tsmithv11 commited on
Commit
896ebdb
·
verified ·
1 Parent(s): 52bfaac

Publish groot-x-embodiment-sim in Lance format

Browse files

Includes the source splits, attribution, vector search example, and browser preview.

Files changed (29) hide show
  1. .gitattributes +8 -0
  2. README.md +69 -135
  3. lance/assets.lance/_indices/8bdc4fb8-2641-4a3a-9e34-94db794bec23/page_data.lance +3 -0
  4. lance/assets.lance/_indices/8bdc4fb8-2641-4a3a-9e34-94db794bec23/page_lookup.lance +0 -0
  5. lance/assets.lance/_transactions/0-1ea39d82-98e5-4589-aee2-de89ac6405d8.txn +0 -0
  6. lance/assets.lance/_transactions/1-184e0ee1-69a7-43f0-b633-18c5758dcdf2.txn +0 -0
  7. lance/assets.lance/_versions/18446744073709551613.manifest +0 -0
  8. lance/assets.lance/_versions/18446744073709551614.manifest +0 -0
  9. lance/assets.lance/_versions/latest_version_hint.json +1 -0
  10. lance/assets.lance/data/0001000100000010011100118dadea49a9908075eab1cf8255.lance +3 -0
  11. lance/assets.lance/data/101010010000001100101101a7fe4142eea2cdb9c542fdfa0f.lance +3 -0
  12. lance/assets.lance/data/10110010010001111101100098ad0f41a8879d4a6f8f3f20bf.lance +3 -0
  13. lance/assets.lance/data/11111001011101001010001196b58e4f43901edcf2e747ba22.lance +3 -0
  14. lance/episodes.lance/_indices/c63eab46-752b-4f8f-b72e-18af51966883/bitmap_page_lookup.lance +3 -0
  15. lance/episodes.lance/_transactions/0-052be1df-46ee-4e55-baf6-e26f22a99c26.txn +0 -0
  16. lance/episodes.lance/_transactions/1-e47e2681-2049-4bea-89cd-f27c89f579c3.txn +0 -0
  17. lance/episodes.lance/_versions/18446744073709551613.manifest +0 -0
  18. lance/episodes.lance/_versions/18446744073709551614.manifest +0 -0
  19. lance/episodes.lance/_versions/latest_version_hint.json +1 -0
  20. lance/episodes.lance/data/111000000010001101100111eda73a4c94b055dcb02234099f.lance +3 -0
  21. lance/release.json +135 -0
  22. lance/subsets.lance/_transactions/0-fcee01c7-c797-4430-bf22-3c582f2d0578.txn +0 -0
  23. lance/subsets.lance/_versions/18446744073709551614.manifest +0 -0
  24. lance/subsets.lance/_versions/latest_version_hint.json +1 -0
  25. lance/subsets.lance/data/111111000010011010011011453e994f738b9093f4cf9cc588.lance +3 -0
  26. source/BEHAVIOR-2025-README.md +319 -0
  27. source/README.md +149 -0
  28. viewer/episodes.parquet +3 -0
  29. viewer/subsets.parquet +3 -0
.gitattributes CHANGED
@@ -105,3 +105,11 @@ single_panda_gripper.OpenDoubleDoor/videos/chunk-000/observation.images.left_vie
105
  single_panda_gripper.OpenDoubleDoor/videos/chunk-000/observation.images.left_view/episode_000757.mp41875097921 filter=lfs diff=lfs merge=lfs -text
106
  single_panda_gripper.OpenDoubleDoor/videos/chunk-000/observation.images.left_view/episode_000901.mp43884489326 filter=lfs diff=lfs merge=lfs -text
107
  single_panda_gripper.OpenDoubleDoor/videos/chunk-000/observation.images.left_view/episode_000485.mp42911758261 filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
 
 
105
  single_panda_gripper.OpenDoubleDoor/videos/chunk-000/observation.images.left_view/episode_000757.mp41875097921 filter=lfs diff=lfs merge=lfs -text
106
  single_panda_gripper.OpenDoubleDoor/videos/chunk-000/observation.images.left_view/episode_000901.mp43884489326 filter=lfs diff=lfs merge=lfs -text
107
  single_panda_gripper.OpenDoubleDoor/videos/chunk-000/observation.images.left_view/episode_000485.mp42911758261 filter=lfs diff=lfs merge=lfs -text
108
+ lance/assets.lance/_indices/8bdc4fb8-2641-4a3a-9e34-94db794bec23/page_data.lance filter=lfs diff=lfs merge=lfs -text
109
+ lance/assets.lance/data/0001000100000010011100118dadea49a9908075eab1cf8255.lance filter=lfs diff=lfs merge=lfs -text
110
+ lance/assets.lance/data/101010010000001100101101a7fe4142eea2cdb9c542fdfa0f.lance filter=lfs diff=lfs merge=lfs -text
111
+ lance/assets.lance/data/10110010010001111101100098ad0f41a8879d4a6f8f3f20bf.lance filter=lfs diff=lfs merge=lfs -text
112
+ lance/assets.lance/data/11111001011101001010001196b58e4f43901edcf2e747ba22.lance filter=lfs diff=lfs merge=lfs -text
113
+ lance/episodes.lance/_indices/c63eab46-752b-4f8f-b72e-18af51966883/bitmap_page_lookup.lance filter=lfs diff=lfs merge=lfs -text
114
+ lance/episodes.lance/data/111000000010001101100111eda73a4c94b055dcb02234099f.lance filter=lfs diff=lfs merge=lfs -text
115
+ lance/subsets.lance/data/111111000010011010011011453e994f738b9093f4cf9cc588.lance filter=lfs diff=lfs merge=lfs -text
README.md CHANGED
@@ -1,149 +1,83 @@
1
  ---
2
  license: cc-by-4.0
 
 
 
3
  task_categories:
4
  - robotics
5
  tags:
6
- - robotics
 
 
 
 
 
 
 
 
 
 
 
 
 
7
  ---
8
- # PhysicalAI-Robotics-GR00T-X-Embodiment-Sim
9
-
10
- ![image/png](https://cdn-uploads.huggingface.co/production/uploads/67b8da81d01134f89899b4a7/PppFm7zvXyJblkMOq-YRQ.png)
11
-
12
- Github Repo: [Isaac GR00T N1](https://github.com/NVIDIA/Isaac-GR00T)
13
-
14
- We provide a set of datasets used for post-training of GR00T N1. Each dataset is a collection of trajectories from different robot embodiments and tasks.
15
-
16
- ### Cross-embodied bimanual manipulation: 9k trajectories
17
-
18
- | Dataset Name | #trajectories |
19
- | - | -|
20
- | bimanual_panda_gripper.Threading | 1000 |
21
- | bimanual_panda_hand.LiftTray | 1000 |
22
- | bimanual_panda_gripper.ThreePieceAssembly | 1000 |
23
- | bimanual_panda_gripper.Transport | 1000 |
24
- | bimanual_panda_hand.BoxCleanup | 1000 |
25
- | bimanual_panda_hand.DrawerCleanup | 1000 |
26
- | gr1_arms_only.CanSort | 1000 |
27
- | gr1_full_upper_body.Coffee | 1000 |
28
- | gr1_full_upper_body.Pouring | 1000 |
29
-
30
-
31
- ### Humanoid robot tabletop manipulation: 240k trajectories
32
-
33
- | Dataset Name | #trajectories |
34
- | - | -|
35
- | gr1_arms_waist.CanToDrawer | 10000 |
36
- | gr1_arms_waist.CupToDrawer | 10000 |
37
- | gr1_arms_waist.CuttingboardToBasket | 10000 |
38
- | gr1_arms_waist.CuttingboardToCardboardBox | 10000 |
39
- | gr1_arms_waist.CuttingboardToPan | 10000 |
40
- | gr1_arms_waist.CuttingboardToPot | 10000 |
41
- | gr1_arms_waist.CuttingboardToTieredBasket | 10000 |
42
- | gr1_arms_waist.PlaceBottleToCabinet | 10000 |
43
- | gr1_arms_waist.PlaceMilkToMicrowave | 10000 |
44
- | gr1_arms_waist.PlacematToBasket | 10000 |
45
- | gr1_arms_waist.PlacematToBowl | 10000 |
46
- | gr1_arms_waist.PlacematToPlate | 10000 |
47
- | gr1_arms_waist.PlacematToTieredShelf | 10000 |
48
- | gr1_arms_waist.PlateToBowl | 10000 |
49
- | gr1_arms_waist.PlateToCardboardBox | 10000 |
50
- | gr1_arms_waist.PlateToPan | 10000 |
51
- | gr1_arms_waist.PlateToPlate | 10000 |
52
- | gr1_arms_waist.PotatoToMicrowave | 10000 |
53
- | gr1_arms_waist.TrayToCardboardBox | 10000 |
54
- | gr1_arms_waist.TrayToPlate | 10000 |
55
- | gr1_arms_waist.TrayToPot | 10000 |
56
- | gr1_arms_waist.TrayToTieredBasket | 10000 |
57
- | gr1_arms_waist.TrayToTieredShelf | 10000 |
58
- | gr1_arms_waist.WineToCabinet | 10000 |
59
-
60
- ### Humanoid robot tabletop manipulation - downsampled: 24k trajectories
61
-
62
- | Dataset Name | #trajectories |
63
- | - | - |
64
- | gr1_unified.PnPBottleToCabinetClose_GR1ArmsAndWaistFourierHands_1000 | 1000 |
65
- | gr1_unified.PnPCanToDrawerClose_GR1ArmsAndWaistFourierHands_1000 | 1000 |
66
- | gr1_unified.PnPCupToDrawerClose_GR1ArmsAndWaistFourierHands_1000 | 1000 |
67
- | gr1_unified.PnPMilkToMicrowaveClose_GR1ArmsAndWaistFourierHands_1000 | 1000 |
68
- | gr1_unified.PnPPotatoToMicrowaveClose_GR1ArmsAndWaistFourierHands_1000 | 1000 |
69
- | gr1_unified.PnPWineToCabinetClose_GR1ArmsAndWaistFourierHands_1000 | 1000 |
70
- | gr1_unified.PosttrainPnPNovelFromCuttingboardToBasketSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
71
- | gr1_unified.PosttrainPnPNovelFromCuttingboardToCardboardboxSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
72
- | gr1_unified.PosttrainPnPNovelFromCuttingboardToPanSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
73
- | gr1_unified.PosttrainPnPNovelFromCuttingboardToPotSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
74
- | gr1_unified.PosttrainPnPNovelFromCuttingboardToTieredbasketSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
75
- | gr1_unified.PosttrainPnPNovelFromPlacematToBasketSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
76
- | gr1_unified.PosttrainPnPNovelFromPlacematToBowlSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
77
- | gr1_unified.PosttrainPnPNovelFromPlacematToPlateSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
78
- | gr1_unified.PosttrainPnPNovelFromPlacematToTieredshelfSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
79
- | gr1_unified.PosttrainPnPNovelFromPlateToBowlSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
80
- | gr1_unified.PosttrainPnPNovelFromPlateToCardboardboxSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
81
- | gr1_unified.PosttrainPnPNovelFromPlateToPanSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
82
- | gr1_unified.PosttrainPnPNovelFromPlateToPlateSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
83
- | gr1_unified.PosttrainPnPNovelFromTrayToCardboardboxSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
84
- | gr1_unified.PosttrainPnPNovelFromTrayToPlateSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
85
- | gr1_unified.PosttrainPnPNovelFromTrayToPotSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
86
- | gr1_unified.PosttrainPnPNovelFromTrayToTieredbasketSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
87
- | gr1_unified.PosttrainPnPNovelFromTrayToTieredshelfSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
88
-
89
- ### Robot Arm Kitchen Manipulation: 72K trajectories
90
- | Dataset Name | #trajectories |
91
- | - | -|
92
- | single_panda_gripper.CloseDoubleDoor | 3000 |
93
- | single_panda_gripper.CloseDrawer | 3000 |
94
- | single_panda_gripper.CloseSingleDoor | 3000 |
95
- | single_panda_gripper.CoffeePressButton | 3000 |
96
- | single_panda_gripper.CoffeeServeMug | 3000 |
97
- | single_panda_gripper.CoffeeSetupMug | 3000 |
98
- | single_panda_gripper.OpenDoubleDoor | 3000 |
99
- | single_panda_gripper.OpenDrawer | 3000 |
100
- | single_panda_gripper.OpenSingleDoor | 3000 |
101
- | single_panda_gripper.PnPCabToCounter | 3000 |
102
- | single_panda_gripper.PnPCounterToCab | 3000 |
103
- | single_panda_gripper.PnPCounterToMicrowave | 3000 |
104
- | single_panda_gripper.PnPCounterToSink | 3000 |
105
- | single_panda_gripper.PnPCounterToStove | 3000 |
106
- | single_panda_gripper.PnPMicrowaveToCounter | 3000 |
107
- | single_panda_gripper.PnPSinkToCounter | 3000 |
108
- | single_panda_gripper.PnPStoveToCounter | 3000 |
109
- | single_panda_gripper.TurnOffMicrowave | 3000 |
110
- | single_panda_gripper.TurnOffSinkFaucet | 3000 |
111
- | single_panda_gripper.TurnOffStove | 3000 |
112
- | single_panda_gripper.TurnOnMicrowave | 3000 |
113
- | single_panda_gripper.TurnOnSinkFaucet | 3000 |
114
- | single_panda_gripper.TurnOnStove | 3000 |
115
- | single_panda_gripper.TurnSinkSpout | 3000 |
116
-
117
- ### Unitree G1 Loco-Manipulation: 102 trajectories
118
- | Dataset Name | #trajectories |
119
- | - | -|
120
- | unitree_g1.LMPnPAppleToPlateDC | 102 |
121
-
122
- ## Download the dataset from Huggingface
123
-
124
- Users can download a specific subset of data by specifying the dataset name.
125
-
126
- 1. Option 1: with `huggingface-cli`
127
 
128
- ```bash
129
- huggingface-cli download nvidia/PhysicalAI-Robotics-GR00T-X-Embodiment-Sim \
130
- --repo-type dataset \
131
- --include "gr1_arms_only.CanSort/**" \
132
- --local-dir $HOME/gr00t_dataset
133
- ```
 
 
 
 
 
134
 
135
- **Replace `gr1_arms_only.CanSort/**` with the dataset name you want to download.**
136
 
137
- 2. Option 2: Github LFS
 
 
 
 
138
 
139
  ```bash
140
- # Clone the repo without downloading files
141
- git clone --filter=blob:none --no-checkout https://huggingface.co/datasets/nvidia/PhysicalAI-Robotics-GR00T-X-Embodiment-Sim
142
- cd PhysicalAI-Robotics-GR00T-X-Embodiment-Sim
143
 
144
- # Configure sparse-checkout for just your folder
145
- git sparse-checkout init --cone
146
- git sparse-checkout set gr1_arms_waist.CupToDrawer << Select the specific dataset to download
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
147
  ```
148
 
149
- Replace the "gr1_arms_waist.CupToDrawer" with the intended dataset folder
 
 
 
 
 
 
 
1
  ---
2
  license: cc-by-4.0
3
+ pretty_name: GR00T X-Embodiment Sim in Lance
4
+ language:
5
+ - en
6
  task_categories:
7
  - robotics
8
  tags:
9
+ - lance
10
+ - lerobot
11
+ - simulation
12
+ - physical-ai
13
+ configs:
14
+ - config_name: subsets
15
+ default: true
16
+ data_files:
17
+ - split: preview
18
+ path: viewer/subsets.parquet
19
+ - config_name: episodes
20
+ data_files:
21
+ - split: preview
22
+ path: viewer/episodes.parquet
23
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
24
 
25
+ # GR00T X-Embodiment Sim in Lance
26
+
27
+ This release converts NVIDIA's GR00T simulation dataset to searchable Lance catalogs. It preserves 132 source subsets across nine robot and task families, including the 50 R1Pro BEHAVIOR subsets present at the pinned revision. All 355,023 original episode metadata records are searchable by subset and episode index. The original frame-level Parquet shards, camera MP4s, and metadata files remain in the same repository at their upstream paths. Lance Blob v2 exposes the original bytes through `lance/assets.lance`. [NVIDIA source dataset at the converted revision](https://huggingface.co/datasets/nvidia/PhysicalAI-Robotics-GR00T-X-Embodiment-Sim/tree/ea7ac0b68f87da62f1e726771bba0fe74300802f)
28
+
29
+ | Lance table | Contents |
30
+ |---|---|
31
+ | `lance/subsets.lance` | 132 source subsets, original task instructions, robot type, episode count, split declarations, and task embeddings |
32
+ | `lance/episodes.lance` | 355,023 original episodes, frame counts, task labels, raw episode metadata, available camera paths, and frame Parquet paths |
33
+ | `lance/assets.lance` | 883,797 original files, recorded sizes and Git or LFS hashes, and native blob ranges into the same repository |
34
+
35
+ `source_split_json` preserves each subset's `meta/info.json` split declaration verbatim as `{"train":"0:100"}`. The conversion preserves original subset boundaries and episode indices. Two paths expected by the metadata are absent upstream, one frame Parquet and one camera MP4. Their file fields stay null while the episode records and expected paths remain present. [Release metadata](lance/release.json)
36
 
37
+ The `viewer/` Parquet files provide a browseable table of all 132 subsets and a separate preview with the first three episodes per subset. The Lance episode table contains the full index. The source metadata and payload remain available in their original LeRobot V2 layouts for existing training consumers.
38
 
39
+ ## Search tasks
40
+
41
+ The release generates one text vector per source subset from its English `meta/tasks.jsonl` instructions. For retrieval, it strips the `unlocked_waist:` prefix, drops empty entries and identifier-only or validation labels, and deduplicates instructions case-insensitively. It encodes each retained text with the `passage: ` prefix, averages the L2-normalized vectors, then normalizes the average. The exact source labels are still available in `source_tasks`, and the retained search inputs are stored in `embedding_inputs`.
42
+
43
+ The encoder is `intfloat/multilingual-e5-small` at revision `614241f622f53c4eeff9890bdc4f31cfecc418b3`. It supports instruction-to-task retrieval using `query: ` for queries. Vectors have **384 float32 dimensions** and use **cosine distance**. The example displays cosine similarity as `1 - distance`. [Pinned embedding model](https://huggingface.co/intfloat/multilingual-e5-small/tree/614241f622f53c4eeff9890bdc4f31cfecc418b3)
44
 
45
  ```bash
46
+ pip install pylance==14.0.3 sentence-transformers==6.3.0
47
+ ```
 
48
 
49
+ ```python
50
+ import lance
51
+ from huggingface_hub import HfApi
52
+ from sentence_transformers import SentenceTransformer
53
+
54
+ repo = "lance-format/groot-x-embodiment-sim-lance"
55
+ root = f"hf://datasets/{repo}"
56
+ revision = HfApi().dataset_info(repo, token=False).sha
57
+ options = {"hf_revision": revision}
58
+ subsets = lance.dataset(f"{root}/lance/subsets.lance", storage_options=options)
59
+
60
+ model = SentenceTransformer(
61
+ "intfloat/multilingual-e5-small",
62
+ revision="614241f622f53c4eeff9890bdc4f31cfecc418b3",
63
+ device="cpu",
64
+ )
65
+ question = "Find robot tasks where an arm picks up bottled water and puts it inside a kitchen cabinet."
66
+ query = model.encode(f"query: {question}", normalize_embeddings=True)
67
+ results = subsets.to_table(
68
+ columns=["subset", "robot_type", "instruction", "episode_count", "_distance"],
69
+ nearest={"column": "task_embedding", "q": query, "k": 5, "metric": "cosine"},
70
+ ).to_pylist()
71
+
72
+ for row in results:
73
+ print(f"{1 - row['_distance']:.4f} {row['subset']} ({row['robot_type']}, "
74
+ f"{row['episode_count']} episodes) {row['instruction']}")
75
  ```
76
 
77
+ Use a result's `subset` to filter `lance/episodes.lance`. Its `frames_path` and camera `source_path` values are keys into `lance/assets.lance`. Open that table with `base_store_params={root: options}` along with `storage_options=options` to read pinned blob ranges with `take_blobs`.
78
+
79
+ ## Source and license
80
+
81
+ Credit NVIDIA and the GR00T X-Embodiment Sim dataset contributors. This conversion retains the source repository's [CC BY 4.0 license](https://creativecommons.org/licenses/by/4.0/) and [original dataset card](source/README.md), and adds the Lance tables, task embeddings, and Parquet previews. The original `.gitattributes` controls Git storage for the mirrored payload. The new root dataset card links to the unchanged upstream copy at `source/README.md`.
82
+
83
+ The R1Pro BEHAVIOR subset families use the BEHAVIOR-1K task benchmark by Chengshu Li and colleagues. Cite the [BEHAVIOR-1K paper](https://arxiv.org/abs/2403.09227) and [MIT-licensed 2025 challenge demonstrations](https://huggingface.co/datasets/behavior-1k/2025-challenge-demos/tree/33639692425296185f245aaa648530178c7b3d2b) when using them, alongside the [NVIDIA Isaac GR00T repository](https://github.com/NVIDIA/Isaac-GR00T). The [original BEHAVIOR dataset card](source/BEHAVIOR-2025-README.md) preserves its MIT declaration and citation.
lance/assets.lance/_indices/8bdc4fb8-2641-4a3a-9e34-94db794bec23/page_data.lance ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:39bc9cf80d0208d9131e7b5f2d49d2c710aa7216d569d39bee4555cff0112d2f
3
+ size 23968137
lance/assets.lance/_indices/8bdc4fb8-2641-4a3a-9e34-94db794bec23/page_lookup.lance ADDED
Binary file (43.1 kB). View file
 
lance/assets.lance/_transactions/0-1ea39d82-98e5-4589-aee2-de89ac6405d8.txn ADDED
Binary file (939 Bytes). View file
 
lance/assets.lance/_transactions/1-184e0ee1-69a7-43f0-b633-18c5758dcdf2.txn ADDED
Binary file (198 Bytes). View file
 
lance/assets.lance/_versions/18446744073709551613.manifest ADDED
Binary file (1.39 kB). View file
 
lance/assets.lance/_versions/18446744073709551614.manifest ADDED
Binary file (1.97 kB). View file
 
lance/assets.lance/_versions/latest_version_hint.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"version":2}
lance/assets.lance/data/0001000100000010011100118dadea49a9908075eab1cf8255.lance ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b5098316cfe95b0110b7255192fcacfbd72450677d9693fb05faf639aeb598dc
3
+ size 45002584
lance/assets.lance/data/101010010000001100101101a7fe4142eea2cdb9c542fdfa0f.lance ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fcf6c69b8de87b372a90f972dbb99d86865cdf131075263c419b2205b6fade73
3
+ size 39765208
lance/assets.lance/data/10110010010001111101100098ad0f41a8879d4a6f8f3f20bf.lance ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:855ea57e05b8e7c89c077b5062556502d04e1df4f6ccd6734aa86b339d3dda88
3
+ size 40117411
lance/assets.lance/data/11111001011101001010001196b58e4f43901edcf2e747ba22.lance ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e7b60c518cc23fc8477d80515f7791a4536716ddc2feae3adf394083cf89968d
3
+ size 21703417
lance/episodes.lance/_indices/c63eab46-752b-4f8f-b72e-18af51966883/bitmap_page_lookup.lance ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:268b94dbf57d11dbc414373280832e1742b78e4da81d3ee7fbd4a403a24f6bab
3
+ size 449660
lance/episodes.lance/_transactions/0-052be1df-46ee-4e55-baf6-e26f22a99c26.txn ADDED
Binary file (709 Bytes). View file
 
lance/episodes.lance/_transactions/1-e47e2681-2049-4bea-89cd-f27c89f579c3.txn ADDED
Binary file (180 Bytes). View file
 
lance/episodes.lance/_versions/18446744073709551613.manifest ADDED
Binary file (1.11 kB). View file
 
lance/episodes.lance/_versions/18446744073709551614.manifest ADDED
Binary file (1.5 kB). View file
 
lance/episodes.lance/_versions/latest_version_hint.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"version":2}
lance/episodes.lance/data/111000000010001101100111eda73a4c94b055dcb02234099f.lance ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ee08ab964528d57c84ab8527d7e30c2117995b4cad9d20db0c46548ec0235495
3
+ size 79964418
lance/release.json ADDED
@@ -0,0 +1,135 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "source_repo": "nvidia/PhysicalAI-Robotics-GR00T-X-Embodiment-Sim",
3
+ "source_revision": "ea7ac0b68f87da62f1e726771bba0fe74300802f",
4
+ "source_license": "cc-by-4.0",
5
+ "subset_count": 132,
6
+ "episode_count": 355023,
7
+ "assets": {
8
+ "files": 883797,
9
+ "bytes": 1913352325228,
10
+ "extensions": {
11
+ ".md": 1,
12
+ ".parquet": 355022,
13
+ ".jsonl": 264,
14
+ ".json": 498,
15
+ ".mp4": 527962,
16
+ ".backup": 2,
17
+ ".mp4199844269": 1,
18
+ ".mp4365425279": 1,
19
+ ".mp41374955119": 1,
20
+ ".mp4215781399": 1,
21
+ ".mp42531598654": 1,
22
+ ".mp41927135579": 1,
23
+ ".mp43103034651": 1,
24
+ ".mp43454667820": 1,
25
+ ".mp43966957131": 1,
26
+ ".mp43833922581": 1,
27
+ ".mp41197474174": 1,
28
+ ".mp43966120224": 1,
29
+ ".mp42241832885": 1,
30
+ ".mp42601844370": 1,
31
+ ".mp42911758261": 1,
32
+ ".mp4317414861": 1,
33
+ ".mp41921342295": 1,
34
+ ".mp42434861555": 1,
35
+ ".mp41101615314": 1,
36
+ ".mp41524419959": 1,
37
+ ".mp41875097921": 1,
38
+ ".mp41510377838": 1,
39
+ ".mp42524313976": 1,
40
+ ".mp43040948088": 1,
41
+ ".mp42105851684": 1,
42
+ ".mp41887828425": 1,
43
+ ".mp42809235912": 1,
44
+ ".mp43858948358": 1,
45
+ ".mp4608237762": 1,
46
+ ".mp43233863005": 1,
47
+ ".mp4468609836": 1,
48
+ ".mp43884489326": 1,
49
+ ".mp4180850661": 1,
50
+ ".mp42651938403": 1,
51
+ ".mp41248192570": 1,
52
+ ".mp43116967891": 1,
53
+ ".mp42037103196": 1,
54
+ ".mp44113730617": 1,
55
+ ".mp4870301939": 1,
56
+ ".mp4913918194": 1,
57
+ ".mp42731591196": 1,
58
+ ".mp44093659278": 1,
59
+ ".mp42220225962": 1,
60
+ ".mp4114851381": 1,
61
+ ".mp42159688174": 1,
62
+ ".mp4699657822": 1,
63
+ ".mp4465068121": 1,
64
+ ".mp44107557110": 1
65
+ }
66
+ },
67
+ "families": {
68
+ "bimanual_panda_gripper": 3,
69
+ "bimanual_panda_hand": 3,
70
+ "gr1_arms_only": 1,
71
+ "gr1_arms_waist": 24,
72
+ "gr1_full_upper_body": 2,
73
+ "gr1_unified": 24,
74
+ "sim_behavior_r1_pro": 50,
75
+ "single_panda_gripper": 24,
76
+ "unitree_g1": 1
77
+ },
78
+ "missing_episode_parquets": 1,
79
+ "missing_expected_videos": 1,
80
+ "source_split_declarations": [
81
+ "{\"train\":\"0:100\"}"
82
+ ],
83
+ "missing_by_subset": {
84
+ "gr1_arms_waist.CuttingboardToTieredBasket": [
85
+ 0,
86
+ 1
87
+ ],
88
+ "single_panda_gripper.PnPCounterToStove": [
89
+ 1,
90
+ 0
91
+ ]
92
+ },
93
+ "embedding_model": "intfloat/multilingual-e5-small",
94
+ "embedding_revision": "614241f622f53c4eeff9890bdc4f31cfecc418b3",
95
+ "embedding_dimensions": 384,
96
+ "retrieval_metric": "cosine",
97
+ "example_query": "Find robot tasks where an arm picks up bottled water and puts it inside a kitchen cabinet.",
98
+ "verified_local_hits": [
99
+ {
100
+ "subset": "gr1_arms_waist.PlaceBottleToCabinet",
101
+ "robot_type": "GR1ArmsAndWaistFourierHands",
102
+ "instruction": "pick up the bottled water, place it into the cabinet and close the cabinet",
103
+ "episode_count": 10032,
104
+ "_distance": 0.12890499830245972
105
+ },
106
+ {
107
+ "subset": "gr1_unified.PnPBottleToCabinetClose_GR1ArmsAndWaistFourierHands_1000",
108
+ "robot_type": "GR1ArmsAndWaistFourierHands",
109
+ "instruction": "pick up the bottled water, place it into the cabinet and close the cabinet",
110
+ "episode_count": 1000,
111
+ "_distance": 0.12890499830245972
112
+ },
113
+ {
114
+ "subset": "single_panda_gripper.PnPCounterToCab",
115
+ "robot_type": "Panda+OmronBase",
116
+ "instruction": "pick the water bottle from the counter and place it in the cabinet",
117
+ "episode_count": 3000,
118
+ "_distance": 0.14401596784591675
119
+ },
120
+ {
121
+ "subset": "single_panda_gripper.OpenSingleDoor",
122
+ "robot_type": "Panda+OmronBase",
123
+ "instruction": "open the microwave door",
124
+ "episode_count": 3000,
125
+ "_distance": 0.14689499139785767
126
+ },
127
+ {
128
+ "subset": "sim_behavior_r1_pro.task-0008_rearranging_kitchen_furniture",
129
+ "robot_type": "R1Pro",
130
+ "instruction": "Rearranging kitchen furniture.",
131
+ "episode_count": 200,
132
+ "_distance": 0.14945638179779053
133
+ }
134
+ ]
135
+ }
lance/subsets.lance/_transactions/0-fcee01c7-c797-4430-bf22-3c582f2d0578.txn ADDED
Binary file (869 Bytes). View file
 
lance/subsets.lance/_versions/18446744073709551614.manifest ADDED
Binary file (1.82 kB). View file
 
lance/subsets.lance/_versions/latest_version_hint.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"version":1}
lance/subsets.lance/data/111111000010011010011011453e994f738b9093f4cf9cc588.lance ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5612e31af686cd1db163e46604db71460734d4da83051887d73d8b4f481984b0
3
+ size 269007
source/BEHAVIOR-2025-README.md ADDED
@@ -0,0 +1,319 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ task_categories:
4
+ - robotics
5
+ tags:
6
+ - LeRobot
7
+ - v
8
+ - '2'
9
+ - .
10
+ - '1'
11
+ configs:
12
+ - config_name: default
13
+ data_files: data/*/*.parquet
14
+ ---
15
+
16
+ This dataset was created using [LeRobot](https://github.com/huggingface/lerobot).
17
+
18
+ ## Dataset Description
19
+
20
+
21
+
22
+ - **Homepage:** [More Information Needed]
23
+ - **Paper:** [More Information Needed]
24
+ - **License:** mit
25
+
26
+ ## Dataset Structure
27
+
28
+ [meta/info.json](meta/info.json):
29
+ ```json
30
+ {
31
+ "codebase_version": "v2.1",
32
+ "robot_type": "R1Pro",
33
+ "total_episodes": 10000,
34
+ "total_frames": 119094660,
35
+ "total_tasks": 50,
36
+ "total_videos": 90000,
37
+ "chunks_size": 10000,
38
+ "fps": 30,
39
+ "splits": {
40
+ "train": "0:10000"
41
+ },
42
+ "data_path": "data/task-{episode_chunk:04d}/episode_{episode_index:08d}.parquet",
43
+ "video_path": "videos/task-{episode_chunk:04d}/{video_key}/episode_{episode_index:08d}.mp4",
44
+ "metainfo_path": "meta/episodes/task-{episode_chunk:04d}/episode_{episode_index:08d}.json",
45
+ "annotation_path": "annotations/task-{episode_chunk:04d}/episode_{episode_index:08d}.json",
46
+ "features": {
47
+ "observation.images.rgb.left_wrist": {
48
+ "dtype": "video",
49
+ "shape": [
50
+ 480,
51
+ 480,
52
+ 3
53
+ ],
54
+ "names": [
55
+ "height",
56
+ "width",
57
+ "rgb"
58
+ ],
59
+ "info": {
60
+ "video.fps": 30.0,
61
+ "video.height": 480,
62
+ "video.width": 480,
63
+ "video.channels": 3,
64
+ "video.codec": "libx265",
65
+ "video.pix_fmt": "yuv420p",
66
+ "video.is_depth_map": false,
67
+ "has_audio": false
68
+ }
69
+ },
70
+ "observation.images.rgb.right_wrist": {
71
+ "dtype": "video",
72
+ "shape": [
73
+ 480,
74
+ 480,
75
+ 3
76
+ ],
77
+ "names": [
78
+ "height",
79
+ "width",
80
+ "rgb"
81
+ ],
82
+ "info": {
83
+ "video.fps": 30.0,
84
+ "video.height": 480,
85
+ "video.width": 480,
86
+ "video.channels": 3,
87
+ "video.codec": "libx265",
88
+ "video.pix_fmt": "yuv420p",
89
+ "video.is_depth_map": false,
90
+ "has_audio": false
91
+ }
92
+ },
93
+ "observation.images.rgb.head": {
94
+ "dtype": "video",
95
+ "shape": [
96
+ 720,
97
+ 720,
98
+ 3
99
+ ],
100
+ "names": [
101
+ "height",
102
+ "width",
103
+ "rgb"
104
+ ],
105
+ "info": {
106
+ "video.fps": 30.0,
107
+ "video.height": 720,
108
+ "video.width": 720,
109
+ "video.channels": 3,
110
+ "video.codec": "libx265",
111
+ "video.pix_fmt": "yuv420p",
112
+ "video.is_depth_map": false,
113
+ "has_audio": false
114
+ }
115
+ },
116
+ "observation.images.depth.left_wrist": {
117
+ "dtype": "video",
118
+ "shape": [
119
+ 480,
120
+ 480,
121
+ 3
122
+ ],
123
+ "names": [
124
+ "height",
125
+ "width",
126
+ "depth"
127
+ ],
128
+ "info": {
129
+ "video.fps": 30.0,
130
+ "video.height": 480,
131
+ "video.width": 480,
132
+ "video.channels": 3,
133
+ "video.codec": "libx265",
134
+ "video.pix_fmt": "yuv420p16le",
135
+ "video.is_depth_map": true,
136
+ "has_audio": false
137
+ }
138
+ },
139
+ "observation.images.depth.right_wrist": {
140
+ "dtype": "video",
141
+ "shape": [
142
+ 480,
143
+ 480,
144
+ 3
145
+ ],
146
+ "names": [
147
+ "height",
148
+ "width",
149
+ "depth"
150
+ ],
151
+ "info": {
152
+ "video.fps": 30.0,
153
+ "video.height": 480,
154
+ "video.width": 480,
155
+ "video.channels": 3,
156
+ "video.codec": "libx265",
157
+ "video.pix_fmt": "yuv420p16le",
158
+ "video.is_depth_map": true,
159
+ "has_audio": false
160
+ }
161
+ },
162
+ "observation.images.depth.head": {
163
+ "dtype": "video",
164
+ "shape": [
165
+ 720,
166
+ 720,
167
+ 3
168
+ ],
169
+ "names": [
170
+ "height",
171
+ "width",
172
+ "depth"
173
+ ],
174
+ "info": {
175
+ "video.fps": 30.0,
176
+ "video.height": 720,
177
+ "video.width": 720,
178
+ "video.channels": 3,
179
+ "video.codec": "libx265",
180
+ "video.pix_fmt": "yuv420p16le",
181
+ "video.is_depth_map": true,
182
+ "has_audio": false
183
+ }
184
+ },
185
+ "observation.images.seg_instance_id.left_wrist": {
186
+ "dtype": "video",
187
+ "shape": [
188
+ 480,
189
+ 480,
190
+ 3
191
+ ],
192
+ "names": [
193
+ "height",
194
+ "width",
195
+ "rgb"
196
+ ],
197
+ "info": {
198
+ "video.fps": 30.0,
199
+ "video.height": 480,
200
+ "video.width": 480,
201
+ "video.channels": 3,
202
+ "video.codec": "libx265",
203
+ "video.pix_fmt": "yuv420p",
204
+ "video.is_depth_map": false,
205
+ "has_audio": false
206
+ }
207
+ },
208
+ "observation.images.seg_instance_id.right_wrist": {
209
+ "dtype": "video",
210
+ "shape": [
211
+ 480,
212
+ 480,
213
+ 3
214
+ ],
215
+ "names": [
216
+ "height",
217
+ "width",
218
+ "rgb"
219
+ ],
220
+ "info": {
221
+ "video.fps": 30.0,
222
+ "video.height": 480,
223
+ "video.width": 480,
224
+ "video.channels": 3,
225
+ "video.codec": "libx265",
226
+ "video.pix_fmt": "yuv420p",
227
+ "video.is_depth_map": false,
228
+ "has_audio": false
229
+ }
230
+ },
231
+ "observation.images.seg_instance_id.head": {
232
+ "dtype": "video",
233
+ "shape": [
234
+ 720,
235
+ 720,
236
+ 3
237
+ ],
238
+ "names": [
239
+ "height",
240
+ "width",
241
+ "rgb"
242
+ ],
243
+ "info": {
244
+ "video.fps": 30.0,
245
+ "video.height": 720,
246
+ "video.width": 720,
247
+ "video.channels": 3,
248
+ "video.codec": "libx265",
249
+ "video.pix_fmt": "yuv420p",
250
+ "video.is_depth_map": false,
251
+ "has_audio": false
252
+ }
253
+ },
254
+ "action": {
255
+ "dtype": "float32",
256
+ "shape": [
257
+ 23
258
+ ],
259
+ "names": null
260
+ },
261
+ "timestamp": {
262
+ "dtype": "float64",
263
+ "shape": [
264
+ 1
265
+ ],
266
+ "names": null
267
+ },
268
+ "episode_index": {
269
+ "dtype": "int64",
270
+ "shape": [
271
+ 1
272
+ ],
273
+ "names": null
274
+ },
275
+ "index": {
276
+ "dtype": "int64",
277
+ "shape": [
278
+ 1
279
+ ],
280
+ "names": null
281
+ },
282
+ "observation.cam_rel_poses": {
283
+ "dtype": "float32",
284
+ "shape": [
285
+ 21
286
+ ],
287
+ "names": null
288
+ },
289
+ "observation.state": {
290
+ "dtype": "float32",
291
+ "shape": [
292
+ 256
293
+ ],
294
+ "names": null
295
+ },
296
+ "observation.task_info": {
297
+ "dtype": "float32",
298
+ "shape": [
299
+ null
300
+ ],
301
+ "names": null
302
+ }
303
+ }
304
+ }
305
+ ```
306
+
307
+
308
+ ## Citation
309
+
310
+ **BibTeX:**
311
+
312
+ ```bibtex
313
+ @article{li2024behavior,
314
+ title={Behavior-1k: A human-centered, embodied ai benchmark with 1,000 everyday activities and realistic simulation},
315
+ author={Li, Chengshu and Zhang, Ruohan and Wong, Josiah and Gokmen, Cem and Srivastava, Sanjana and Mart{'i}n-Mart{'i}n, Roberto and Wang, Chen and Levine, Gabrael and Ai, Wensi and Martinez, Benjamin and Yin, Hang and Lingelbach, Michael and Hwang, Minjune and Hiranaka, Ayano and Garlanka, Sujay and Aydin, Arman and Lee, Sharon and Sun, Jiankai and Anvari, Mona and Sharma, Manasi and Bansal, Dhruva and Hunter, Samuel and Kim, Kyu-Young and Lou, Alan and Matthews, Caleb R. and Villa-Renteria, Ivan and Tang, Jerry Huayang and Tang, Claire and Xia, Fei and Li, Yunzhu and Savarese, Silvio and Gweon, Hyowon and Liu, C. Karen and Wu, Jiajun and Fei-Fei, Li},
316
+ journal={arXiv preprint arXiv:2403.09227},
317
+ year={2024}
318
+ }
319
+ ```
source/README.md ADDED
@@ -0,0 +1,149 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-4.0
3
+ task_categories:
4
+ - robotics
5
+ tags:
6
+ - robotics
7
+ ---
8
+ # PhysicalAI-Robotics-GR00T-X-Embodiment-Sim
9
+
10
+ ![image/png](https://cdn-uploads.huggingface.co/production/uploads/67b8da81d01134f89899b4a7/PppFm7zvXyJblkMOq-YRQ.png)
11
+
12
+ Github Repo: [Isaac GR00T N1](https://github.com/NVIDIA/Isaac-GR00T)
13
+
14
+ We provide a set of datasets used for post-training of GR00T N1. Each dataset is a collection of trajectories from different robot embodiments and tasks.
15
+
16
+ ### Cross-embodied bimanual manipulation: 9k trajectories
17
+
18
+ | Dataset Name | #trajectories |
19
+ | - | -|
20
+ | bimanual_panda_gripper.Threading | 1000 |
21
+ | bimanual_panda_hand.LiftTray | 1000 |
22
+ | bimanual_panda_gripper.ThreePieceAssembly | 1000 |
23
+ | bimanual_panda_gripper.Transport | 1000 |
24
+ | bimanual_panda_hand.BoxCleanup | 1000 |
25
+ | bimanual_panda_hand.DrawerCleanup | 1000 |
26
+ | gr1_arms_only.CanSort | 1000 |
27
+ | gr1_full_upper_body.Coffee | 1000 |
28
+ | gr1_full_upper_body.Pouring | 1000 |
29
+
30
+
31
+ ### Humanoid robot tabletop manipulation: 240k trajectories
32
+
33
+ | Dataset Name | #trajectories |
34
+ | - | -|
35
+ | gr1_arms_waist.CanToDrawer | 10000 |
36
+ | gr1_arms_waist.CupToDrawer | 10000 |
37
+ | gr1_arms_waist.CuttingboardToBasket | 10000 |
38
+ | gr1_arms_waist.CuttingboardToCardboardBox | 10000 |
39
+ | gr1_arms_waist.CuttingboardToPan | 10000 |
40
+ | gr1_arms_waist.CuttingboardToPot | 10000 |
41
+ | gr1_arms_waist.CuttingboardToTieredBasket | 10000 |
42
+ | gr1_arms_waist.PlaceBottleToCabinet | 10000 |
43
+ | gr1_arms_waist.PlaceMilkToMicrowave | 10000 |
44
+ | gr1_arms_waist.PlacematToBasket | 10000 |
45
+ | gr1_arms_waist.PlacematToBowl | 10000 |
46
+ | gr1_arms_waist.PlacematToPlate | 10000 |
47
+ | gr1_arms_waist.PlacematToTieredShelf | 10000 |
48
+ | gr1_arms_waist.PlateToBowl | 10000 |
49
+ | gr1_arms_waist.PlateToCardboardBox | 10000 |
50
+ | gr1_arms_waist.PlateToPan | 10000 |
51
+ | gr1_arms_waist.PlateToPlate | 10000 |
52
+ | gr1_arms_waist.PotatoToMicrowave | 10000 |
53
+ | gr1_arms_waist.TrayToCardboardBox | 10000 |
54
+ | gr1_arms_waist.TrayToPlate | 10000 |
55
+ | gr1_arms_waist.TrayToPot | 10000 |
56
+ | gr1_arms_waist.TrayToTieredBasket | 10000 |
57
+ | gr1_arms_waist.TrayToTieredShelf | 10000 |
58
+ | gr1_arms_waist.WineToCabinet | 10000 |
59
+
60
+ ### Humanoid robot tabletop manipulation - downsampled: 24k trajectories
61
+
62
+ | Dataset Name | #trajectories |
63
+ | - | - |
64
+ | gr1_unified.PnPBottleToCabinetClose_GR1ArmsAndWaistFourierHands_1000 | 1000 |
65
+ | gr1_unified.PnPCanToDrawerClose_GR1ArmsAndWaistFourierHands_1000 | 1000 |
66
+ | gr1_unified.PnPCupToDrawerClose_GR1ArmsAndWaistFourierHands_1000 | 1000 |
67
+ | gr1_unified.PnPMilkToMicrowaveClose_GR1ArmsAndWaistFourierHands_1000 | 1000 |
68
+ | gr1_unified.PnPPotatoToMicrowaveClose_GR1ArmsAndWaistFourierHands_1000 | 1000 |
69
+ | gr1_unified.PnPWineToCabinetClose_GR1ArmsAndWaistFourierHands_1000 | 1000 |
70
+ | gr1_unified.PosttrainPnPNovelFromCuttingboardToBasketSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
71
+ | gr1_unified.PosttrainPnPNovelFromCuttingboardToCardboardboxSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
72
+ | gr1_unified.PosttrainPnPNovelFromCuttingboardToPanSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
73
+ | gr1_unified.PosttrainPnPNovelFromCuttingboardToPotSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
74
+ | gr1_unified.PosttrainPnPNovelFromCuttingboardToTieredbasketSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
75
+ | gr1_unified.PosttrainPnPNovelFromPlacematToBasketSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
76
+ | gr1_unified.PosttrainPnPNovelFromPlacematToBowlSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
77
+ | gr1_unified.PosttrainPnPNovelFromPlacematToPlateSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
78
+ | gr1_unified.PosttrainPnPNovelFromPlacematToTieredshelfSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
79
+ | gr1_unified.PosttrainPnPNovelFromPlateToBowlSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
80
+ | gr1_unified.PosttrainPnPNovelFromPlateToCardboardboxSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
81
+ | gr1_unified.PosttrainPnPNovelFromPlateToPanSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
82
+ | gr1_unified.PosttrainPnPNovelFromPlateToPlateSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
83
+ | gr1_unified.PosttrainPnPNovelFromTrayToCardboardboxSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
84
+ | gr1_unified.PosttrainPnPNovelFromTrayToPlateSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
85
+ | gr1_unified.PosttrainPnPNovelFromTrayToPotSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
86
+ | gr1_unified.PosttrainPnPNovelFromTrayToTieredbasketSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
87
+ | gr1_unified.PosttrainPnPNovelFromTrayToTieredshelfSplitA_GR1ArmsAndWaistFourierHands_1000 | 1000 |
88
+
89
+ ### Robot Arm Kitchen Manipulation: 72K trajectories
90
+ | Dataset Name | #trajectories |
91
+ | - | -|
92
+ | single_panda_gripper.CloseDoubleDoor | 3000 |
93
+ | single_panda_gripper.CloseDrawer | 3000 |
94
+ | single_panda_gripper.CloseSingleDoor | 3000 |
95
+ | single_panda_gripper.CoffeePressButton | 3000 |
96
+ | single_panda_gripper.CoffeeServeMug | 3000 |
97
+ | single_panda_gripper.CoffeeSetupMug | 3000 |
98
+ | single_panda_gripper.OpenDoubleDoor | 3000 |
99
+ | single_panda_gripper.OpenDrawer | 3000 |
100
+ | single_panda_gripper.OpenSingleDoor | 3000 |
101
+ | single_panda_gripper.PnPCabToCounter | 3000 |
102
+ | single_panda_gripper.PnPCounterToCab | 3000 |
103
+ | single_panda_gripper.PnPCounterToMicrowave | 3000 |
104
+ | single_panda_gripper.PnPCounterToSink | 3000 |
105
+ | single_panda_gripper.PnPCounterToStove | 3000 |
106
+ | single_panda_gripper.PnPMicrowaveToCounter | 3000 |
107
+ | single_panda_gripper.PnPSinkToCounter | 3000 |
108
+ | single_panda_gripper.PnPStoveToCounter | 3000 |
109
+ | single_panda_gripper.TurnOffMicrowave | 3000 |
110
+ | single_panda_gripper.TurnOffSinkFaucet | 3000 |
111
+ | single_panda_gripper.TurnOffStove | 3000 |
112
+ | single_panda_gripper.TurnOnMicrowave | 3000 |
113
+ | single_panda_gripper.TurnOnSinkFaucet | 3000 |
114
+ | single_panda_gripper.TurnOnStove | 3000 |
115
+ | single_panda_gripper.TurnSinkSpout | 3000 |
116
+
117
+ ### Unitree G1 Loco-Manipulation: 102 trajectories
118
+ | Dataset Name | #trajectories |
119
+ | - | -|
120
+ | unitree_g1.LMPnPAppleToPlateDC | 102 |
121
+
122
+ ## Download the dataset from Huggingface
123
+
124
+ Users can download a specific subset of data by specifying the dataset name.
125
+
126
+ 1. Option 1: with `huggingface-cli`
127
+
128
+ ```bash
129
+ huggingface-cli download nvidia/PhysicalAI-Robotics-GR00T-X-Embodiment-Sim \
130
+ --repo-type dataset \
131
+ --include "gr1_arms_only.CanSort/**" \
132
+ --local-dir $HOME/gr00t_dataset
133
+ ```
134
+
135
+ **Replace `gr1_arms_only.CanSort/**` with the dataset name you want to download.**
136
+
137
+ 2. Option 2: Github LFS
138
+
139
+ ```bash
140
+ # Clone the repo without downloading files
141
+ git clone --filter=blob:none --no-checkout https://huggingface.co/datasets/nvidia/PhysicalAI-Robotics-GR00T-X-Embodiment-Sim
142
+ cd PhysicalAI-Robotics-GR00T-X-Embodiment-Sim
143
+
144
+ # Configure sparse-checkout for just your folder
145
+ git sparse-checkout init --cone
146
+ git sparse-checkout set gr1_arms_waist.CupToDrawer << Select the specific dataset to download
147
+ ```
148
+
149
+ Replace the "gr1_arms_waist.CupToDrawer" with the intended dataset folder
viewer/episodes.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:630ff9384fca29b9b92614ba0f99a322c667db41f21abc2060bb87aacd334bda
3
+ size 11298
viewer/subsets.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:276ea91701d4812593333426aca019332eb734c1e8822abaff543b1deb24cf01
3
+ size 10378