misc 0 0 0

linkenv/humanoidmimicgen-g1-push-button-action-replay

published by @linkenv

Alignment dataset tracked in the /datasets sub-catalog.

Lifetime downloads
111
Last 30 days
47
Likes
-
Size
92 KB
Created on HF
2026-09-09
Age
1 months ago

Description

HumanoidMimicGen G1 Push Button Action Replay Fixture

This is a small public validation fixture for the LMPushButton environment in

HumanoidMimicGen. It contains 50

recorded G1 low-level actions and MuJoCo states at 50 Hz. It is intended to test

recorded-action physics replay, not policy training or benchmark success.

The fixture contains no images, checkpoint weights, internal repository code,

machine paths, or service addresses.

Replay

hf download… See the full description on the dataset page: https://huggingface.co/datasets/linkenv/humanoidmimicgen-g1-push-button-action-replay.

Tags

license:otherregion:usroboticshumanoidmujocolerobot

README current version from Hugging Face


license: other
pretty_name: HumanoidMimicGen G1 Push Button Action Replay Fixture
tags:

  • robotics
  • humanoid
  • mujoco
  • lerobot

HumanoidMimicGen G1 Push Button Action Replay Fixture

This is a small public validation fixture for the LMPushButton environment in
HumanoidMimicGen. It contains 50
recorded G1 low-level actions and MuJoCo states at 50 Hz. It is intended to test
recorded-action physics replay, not policy training or benchmark success.

The fixture contains no images, checkpoint weights, internal repository code,
machine paths, or service addresses.

Replay

hf download linkenv/humanoidmimicgen-g1-push-button-action-replay --repo-type dataset --local-dir /tmp/hmg-push-button-replay

git clone https://github.com/NVlabs/humanoidmimicgen.git
cd humanoidmimicgen
python -m pip install -e ".[wbc-replay]"
python scripts/playback_dataset.py /tmp/hmg-push-button-replay \
  --action-source recorded --num-episodes 1 --state-atol 0

The player restores only the initial state. It then applies each recorded
low-level joint action and base command without replacing later states, and
compares every resulting MuJoCo state with the next recording.

Provenance and scope

  • Environment: LMPushButton
  • Robot: G1
  • Runtime: MuJoCo 3.2.6, RoboSuite 1.5.1
  • Source HMG revision: 4d84ac37deda3e2720ea5e0774d64d426fe5909f
  • 50 rows, 49 checked state transitions
  • No task-success claim; this is a physics replay fixture

Dataset use is subject to the licenses and third-party notices in the
HumanoidMimicGen repository.

README history 2 revisions

Every night we snapshot the README of every dataset in the catalog. When the SHA changes we archive the new version and diff it against the last. This is the evolving thought record of the alignment-data field: what the author decided to say about the corpus, and how that framing shifted over time.

  1. 2026-09-09Add files using upload-large-folder tool7ab1e4f1.6 KB
    Loading...
  2. 2026-09-09initial commit70bca140 B
    Loading...

Used by abliterated models none yet

No abliterated models in our catalog have declared this dataset in their YAML metadata yet. This can mean the dataset is used but not declared, is used outside abliteration workflows, or is new. The link will populate automatically as models are indexed.

Related datasets same stage · misc

Read further

Metadata

License
other

Download the dataset

0

Copy any of these snippets into your notebook or terminal. All three fetch directly from Hugging Face using your own credentials.

Recommended: datasets library (Python)
from datasets import load_dataset
ds = load_dataset("linkenv/humanoidmimicgen-g1-push-button-action-replay")
Raw snapshot (Python)
from huggingface_hub import snapshot_download
snapshot_download(repo_id="linkenv/humanoidmimicgen-g1-push-button-action-replay", repo_type="dataset")
git clone (requires git-lfs)
git clone https://huggingface.co/datasets/linkenv/humanoidmimicgen-g1-push-button-action-replay
Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Open in Abliteration