community_dataset_v3_v3

HuggingFaceVLA/community_dataset_v3 with every sub-dataset migrated to the LeRobot v3.0 dataset format, so it loads directly with lerobot ≥ 0.4 (LeRobotDataset(repo_id, root=...)) without per-dataset conversion.

Conversion is fully reproducible with lerobot-dataset-tools; the exact per-dataset log is in conversion_manifest.jsonl.

Usage

This is an umbrella repo: every <user>/<dataset> subtree is an independent LeRobotDataset (hence no dataset viewer). Download a subtree and point lerobot at it:

from huggingface_hub import snapshot_download
from lerobot.datasets.lerobot_dataset import LeRobotDataset

name = "1g0rrr/demo2_frame_holder"  # any sub-dataset in this repo
root = snapshot_download(
    "mcobzarenco/community_dataset_v3_v3",
    repo_type="dataset",
    allow_patterns=[f"{name}/**"],
    local_dir="./data",
)
dataset = LeRobotDataset(name, root=f"./data/{name}")

or with the CLI:

hf download mcobzarenco/community_dataset_v3_v3 --repo-type=dataset --include "1g0rrr/demo2_frame_holder/**" --local-dir ./data

Statistics

	source	converted
sub-datasets	791 (791× v2.1)	791 (all v3.0)
episodes	50,622	50,614
frames	25,971,082	25,941,870
size	757.7 GB	737.3 GB

What the conversion does

Per sub-dataset (see the tools repo for code):

1. Copy the dataset, skipping stray duplicate video trees and *.bak files.
2. Drop parquet columns not declared in info.json features (legacy leftovers such as next.done break v3.0 schema casting).
3. Repair the collection-wide stats-key bug (stats keyed observation.<cam> while features are observation.images.<cam>).
4. For v2.0 sources: synthesize meta/episodes_stats.jsonl (numeric stats exact from parquet; per-episode image stats from ~100 sampled frames per camera, matching lerobot's own conventions) and bump to v2.1.
5. Run the official lerobot.scripts.convert_dataset_v21_to_v30 converter (with an ffmpeg concat fallback for episodes whose packets carry duplicate/non-monotonic DTS, which PyAV refuses to mux).
6. Validate: reload with LeRobotDataset, check episode/frame counts, and verify per-episode length == index-range == video-span × fps per camera.

Reproduce

git clone https://github.com/mcobzarenco/lerobot-dataset-tools && cd lerobot-dataset-tools
uv run hf download HuggingFaceVLA/community_dataset_v3 --repo-type=dataset --local-dir ./source
uv run python -m ldtools.convert_collection \
    --source ./source --output ./converted --workers 12
uv run python -m ldtools.dataset_card \
    --source ./source --output ./converted \
    --source-repo HuggingFaceVLA/community_dataset_v3 --target-repo mcobzarenco/community_dataset_v3_v3 --write

Source-data corrections

Some source datasets ship metadata that contradicts their own data (e.g. inflated total_frames counters, which make LeRobotDataset.__len__ overreport and crash shuffled samplers with IndexError). The converter recomputes every aggregate from the actual episode table and records each correction in the manifest. Corrections applied in this collection:

dataset	correction
danaaubakirova/svla_so100_task4_v3_clean	splits recomputed {'train': '0:56'} -> {'train': '0:50'}
dopaul/1500_chess_moves	dropped v3-incompatible features ['observation.images.image2']
kumarhans/ethernetGrab01	splits recomputed {'train': '0:47'} -> {'train': '0:48'}
kumarhans/ethernetGrab01234	splits recomputed {'train': '0:100'} -> {'train': '0:101'}
kumarhans/ethernetGrab01234567	splits recomputed {'train': '0:151'} -> {'train': '0:152'}
kumarhans/ethernetGrab23	splits recomputed {'train': '0:27'} -> {'train': '0:28'}
kumarhans/ethernetInsertMerged063025	splits recomputed {'train': '0:51'} -> {'train': '0:52'}
kumarhans/ethernetInsertMerged063025Processed	splits recomputed {'train': '0:51'} -> {'train': '0:52'}
kumarhans/ethernetInsertMerged070725	splits recomputed {'train': '0:123'} -> {'train': '0:124'}
siqiLi/so100_test15	splits recomputed {'train': '0:44'} -> {'train': '0:45'}
siqiLi/so100_test2	splits recomputed {'train': '0:44'} -> {'train': '0:45'}
villekuosmanen/dAgger_coffee_prop	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/dAgger_coffee_prop_2.1.0	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/dAgger_coffee_prop_2.3.0	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/dAgger_coffee_prop_2.4.0	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/dAgger_pack_easter_eggs_into_basket_2.0.1	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/dAgger_plus_coffee_prop	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/fold_clothes_dining	splits recomputed {'train': '0:40'} -> {'train': '0:38'}
villekuosmanen/fold_teat_towel_desk	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/insert_paper_towel	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/kitchen_interactions	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/lift_cup_remove_capsule_	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/manipulate_glasses	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/measure_cat_food	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/move_glass_cups	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/move_objects_multitask	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/open_close_zipper_bags	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/open_snackbar_box	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/pack_backpack	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/pack_easter_eggs_into_basket	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/pack_electric_toothbrush	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/pack_toothbrush	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/pair_up_socks	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/pair_up_socks_dining_table	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/pass_laser_pointer	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/pick_1_snackbar	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/pick_2_snackbars	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/pick_3_snackbars	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/pick_coffee_capsule_under_dome	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/pick_coffee_prop_20Hz	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/pick_coffee_prop_20Hz_3	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/pick_coffee_prop_20Hz_alt	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/rice-cooker_step_0	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/rice-cooker_step_1_truncated_x	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/rice-cooker_step_2	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/rice-cooker_step_3	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/sort_coloured_pins	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/spread_snackbar_box_flaps	dropped v3-incompatible features ['subtask', 'expanded_subtasks']
villekuosmanen/stack_pepsi_cans	dropped v3-incompatible features ['subtask', 'expanded_subtasks']

Remaining known quirk: a handful of episodes across the collections have a few trailing video frames beyond their declared length (source recording artifacts, not repairable without re-encoding). They load, decode and train normally — frames are indexed via episode metadata, so the tail frames are simply never read.

Provenance & license

All data originates from HuggingFaceVLA/community_dataset_v3 (community-contributed robot demonstrations, Apache-2.0). This repo changes the storage format only: actions, states, task strings and video content are unmodified (videos are stream-copied; v2.0 per-episode image stats are sampled approximations).

Generated 2026-07-24 by lerobot-dataset-tools.