Skip to content
Robotics

act_stack_c50_h200_80k

robotic arm control policy

2 min readupdated

This model is trained with the ACT algorithm (Action Chunking with Transformers, an imitation learning method that predicts short action sequences instead of single steps) and the LeRobot library to control an so_follower robot. Its task is to stack red, yellow, and purple cubes onto a designated spot on a table. The policy takes the robot's state and high-resolution wrist camera feed as input to output control signals. Trained on a 74-episode dataset, it is aimed at robotics researchers working with hardware compatible with the LeRobot ecosystem.

Key features

  • Robotics: Built for a physical so_follower robotic arm.
  • Computer vision: Uses 640x480 wrist camera input (`observation.images.wrist`).
  • ACT architecture: Predicts action chunks via imitation learning.
  • Dataset: Trained on 74 episodes (over 100k frames) for color-coded cube stacking.

What you can do

  • Run on a robot: Execute the cube-stacking task using the `lerobot-rollout` command.
  • Study the training setup: Reuse the configuration to train your own manipulation policies.
  • Inspect the dataset: Explore the recorded demonstrations via Hugging Face Spaces visualizers.

Example

lerobot-rollout \
  --strategy.type=base \
  --robot.type=so_follower \
  --robot.port= \
  --robot.cameras="{ : {type: opencv, index_or_path: , width: 640, height: 480, fps: 30}}" \
  --policy.path=asma-4a/act_stack_c50_h200_80k \
  --task="Stack the red, yellow and purple cubes on the table mark" \
  --duration=60
Read moreDetails
Training steps: 80000
Batch size: 32
Optimizer: adamw (learning rate: 1e-05)
LeRobot version: 0.7.1
Input features: observation.state (6,) and observation.images.wrist (3, 640, 480)
Discuss Telegram 0 11

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.

    # asma-4a/act_stack_c50_h200_80k
    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 asma-4a/act_stack_c50_h200_80k --local-dir ~/models/act_stack_c50_h200_80k
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to act_stack_c50_h200_80k
  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_stack_c50_h200_80k (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

asma-4a/act_stack_c50_h200_80k
Author
asma-4a
Category
Robotics
Released
Posted

On NeyroLearn

Hugging Face

110

Ask AI about this model

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

Comments