Skip to content
Robotics

microduck-sock-heist-composite-v10

updated strategy for the Microduck Arena robotics simulation

2 min readupdated

This competitive checkpoint for the 6m Sock Heist simulation in Microduck Arena combines winning trajectories from other authors with targeted optimization of a key step. The model clocks an average runtime of 4.780 seconds across five official seeds, beating previous benchmarks by applying CMA-ES (Covariance Matrix Adaptation Evolution Strategy) search to the fourth seed. The remaining four trajectories are taken from the nov17-v3 model by HlibPlakhtii. This setup is built strictly for the x86-64 simulator and is not meant to be deployed on physical robots.

Key features

  • Trajectory optimization: cuts seed 4 runtime down to 4.680 s using CMA-ES search.
  • Hybrid setup: pairs four trajectories from nov17-v3 with one newly optimized trajectory.
  • Reliable execution: clears all five official simulator seeds cleanly.

What you can do

  • Run the simulation in Microduck Arena on an x86-64 machine to benchmark completion times.
  • Inspect the recorded trajectories and lookup tables to look for further optimization opportunities.
  • Use this configuration as a baseline to push for better scores in the robotics competition.

Example

Because the model consists of recorded trajectories and lookup tables for the simulator, standard inference code is replaced by direct evaluation in the Arena engine. Per-seed results:
Seed 1: 4.635 | Seed 2: 5.095 | Seed 3: 4.675 | Seed 4: 4.680 | Seed 5: 4.815

Read moreDetails
Average completion time: 4.780 seconds across 5 seeds.
Hardware and environment requirements: x86-64 CPU, pinned versions of MuJoCo 3.10.0, onnxruntime 1.24.4, and numpy 2.4.1.
Limitations: runs in the simulator only; results on Mac or other hardware architectures may vary.
License: Apache-2.0 applies only to the modified parts and markup.
Discuss Telegram 0 17

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.0 GB · ready files0.5 GB
  • FP8 / INT8recommendedweights 0.0 GB0.5 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 · BF16 — 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.

    # rumezhak/microduck-sock-heist-composite-v10
    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 rumezhak/microduck-sock-heist-composite-v10 --local-dir ~/models/microduck-sock-heist-composite-v10
    # 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. microduck-sock-heist-composite-v3family root
  2. microduck-sock-heist-composite-v4
  3. microduck-sock-heist-composite-v5
  4. microduck-sock-heist-composite-v9
  5. microduck-sock-heist-composite-v10

New posts about versions of microduck-sock-heist-composite-v3 (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

rumezhak/microduck-sock-heist-composite-v10
Author
rumezhak
Category
Robotics
Released
Posted

On NeyroLearn

Hugging Face

170

Ask AI about this model

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

Comments