Skip to content
Robotics

HRL-Group/stack_100

policy for robotic manipulators

2 min readupdated

This is a ready-to-use policy (control model) for robotics, trained using the LeRobot framework on a given dataset. The model uses the ACT method (Action Chunking with Transformers — an imitation learning approach that predicts short sequences of actions instead of individual steps) to perform block manipulation tasks. It takes robot state data and an image from a wrist camera as inputs, and outputs control commands for the manipulator. The project targets researchers and developers working with physical robots like so_follower.

Key features

  • Trained via the ACT method for efficient action sequence prediction
  • Uses visual data from a wrist camera and robot joint state data
  • Designed to work with the so_follower hardware configuration
  • Based on a dataset of 92 episodes (over 218,000 frames at 30 FPS)

What you can do

  • Run the ready-to-use control policy on a real robot using LeRobot utilities
  • Fine-tune your own model on a similar dataset for new manipulation tasks
  • Integrate the model into custom robotic scenarios involving object grasping and movement

Example

Running the ready-to-use policy on a robot via the command line:

lerobot-rollout \
  --strategy.type=base \
  --robot.type=so_follower \
  --robot.port= \
  --robot.cameras="{ : {type: opencv, index_or_path: , width: 640, height: 480, fps: 30}, : {type: opencv, index_or_path: , width: 640, height: 480, fps: 30}}" \
  --policy.path=HRL-Group/stack_100 \
  --task="Pick up one cube and stack it on top of the other cube" \
  --duration=60
Read moreDetails
Training parameters: 100,000 steps, batch size 16, adamw optimizer, learning rate 1e-05.
Inputs: observation.state with shape (6,) and observation.images.wrist with shape (3, 480, 640).
Outputs: action with shape (6,).
Dataset: 92 episodes, frame rate 30 FPS.
Discuss Telegram 0

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.

    # HRL-Group/stack_100
    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 HRL-Group/stack_100 --local-dir ~/models/stack_100
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to stack_100
  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 stack_100 (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

HRL-Group/stack_100
Author
HRL-Group
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