Skip to content
Robotics

ACT policy for SO-101 robotic arm - picking and placing a cube

2 min read

The model implements the ACT (Action Chunking with Transformers, a robot control algorithm) policy based on the LeRobot library for the SO-101 robotic arm. It is trained to pick a green cube from a fixed table position and place it into a box, handling scenarios where other cubes are already inside the box. The model accepts video input from two cameras (top-down and wrist-mounted) along with current joint states, and outputs a sequence of target positions for the robot's 6 joints.

Key features

  • Training for SO-101: the model is built to control the specific manipulator model via the LeRobot framework.
  • Multimodal input: uses data from two cameras (640x480 at 30 fps) and 6-joint parameters.
  • Fine-tuned scenarios: trained on demonstrations with cubes already present inside the target box.
  • Hardware integration: runs using standard rollout utility tools for real robotics tests.

What you can do

  • Run on a robotic arm: apply the policy to automate object picking on compatible SO-101 hardware.
  • Test in simulation: verify the playback of recorded manipulator movement trajectories.
  • Integrate into a setup: use alongside external color detectors to implement conditional sorting scenarios.

Example

lerobot-rollout --strategy.type=base --policy.path=MaxPrain/act_pick_place_green_v3_40000 --device=mps \
  --robot.type=so101_follower --robot.port= --robot.id= --robot.max_relative_target=10 \
  --robot.cameras="{ top: {type: opencv, index_or_path: 1, width: 640, height: 480, fps: 30}, wrist: {type: opencv, index_or_path: 0, width: 640, height: 480, fps: 30} }" \
  --task="pick the green cube and place it in the box" --fps=30 --duration=12
Read moreDetails
Used the LeRobot 0.7.0 framework and the MaxPrain/pick_place_green_v3 dataset (30 episodes).
Training took 40,000 steps via Hugging Face Jobs (a10g-small).
Important: the model reproduces a fixed coordinate-based trajectory and does not recognize color directly.
Discuss Telegram 0 2

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 0.2 GB · ready files0.7 GB
  • BF16weights 0.1 GB0.6 GB
  • FP8 / INT8recommendedweights 0.0 GB0.6 GB
  • 4-bitweights 0.0 GB0.5 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 2060

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.

    # MaxPrain/act_pick_place_green_v3_40000
    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 MaxPrain/act_pick_place_green_v3_40000 --local-dir ~/models/act_pick_place_green_v3_40000
    # 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 act_pick_place_green_v3_40000 (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

MaxPrain/act_pick_place_green_v3_40000
Author
MaxPrain
Category
Robotics
Released
Posted

On NeyroLearn

Hugging Face

20

Ask AI about this model

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

Comments