic arm scenarios through a single pipeline.
  • Uses subtask prediction for accurate action planning.
  • Trained on combined demonstrations (GRC, Yonsei, Hyundai) with a batch size of 128.
  • Leverages bf16 precision and gradient checkpointing for stable training.

What you can do

  • Run model inference to control a robotic arm for assembly and mounting tasks.
  • Test multi-task prompting to switch between inserting a board and grasping a cable.
  • Integrate the model with the LeRobot plugin to customize it for your own robotics testbeds.

Example

# pip install git+[https://github.com/HDR-SRL/lerobot_policy_pi05_subtask@multitask](https://github.com/HDR-SRL/lerobot_policy_pi05_subtask@multitask)
import lerobot_policy_pi05_subtask
from lerobot.policies.factory import make_pre_post_processors
from lerobot_policy_pi05_subtask import PI05SubtaskPolicy
policy = PI05SubtaskPolicy.from_pretrained("chomeed/task1task2_subtask_rel_chunk30_sirius_bs128_multitask_fromR2mixed_1118_20k")
pre, post = make_pre_post_processors(policy.config, pretrained_path="chomeed/task1task2_subtask_rel_chunk30_sirius_bs128_multitask_fromR2mixed_1118_20k")
batch = pre({"observation.state": state, "observation.images.d405_left": ..., "task": "insert the cable into the control module"})
actions = post(policy.predict_action_chunk(batch))
Read moreDetails
Train loss: 0.002 at step 20,000
Architecture: LeRobot 0.6.2 pi0.5_subtask + lerobot_sirius
Inputs: 3 cameras and 24-dimensional state
Output: 30-step action sequences (11-D)
Discuss Telegram 0 13

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 15 GB · ready files17 GB
  • BF16weights 7.7 GB9.0 GB
  • FP8 / INT8recommendedweights 3.9 GB4.7 GB
  • 4-bitweights 2.2 GB2.9 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 · F32 — 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.

    # chomeed/task1task2_subtask_rel_chunk30_sirius_bs128_multitask_fromR2mixed_1118_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 chomeed/task1task2_subtask_rel_chunk30_sirius_bs128_multitask_fromR2mixed_1118_20k --local-dir ~/models/task1task2_subtask_rel_chunk30_sirius_bs128_multitask_fromR2mixed_1118_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

  1. paligemma-3b-pt-224family root
  2. pi05_base
  3. task1task2_subtask_rel_chunk30_sirius_bs128_multitask_fromR2mixed_1118_20k

New posts about versions of paligemma-3b-pt-224 (quants, finetunes, LoRAs) will show up under the bell.

Fine-tunes 12All on Hugging Face
LoRAs and adapters 12All on Hugging Face

About the model

chomeed/task1task2_subtask_rel_chunk30_sirius_bs128_multitask_fromR2mixed_1118_20k
Author
chomeed
Category
Robotics
Released
Posted

On NeyroLearn

Hugging Face

130

Ask AI about this model

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

Comments