mliu48 commited on
Commit
5799abc
·
verified ·
1 Parent(s): 34ee96b

Env moved to gymnasium_env/telearms_env.py

Browse files
Files changed (31) hide show
  1. CubePatchPlacement/elo-1500-v0/data/metadata.json +1 -1
  2. CubePatchPlacement/random-1500-v0/data/metadata.json +1 -1
  3. CubePatchPlacement/score-1500-v0/data/metadata.json +1 -1
  4. FrontEdgeCubeRow/elo-1500-v0/data/metadata.json +1 -1
  5. FrontEdgeCubeRow/random-1500-v0/data/metadata.json +1 -1
  6. FrontEdgeCubeRow/score-1500-v0/data/metadata.json +1 -1
  7. FrontToBackCubeLine/elo-1500-v0/data/metadata.json +1 -1
  8. FrontToBackCubeLine/random-1500-v0/data/metadata.json +1 -1
  9. FrontToBackCubeLine/score-1500-v0/data/metadata.json +1 -1
  10. LCubeArrangement/elo-1500-v0/data/metadata.json +1 -1
  11. LCubeArrangement/random-1500-v0/data/metadata.json +1 -1
  12. LCubeArrangement/score-1500-v0/data/metadata.json +1 -1
  13. LetterFaceAlignment/elo-1500-v0/data/metadata.json +1 -1
  14. LetterFaceAlignment/random-1500-v0/data/metadata.json +1 -1
  15. LetterFaceAlignment/score-1500-v0/data/metadata.json +1 -1
  16. README.md +2 -2
  17. RedCubeInFrontOfBlueCube/elo-1500-v0/data/metadata.json +1 -1
  18. RedCubeInFrontOfBlueCube/random-1500-v0/data/metadata.json +1 -1
  19. RedCubeInFrontOfBlueCube/score-1500-v0/data/metadata.json +1 -1
  20. RedLeftOfBluePlacement/elo-1500-v0/data/metadata.json +1 -1
  21. RedLeftOfBluePlacement/random-1500-v0/data/metadata.json +1 -1
  22. RedLeftOfBluePlacement/score-1500-v0/data/metadata.json +1 -1
  23. SemanticCubeOnRedPatchAUp/elo-1500-v0/data/metadata.json +1 -1
  24. SemanticCubeOnRedPatchAUp/random-1500-v0/data/metadata.json +1 -1
  25. SemanticCubeOnRedPatchAUp/score-1500-v0/data/metadata.json +1 -1
  26. SequentialPushBlockToTarget/elo-1500-v0/data/metadata.json +1 -1
  27. SequentialPushBlockToTarget/random-1500-v0/data/metadata.json +1 -1
  28. SequentialPushBlockToTarget/score-1500-v0/data/metadata.json +1 -1
  29. ThreeCubeWallRow/elo-1500-v0/data/metadata.json +1 -1
  30. ThreeCubeWallRow/random-1500-v0/data/metadata.json +1 -1
  31. ThreeCubeWallRow/score-1500-v0/data/metadata.json +1 -1
CubePatchPlacement/elo-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 816978, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [54], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/CubePatchPlacement\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Cube Patch Placement\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/CubePatchPlacement/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Cube Patch Placement', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 54-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 267.5}
 
1
+ {"total_episodes": 1500, "total_steps": 816978, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [54], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/CubePatchPlacement\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Cube Patch Placement\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/CubePatchPlacement/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Cube Patch Placement', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 54-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 267.5}
CubePatchPlacement/random-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 1123302, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [54], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/CubePatchPlacement\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Cube Patch Placement\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/CubePatchPlacement/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Cube Patch Placement', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 54-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 358.8}
 
1
+ {"total_episodes": 1500, "total_steps": 1123302, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [54], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/CubePatchPlacement\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Cube Patch Placement\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/CubePatchPlacement/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Cube Patch Placement', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 54-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 358.8}
CubePatchPlacement/score-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 690766, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [54], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/CubePatchPlacement\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Cube Patch Placement\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/CubePatchPlacement/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Cube Patch Placement', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 54-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 230.2}
 
1
+ {"total_episodes": 1500, "total_steps": 690766, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [54], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/CubePatchPlacement\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Cube Patch Placement\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/CubePatchPlacement/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Cube Patch Placement', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 54-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 230.2}
FrontEdgeCubeRow/elo-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 1258916, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [53], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/FrontEdgeCubeRow\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Front Edge Cube Row\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/FrontEdgeCubeRow/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Front Edge Cube Row', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 53-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 397.5}
 
1
+ {"total_episodes": 1500, "total_steps": 1258916, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [53], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/FrontEdgeCubeRow\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Front Edge Cube Row\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/FrontEdgeCubeRow/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Front Edge Cube Row', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 53-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 397.5}
FrontEdgeCubeRow/random-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 1761905, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [53], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/FrontEdgeCubeRow\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Front Edge Cube Row\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/FrontEdgeCubeRow/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Front Edge Cube Row', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 53-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 548.5}
 
1
+ {"total_episodes": 1500, "total_steps": 1761905, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [53], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/FrontEdgeCubeRow\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Front Edge Cube Row\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/FrontEdgeCubeRow/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Front Edge Cube Row', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 53-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 548.5}
FrontEdgeCubeRow/score-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 1334777, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [53], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/FrontEdgeCubeRow\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Front Edge Cube Row\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/FrontEdgeCubeRow/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Front Edge Cube Row', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 53-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 420.0}
 
1
+ {"total_episodes": 1500, "total_steps": 1334777, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [53], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/FrontEdgeCubeRow\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Front Edge Cube Row\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/FrontEdgeCubeRow/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Front Edge Cube Row', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 53-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 420.0}
FrontToBackCubeLine/elo-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 1068287, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [53], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/FrontToBackCubeLine\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Front To Back Cube Line\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/FrontToBackCubeLine/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Front To Back Cube Line', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 53-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 341.7}
 
1
+ {"total_episodes": 1500, "total_steps": 1068287, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [53], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/FrontToBackCubeLine\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Front To Back Cube Line\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/FrontToBackCubeLine/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Front To Back Cube Line', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 53-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 341.7}
FrontToBackCubeLine/random-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 1602767, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [53], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/FrontToBackCubeLine\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Front To Back Cube Line\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/FrontToBackCubeLine/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Front To Back Cube Line', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 53-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 501.6}
 
1
+ {"total_episodes": 1500, "total_steps": 1602767, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [53], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/FrontToBackCubeLine\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Front To Back Cube Line\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/FrontToBackCubeLine/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Front To Back Cube Line', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 53-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 501.6}
FrontToBackCubeLine/score-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 949023, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [53], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/FrontToBackCubeLine\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Front To Back Cube Line\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/FrontToBackCubeLine/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Front To Back Cube Line', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 53-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 306.6}
 
1
+ {"total_episodes": 1500, "total_steps": 949023, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [53], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/FrontToBackCubeLine\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Front To Back Cube Line\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/FrontToBackCubeLine/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Front To Back Cube Line', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 53-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 306.6}
LCubeArrangement/elo-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 1790704, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [66], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/LCubeArrangement\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"L Cube Arrangement\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/LCubeArrangement/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'L Cube Arrangement', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 66-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 643.9}
 
1
+ {"total_episodes": 1500, "total_steps": 1790704, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [66], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/LCubeArrangement\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"L Cube Arrangement\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/LCubeArrangement/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'L Cube Arrangement', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 66-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 643.9}
LCubeArrangement/random-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 2515156, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [66], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/LCubeArrangement\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"L Cube Arrangement\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/LCubeArrangement/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'L Cube Arrangement', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 66-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 910.0}
 
1
+ {"total_episodes": 1500, "total_steps": 2515156, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [66], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/LCubeArrangement\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"L Cube Arrangement\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/LCubeArrangement/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'L Cube Arrangement', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 66-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 910.0}
LCubeArrangement/score-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 1725123, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [66], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/LCubeArrangement\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"L Cube Arrangement\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/LCubeArrangement/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'L Cube Arrangement', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 66-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 622.2}
 
1
+ {"total_episodes": 1500, "total_steps": 1725123, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [66], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/LCubeArrangement\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"L Cube Arrangement\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/LCubeArrangement/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'L Cube Arrangement', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 66-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 622.2}
LetterFaceAlignment/elo-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 798595, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [27], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/LetterFaceAlignment\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Letter Face Alignment\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/LetterFaceAlignment/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Letter Face Alignment', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 27-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 173.7}
 
1
+ {"total_episodes": 1500, "total_steps": 798595, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [27], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/LetterFaceAlignment\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Letter Face Alignment\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/LetterFaceAlignment/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Letter Face Alignment', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 27-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 173.7}
LetterFaceAlignment/random-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 1234005, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [27], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/LetterFaceAlignment\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Letter Face Alignment\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/LetterFaceAlignment/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Letter Face Alignment', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 27-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 254.9}
 
1
+ {"total_episodes": 1500, "total_steps": 1234005, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [27], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/LetterFaceAlignment\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Letter Face Alignment\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/LetterFaceAlignment/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Letter Face Alignment', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 27-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 254.9}
LetterFaceAlignment/score-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 629146, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [27], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/LetterFaceAlignment\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Letter Face Alignment\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/LetterFaceAlignment/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Letter Face Alignment', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 27-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 143.0}
 
1
+ {"total_episodes": 1500, "total_steps": 629146, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [27], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/LetterFaceAlignment\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Letter Face Alignment\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/LetterFaceAlignment/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Letter Face Alignment', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 27-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 143.0}
README.md CHANGED
@@ -21,11 +21,11 @@ YAM cube tasks in MuJoCo: each task in three subsets of 1,500 demos, picked at r
21
  import minari
22
  minari.download_dataset("hf://mliu48/telearms/CubePatchPlacement/random-1500-v0")
23
  dataset = minari.load_dataset("telearms/CubePatchPlacement/random-1500-v0")
24
- env = dataset.recover_environment() # needs training/telearms_env.py on the path
25
  ```
26
 
27
  The environment is `TeleArmsEnv(task, obs_mode="privileged")` in
28
- [training/telearms_env.py](https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env). On a machine without EGL, set `MUJOCO_GL=disable`
29
  before importing it.
30
 
31
  **Observations** are the simulator's state, not images: the robot's 14 joint and gripper values,
 
21
  import minari
22
  minari.download_dataset("hf://mliu48/telearms/CubePatchPlacement/random-1500-v0")
23
  dataset = minari.load_dataset("telearms/CubePatchPlacement/random-1500-v0")
24
+ env = dataset.recover_environment() # needs the repo's gymnasium_env/ on the path
25
  ```
26
 
27
  The environment is `TeleArmsEnv(task, obs_mode="privileged")` in
28
+ [gymnasium_env/telearms_env.py](https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env/gymnasium_env). On a machine without EGL, set `MUJOCO_GL=disable`
29
  before importing it.
30
 
31
  **Observations** are the simulator's state, not images: the robot's 14 joint and gripper values,
RedCubeInFrontOfBlueCube/elo-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 533290, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [40], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/RedCubeInFrontOfBlueCube\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Red Cube In Front Of Blue Cube\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/RedCubeInFrontOfBlueCube/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Red Cube In Front Of Blue Cube', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 40-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 151.8}
 
1
+ {"total_episodes": 1500, "total_steps": 533290, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [40], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/RedCubeInFrontOfBlueCube\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Red Cube In Front Of Blue Cube\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/RedCubeInFrontOfBlueCube/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Red Cube In Front Of Blue Cube', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 40-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 151.8}
RedCubeInFrontOfBlueCube/random-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 629595, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [40], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/RedCubeInFrontOfBlueCube\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Red Cube In Front Of Blue Cube\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/RedCubeInFrontOfBlueCube/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Red Cube In Front Of Blue Cube', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 40-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 173.9}
 
1
+ {"total_episodes": 1500, "total_steps": 629595, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [40], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/RedCubeInFrontOfBlueCube\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Red Cube In Front Of Blue Cube\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/RedCubeInFrontOfBlueCube/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Red Cube In Front Of Blue Cube', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 40-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 173.9}
RedCubeInFrontOfBlueCube/score-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 420706, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [40], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/RedCubeInFrontOfBlueCube\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Red Cube In Front Of Blue Cube\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/RedCubeInFrontOfBlueCube/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Red Cube In Front Of Blue Cube', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 40-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 126.1}
 
1
+ {"total_episodes": 1500, "total_steps": 420706, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [40], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/RedCubeInFrontOfBlueCube\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Red Cube In Front Of Blue Cube\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/RedCubeInFrontOfBlueCube/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Red Cube In Front Of Blue Cube', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 40-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 126.1}
RedLeftOfBluePlacement/elo-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 483722, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [40], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/RedLeftOfBluePlacement\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Red Left Of Blue Placement\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/RedLeftOfBluePlacement/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Red Left Of Blue Placement', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 40-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 140.5}
 
1
+ {"total_episodes": 1500, "total_steps": 483722, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [40], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/RedLeftOfBluePlacement\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Red Left Of Blue Placement\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/RedLeftOfBluePlacement/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Red Left Of Blue Placement', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 40-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 140.5}
RedLeftOfBluePlacement/random-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 619378, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [40], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/RedLeftOfBluePlacement\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Red Left Of Blue Placement\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/RedLeftOfBluePlacement/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Red Left Of Blue Placement', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 40-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 171.7}
 
1
+ {"total_episodes": 1500, "total_steps": 619378, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [40], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/RedLeftOfBluePlacement\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Red Left Of Blue Placement\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/RedLeftOfBluePlacement/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Red Left Of Blue Placement', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 40-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 171.7}
RedLeftOfBluePlacement/score-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 404433, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [40], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/RedLeftOfBluePlacement\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Red Left Of Blue Placement\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/RedLeftOfBluePlacement/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Red Left Of Blue Placement', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 40-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 122.4}
 
1
+ {"total_episodes": 1500, "total_steps": 404433, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [40], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/RedLeftOfBluePlacement\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Red Left Of Blue Placement\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/RedLeftOfBluePlacement/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Red Left Of Blue Placement', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 40-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 122.4}
SemanticCubeOnRedPatchAUp/elo-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 1291808, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [34], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/SemanticCubeOnRedPatchAUp\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Semantic Cube On Red Patch A Up\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/SemanticCubeOnRedPatchAUp/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Semantic Cube On Red Patch A Up', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 34-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 302.2}
 
1
+ {"total_episodes": 1500, "total_steps": 1291808, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [34], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/SemanticCubeOnRedPatchAUp\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Semantic Cube On Red Patch A Up\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/SemanticCubeOnRedPatchAUp/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Semantic Cube On Red Patch A Up', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 34-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 302.2}
SemanticCubeOnRedPatchAUp/random-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 1939049, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [34], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/SemanticCubeOnRedPatchAUp\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Semantic Cube On Red Patch A Up\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/SemanticCubeOnRedPatchAUp/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Semantic Cube On Red Patch A Up', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 34-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 444.5}
 
1
+ {"total_episodes": 1500, "total_steps": 1939049, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [34], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/SemanticCubeOnRedPatchAUp\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Semantic Cube On Red Patch A Up\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/SemanticCubeOnRedPatchAUp/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Semantic Cube On Red Patch A Up', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 34-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 444.5}
SemanticCubeOnRedPatchAUp/score-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 1133792, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [34], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/SemanticCubeOnRedPatchAUp\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Semantic Cube On Red Patch A Up\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/SemanticCubeOnRedPatchAUp/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Semantic Cube On Red Patch A Up', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 34-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 268.3}
 
1
+ {"total_episodes": 1500, "total_steps": 1133792, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [34], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/SemanticCubeOnRedPatchAUp\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Semantic Cube On Red Patch A Up\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/SemanticCubeOnRedPatchAUp/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Semantic Cube On Red Patch A Up', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 34-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 268.3}
SequentialPushBlockToTarget/elo-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 920421, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [54], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/SequentialPushBlockToTarget\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Sequential Push Block to Target\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/SequentialPushBlockToTarget/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Sequential Push Block to Target', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 54-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 297.9}
 
1
+ {"total_episodes": 1500, "total_steps": 920421, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [54], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/SequentialPushBlockToTarget\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Sequential Push Block to Target\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/SequentialPushBlockToTarget/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Sequential Push Block to Target', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 54-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 297.9}
SequentialPushBlockToTarget/random-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 1155646, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [54], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/SequentialPushBlockToTarget\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Sequential Push Block to Target\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/SequentialPushBlockToTarget/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Sequential Push Block to Target', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 54-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 368.2}
 
1
+ {"total_episodes": 1500, "total_steps": 1155646, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [54], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/SequentialPushBlockToTarget\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Sequential Push Block to Target\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/SequentialPushBlockToTarget/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Sequential Push Block to Target', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 54-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 368.2}
SequentialPushBlockToTarget/score-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 702845, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [54], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/SequentialPushBlockToTarget\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Sequential Push Block to Target\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/SequentialPushBlockToTarget/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Sequential Push Block to Target', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 54-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 233.5}
 
1
+ {"total_episodes": 1500, "total_steps": 702845, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [54], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/SequentialPushBlockToTarget\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Sequential Push Block to Target\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/SequentialPushBlockToTarget/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Sequential Push Block to Target', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 54-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 233.5}
ThreeCubeWallRow/elo-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 1081069, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [53], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/ThreeCubeWallRow\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Three Cube Wall Row\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/ThreeCubeWallRow/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Three Cube Wall Row', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 53-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 345.6}
 
1
+ {"total_episodes": 1500, "total_steps": 1081069, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [53], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/ThreeCubeWallRow\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Three Cube Wall Row\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/ThreeCubeWallRow/elo-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Three Cube Wall Row', human teleoperation demos in simulation: the 1,500 demos per task from the players with the highest ELO, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-elo-1500 for this task. State-based (privileged) version: observations are the 53-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 345.6}
ThreeCubeWallRow/random-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 1420035, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [53], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/ThreeCubeWallRow\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Three Cube Wall Row\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/ThreeCubeWallRow/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Three Cube Wall Row', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 53-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 446.6}
 
1
+ {"total_episodes": 1500, "total_steps": 1420035, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [53], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/ThreeCubeWallRow\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Three Cube Wall Row\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/ThreeCubeWallRow/random-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Three Cube Wall Row', human teleoperation demos in simulation: 1,500 demos per task drawn at random, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-random-1500 for this task. State-based (privileged) version: observations are the 53-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 446.6}
ThreeCubeWallRow/score-1500-v0/data/metadata.json CHANGED
@@ -1 +1 @@
1
- {"total_episodes": 1500, "total_steps": 932903, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [53], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/ThreeCubeWallRow\", \"entry_point\": \"training.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Three Cube Wall Row\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/ThreeCubeWallRow/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Three Cube Wall Row', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 53-dim obs_mode='privileged' vector of training/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 301.9}
 
1
+ {"total_episodes": 1500, "total_steps": 932903, "data_format": "hdf5", "observation_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [53], \"low\": [-Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity, -Infinity], \"high\": [Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity, Infinity]}", "action_space": "{\"type\": \"Box\", \"dtype\": \"float32\", \"shape\": [14], \"low\": [-3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732, -3.1415927410125732], \"high\": [3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732, 3.1415927410125732]}", "env_spec": "{\"id\": \"telearms/ThreeCubeWallRow\", \"entry_point\": \"gymnasium_env.telearms_env:TeleArmsEnv\", \"reward_threshold\": null, \"nondeterministic\": false, \"max_episode_steps\": null, \"order_enforce\": true, \"disable_env_checker\": true, \"kwargs\": {\"task\": \"Three Cube Wall Row\", \"obs_mode\": \"privileged\"}, \"additional_wrappers\": [], \"vector_entry_point\": null}", "dataset_id": "telearms/ThreeCubeWallRow/score-1500-v0", "code_permalink": "https://github.com/helen9975/RoboEval-telearms/tree/gymnasium-env", "author": ["TeleArms"], "algorithm_name": "human teleoperation", "description": "TeleArms YAM bimanual cube task 'Three Cube Wall Row', human teleoperation demos in simulation: the 1,500 demos per task with the highest demo score, the episodes of the LeRobot dataset mliu48/telearms-yam-cube-v5-exp-score-1500 for this task. State-based (privileged) version: observations are the 53-dim obs_mode='privileged' vector of gymnasium_env/telearms_env.py (14-dim policy joint state, each cube's position, wxyz quaternion and 6-dim velocity, each goal patch's position and wxyz quaternion); actions are 14-dim absolute joint targets at 30 Hz. Each demo is cut at its first success; rewards are sparse (1.0 on that step); infos['reward_dense'] is the env's dense reward (1/N per task stage first reached, plus 1.0 on success); demos that never succeed under the env's check end truncated.", "minari_version": "0.5.3", "dataset_size": 301.9}