N053: instruction text -> instructions.tsv@30762bb60 (400 files)

N053/go2_arm/dr_on/seed_2119/rec/check.json

@@ -23,7 +23,7 @@
   },
   "task text = MobileBench instruction": {
    "pass": true,
-   "detail": "task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'"
+   "detail": "task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'"
   },
   "rows = recorded control steps / 2 (= episode length at 25 Hz)": {
    "pass": true,

N053/go2_arm/dr_on/seed_2119/rec/check.txt

@@ -7,7 +7,7 @@ dataset root /mnt/scratch/sz850_atomic/n053_go2_on_2119/out/rec/memorystage/000/
 --- episode file episode_000000.parquet: 1315 rows, columns ['action', 'action_is_hold', 'observation.state', 'observation.joint_vel', 'observation.eef_pose', 'observation.environment_state', 'action.command', 'action.applied', 'narration', 'collect_info', 'extra.reward', 'extra.action_info', 'next.truncated', 'sim_step_idx', 'timestamp', 'frame_index', 'episode_index', 'index', 'task_index', 'next.done']
 [PASS] observation.state shape: parquet width 26, info shape [26], 26 names first: ['joint_0', 'joint_1', 'joint_2', 'joint_3', 'joint_4', 'joint_5']
 [PASS] action shape: parquet width 12, info shape [12], 12 names first: ['base_action_0', 'base_action_1', 'base_action_2', 'base_action_3', 'eef0.pos_x', 'eef0.pos_y']
-[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'
+[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'
 rows 1315 at dataset fps 25 = 52.60 s
 [PASS] rows = recorded control steps / 2 (= episode length at 25 Hz): 1315 rows vs ceil(2629/2) = 1315
 [PASS] rows are every 2nd control step (sim_step_idx): sim_step_idx 61..2689, step differences [2]

N053/go2_arm/dr_on/seed_2119/rec/memorystage/000/lerobot_v21/meta/episodes.jsonl

@@ -1 +1 @@
-{"episode_index": 0, "tasks": ["Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."], "length": 1315, "magicsim/control_steps": 2629, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is to the left of the rack. The white st booth r is on top of the rack. The white st booth r is to the right of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is to the right of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front-left, at table height."}
+{"episode_index": 0, "tasks": ["Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."], "length": 1315, "magicsim/control_steps": 2629, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is to the left of the rack. The white st booth r is on top of the rack. The white st booth r is to the right of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is to the right of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front-left, at table height."}

N053/go2_arm/dr_on/seed_2119/rec/memorystage/000/lerobot_v21/meta/extra/episode_000000.json

@@ -534,7 +534,7 @@
   },
   "task": {
    "task_type": "MemoryStage",
-   "instruction": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.",
+   "instruction": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.",
    "template_id": -1,
    "slots": {
     "obj": 1.5707963267948966,

N053/go2_arm/dr_on/seed_2119/rec/memorystage/000/lerobot_v21/meta/tasks.jsonl

@@ -1 +1 @@
-{"task_index": 0, "task": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."}
+{"task_index": 0, "task": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."}

N053/go2_arm/dr_on/seed_2120/rec/check.json

@@ -23,7 +23,7 @@
   },
   "task text = MobileBench instruction": {
    "pass": true,
-   "detail": "task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'"
+   "detail": "task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'"
   },
   "rows = recorded control steps / 2 (= episode length at 25 Hz)": {
    "pass": true,

N053/go2_arm/dr_on/seed_2120/rec/check.txt

@@ -7,7 +7,7 @@ dataset root /mnt/scratch/sz850_atomic/n053_go2_on_2120/out/rec/memorystage/000/
 --- episode file episode_000000.parquet: 1180 rows, columns ['action', 'action_is_hold', 'observation.state', 'observation.joint_vel', 'observation.eef_pose', 'observation.environment_state', 'action.command', 'action.applied', 'narration', 'collect_info', 'extra.reward', 'extra.action_info', 'next.truncated', 'sim_step_idx', 'timestamp', 'frame_index', 'episode_index', 'index', 'task_index', 'next.done']
 [PASS] observation.state shape: parquet width 26, info shape [26], 26 names first: ['joint_0', 'joint_1', 'joint_2', 'joint_3', 'joint_4', 'joint_5']
 [PASS] action shape: parquet width 12, info shape [12], 12 names first: ['base_action_0', 'base_action_1', 'base_action_2', 'base_action_3', 'eef0.pos_x', 'eef0.pos_y']
-[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'
+[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'
 rows 1180 at dataset fps 25 = 47.20 s
 [PASS] rows = recorded control steps / 2 (= episode length at 25 Hz): 1180 rows vs ceil(2359/2) = 1180
 [PASS] rows are every 2nd control step (sim_step_idx): sim_step_idx 61..2419, step differences [2]

N053/go2_arm/dr_on/seed_2120/rec/memorystage/000/lerobot_v21/meta/episodes.jsonl

@@ -1 +1 @@
-{"episode_index": 0, "tasks": ["Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."], "length": 1180, "magicsim/control_steps": 2359, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is to the left of the rack. The white st booth r is on top of the rack. The white st booth r is to the right of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is in front of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front-left, at table height."}
+{"episode_index": 0, "tasks": ["Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."], "length": 1180, "magicsim/control_steps": 2359, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is to the left of the rack. The white st booth r is on top of the rack. The white st booth r is to the right of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is in front of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front-left, at table height."}

N053/go2_arm/dr_on/seed_2120/rec/memorystage/000/lerobot_v21/meta/extra/episode_000000.json

@@ -534,7 +534,7 @@
   },
   "task": {
    "task_type": "MemoryStage",
-   "instruction": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.",
+   "instruction": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.",
    "template_id": -1,
    "slots": {
     "obj": 1.5707963267948966,

N053/go2_arm/dr_on/seed_2120/rec/memorystage/000/lerobot_v21/meta/tasks.jsonl

@@ -1 +1 @@
-{"task_index": 0, "task": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."}
+{"task_index": 0, "task": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."}

N053/go2_arm/dr_on/seed_2121/rec/check.json

@@ -23,7 +23,7 @@
   },
   "task text = MobileBench instruction": {
    "pass": true,
-   "detail": "task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'"
+   "detail": "task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'"
   },
   "rows = recorded control steps / 2 (= episode length at 25 Hz)": {
    "pass": true,

N053/go2_arm/dr_on/seed_2121/rec/check.txt

@@ -7,7 +7,7 @@ dataset root /mnt/scratch/sz850_atomic/n053_go2_on_2121/out/rec/memorystage/000/
 --- episode file episode_000000.parquet: 1174 rows, columns ['action', 'action_is_hold', 'observation.state', 'observation.joint_vel', 'observation.eef_pose', 'observation.environment_state', 'action.command', 'action.applied', 'narration', 'collect_info', 'extra.reward', 'extra.action_info', 'next.truncated', 'sim_step_idx', 'timestamp', 'frame_index', 'episode_index', 'index', 'task_index', 'next.done']
 [PASS] observation.state shape: parquet width 26, info shape [26], 26 names first: ['joint_0', 'joint_1', 'joint_2', 'joint_3', 'joint_4', 'joint_5']
 [PASS] action shape: parquet width 12, info shape [12], 12 names first: ['base_action_0', 'base_action_1', 'base_action_2', 'base_action_3', 'eef0.pos_x', 'eef0.pos_y']
-[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'
+[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'
 rows 1174 at dataset fps 25 = 46.96 s
 [PASS] rows = recorded control steps / 2 (= episode length at 25 Hz): 1174 rows vs ceil(2347/2) = 1174
 [PASS] rows are every 2nd control step (sim_step_idx): sim_step_idx 61..2407, step differences [2]

N053/go2_arm/dr_on/seed_2121/rec/memorystage/000/lerobot_v21/meta/episodes.jsonl

@@ -1 +1 @@
-{"episode_index": 0, "tasks": ["Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."], "length": 1174, "magicsim/control_steps": 2347, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is to the left of the rack. The white st booth r is on top of the rack. The white st booth r is to the right of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is to the right of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front, at table height."}
+{"episode_index": 0, "tasks": ["Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."], "length": 1174, "magicsim/control_steps": 2347, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is to the left of the rack. The white st booth r is on top of the rack. The white st booth r is to the right of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is to the right of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front, at table height."}

N053/go2_arm/dr_on/seed_2121/rec/memorystage/000/lerobot_v21/meta/extra/episode_000000.json

@@ -534,7 +534,7 @@
   },
   "task": {
    "task_type": "MemoryStage",
-   "instruction": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.",
+   "instruction": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.",
    "template_id": -1,
    "slots": {
     "obj": 1.5707963267948966,

N053/go2_arm/dr_on/seed_2121/rec/memorystage/000/lerobot_v21/meta/tasks.jsonl

@@ -1 +1 @@
-{"task_index": 0, "task": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."}
+{"task_index": 0, "task": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."}

N053/go2_arm/dr_on/seed_2122/rec/check.json

@@ -23,7 +23,7 @@
   },
   "task text = MobileBench instruction": {
    "pass": true,
-   "detail": "task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'"
+   "detail": "task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'"
   },
   "rows = recorded control steps / 2 (= episode length at 25 Hz)": {
    "pass": true,

N053/go2_arm/dr_on/seed_2122/rec/check.txt

@@ -7,7 +7,7 @@ dataset root /mnt/scratch/sz850_atomic/n053_go2_on_2122_r1/out/rec/memorystage/0
 --- episode file episode_000000.parquet: 1241 rows, columns ['action', 'action_is_hold', 'observation.state', 'observation.joint_vel', 'observation.eef_pose', 'observation.environment_state', 'action.command', 'action.applied', 'narration', 'collect_info', 'extra.reward', 'extra.action_info', 'next.truncated', 'sim_step_idx', 'timestamp', 'frame_index', 'episode_index', 'index', 'task_index', 'next.done']
 [PASS] observation.state shape: parquet width 26, info shape [26], 26 names first: ['joint_0', 'joint_1', 'joint_2', 'joint_3', 'joint_4', 'joint_5']
 [PASS] action shape: parquet width 12, info shape [12], 12 names first: ['base_action_0', 'base_action_1', 'base_action_2', 'base_action_3', 'eef0.pos_x', 'eef0.pos_y']
-[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'
+[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'
 rows 1241 at dataset fps 25 = 49.64 s
 [PASS] rows = recorded control steps / 2 (= episode length at 25 Hz): 1241 rows vs ceil(2482/2) = 1241
 [PASS] rows are every 2nd control step (sim_step_idx): sim_step_idx 61..2541, step differences [2]

N053/go2_arm/dr_on/seed_2122/rec/memorystage/000/lerobot_v21/meta/episodes.jsonl

@@ -1 +1 @@
-{"episode_index": 0, "tasks": ["Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."], "length": 1241, "magicsim/control_steps": 2482, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is to the right of the rack. The white st booth r is on top of the rack. The white st booth r is to the left of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is in front of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front-left, at table height."}
+{"episode_index": 0, "tasks": ["Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."], "length": 1241, "magicsim/control_steps": 2482, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is to the right of the rack. The white st booth r is on top of the rack. The white st booth r is to the left of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is in front of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front-left, at table height."}

N053/go2_arm/dr_on/seed_2122/rec/memorystage/000/lerobot_v21/meta/extra/episode_000000.json

@@ -534,7 +534,7 @@
   },
   "task": {
    "task_type": "MemoryStage",
-   "instruction": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.",
+   "instruction": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.",
    "template_id": -1,
    "slots": {
     "obj": 1.5707963267948966,

N053/go2_arm/dr_on/seed_2122/rec/memorystage/000/lerobot_v21/meta/tasks.jsonl

@@ -1 +1 @@
-{"task_index": 0, "task": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."}
+{"task_index": 0, "task": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."}

N053/go2_arm/dr_on/seed_2123/rec/check.json

@@ -23,7 +23,7 @@
   },
   "task text = MobileBench instruction": {
    "pass": true,
-   "detail": "task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'"
+   "detail": "task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'"
   },
   "rows = recorded control steps / 2 (= episode length at 25 Hz)": {
    "pass": true,

N053/go2_arm/dr_on/seed_2123/rec/check.txt

@@ -7,7 +7,7 @@ dataset root /mnt/scratch/sz850_atomic/n053_go2_on_2123/out/rec/memorystage/000/
 --- episode file episode_000000.parquet: 1172 rows, columns ['action', 'action_is_hold', 'observation.state', 'observation.joint_vel', 'observation.eef_pose', 'observation.environment_state', 'action.command', 'action.applied', 'narration', 'collect_info', 'extra.reward', 'extra.action_info', 'next.truncated', 'sim_step_idx', 'timestamp', 'frame_index', 'episode_index', 'index', 'task_index', 'next.done']
 [PASS] observation.state shape: parquet width 26, info shape [26], 26 names first: ['joint_0', 'joint_1', 'joint_2', 'joint_3', 'joint_4', 'joint_5']
 [PASS] action shape: parquet width 12, info shape [12], 12 names first: ['base_action_0', 'base_action_1', 'base_action_2', 'base_action_3', 'eef0.pos_x', 'eef0.pos_y']
-[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'
+[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'
 rows 1172 at dataset fps 25 = 46.88 s
 [PASS] rows = recorded control steps / 2 (= episode length at 25 Hz): 1172 rows vs ceil(2344/2) = 1172
 [PASS] rows are every 2nd control step (sim_step_idx): sim_step_idx 61..2403, step differences [2]

N053/go2_arm/dr_on/seed_2123/rec/memorystage/000/lerobot_v21/meta/episodes.jsonl

@@ -1 +1 @@
-{"episode_index": 0, "tasks": ["Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."], "length": 1172, "magicsim/control_steps": 2344, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is to the left of the rack. The white st booth r is on top of the rack. The white st booth r is to the right of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is to the right of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front, at table height."}
+{"episode_index": 0, "tasks": ["Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."], "length": 1172, "magicsim/control_steps": 2344, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is to the left of the rack. The white st booth r is on top of the rack. The white st booth r is to the right of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is to the right of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front, at table height."}

N053/go2_arm/dr_on/seed_2123/rec/memorystage/000/lerobot_v21/meta/extra/episode_000000.json

@@ -534,7 +534,7 @@
   },
   "task": {
    "task_type": "MemoryStage",
-   "instruction": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.",
+   "instruction": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.",
    "template_id": -1,
    "slots": {
     "obj": 1.5707963267948966,

N053/go2_arm/dr_on/seed_2123/rec/memorystage/000/lerobot_v21/meta/tasks.jsonl

@@ -1 +1 @@
-{"task_index": 0, "task": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."}
+{"task_index": 0, "task": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."}

N053/go2_arm/dr_on/seed_2124/rec/check.json

@@ -23,7 +23,7 @@
   },
   "task text = MobileBench instruction": {
    "pass": true,
-   "detail": "task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'"
+   "detail": "task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'"
   },
   "rows = recorded control steps / 2 (= episode length at 25 Hz)": {
    "pass": true,

N053/go2_arm/dr_on/seed_2124/rec/check.txt

@@ -7,7 +7,7 @@ dataset root /mnt/scratch/sz850_atomic/n053_go2_on_2124_r1/out/rec/memorystage/0
 --- episode file episode_000000.parquet: 1248 rows, columns ['action', 'action_is_hold', 'observation.state', 'observation.joint_vel', 'observation.eef_pose', 'observation.environment_state', 'action.command', 'action.applied', 'narration', 'collect_info', 'extra.reward', 'extra.action_info', 'next.truncated', 'sim_step_idx', 'timestamp', 'frame_index', 'episode_index', 'index', 'task_index', 'next.done']
 [PASS] observation.state shape: parquet width 26, info shape [26], 26 names first: ['joint_0', 'joint_1', 'joint_2', 'joint_3', 'joint_4', 'joint_5']
 [PASS] action shape: parquet width 12, info shape [12], 12 names first: ['base_action_0', 'base_action_1', 'base_action_2', 'base_action_3', 'eef0.pos_x', 'eef0.pos_y']
-[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'
+[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'
 rows 1248 at dataset fps 25 = 49.92 s
 [PASS] rows = recorded control steps / 2 (= episode length at 25 Hz): 1248 rows vs ceil(2495/2) = 1248
 [PASS] rows are every 2nd control step (sim_step_idx): sim_step_idx 61..2555, step differences [2]

N053/go2_arm/dr_on/seed_2124/rec/memorystage/000/lerobot_v21/meta/episodes.jsonl

@@ -1 +1 @@
-{"episode_index": 0, "tasks": ["Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."], "length": 1248, "magicsim/control_steps": 2495, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is to the right of the rack. The white st booth r is on top of the rack. The white st booth r is to the left of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is to the left of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front-left, at table height."}
+{"episode_index": 0, "tasks": ["Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."], "length": 1248, "magicsim/control_steps": 2495, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is to the right of the rack. The white st booth r is on top of the rack. The white st booth r is to the left of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is to the left of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front-left, at table height."}

N053/go2_arm/dr_on/seed_2124/rec/memorystage/000/lerobot_v21/meta/extra/episode_000000.json

@@ -534,7 +534,7 @@
   },
   "task": {
    "task_type": "MemoryStage",
-   "instruction": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.",
+   "instruction": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.",
    "template_id": -1,
    "slots": {
     "obj": 1.5707963267948966,

N053/go2_arm/dr_on/seed_2124/rec/memorystage/000/lerobot_v21/meta/tasks.jsonl

@@ -1 +1 @@
-{"task_index": 0, "task": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."}
+{"task_index": 0, "task": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."}

N053/go2_arm/dr_on/seed_2125/rec/check.json

@@ -23,7 +23,7 @@
   },
   "task text = MobileBench instruction": {
    "pass": true,
-   "detail": "task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'"
+   "detail": "task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'"
   },
   "rows = recorded control steps / 2 (= episode length at 25 Hz)": {
    "pass": true,

N053/go2_arm/dr_on/seed_2125/rec/check.txt

@@ -7,7 +7,7 @@ dataset root /mnt/scratch/sz850_atomic/n053_go2_on_2125/out/rec/memorystage/000/
 --- episode file episode_000000.parquet: 1215 rows, columns ['action', 'action_is_hold', 'observation.state', 'observation.joint_vel', 'observation.eef_pose', 'observation.environment_state', 'action.command', 'action.applied', 'narration', 'collect_info', 'extra.reward', 'extra.action_info', 'next.truncated', 'sim_step_idx', 'timestamp', 'frame_index', 'episode_index', 'index', 'task_index', 'next.done']
 [PASS] observation.state shape: parquet width 26, info shape [26], 26 names first: ['joint_0', 'joint_1', 'joint_2', 'joint_3', 'joint_4', 'joint_5']
 [PASS] action shape: parquet width 12, info shape [12], 12 names first: ['base_action_0', 'base_action_1', 'base_action_2', 'base_action_3', 'eef0.pos_x', 'eef0.pos_y']
-[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'
+[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'
 rows 1215 at dataset fps 25 = 48.60 s
 [PASS] rows = recorded control steps / 2 (= episode length at 25 Hz): 1215 rows vs ceil(2430/2) = 1215
 [PASS] rows are every 2nd control step (sim_step_idx): sim_step_idx 61..2489, step differences [2]

N053/go2_arm/dr_on/seed_2125/rec/memorystage/000/lerobot_v21/meta/episodes.jsonl

@@ -1 +1 @@
-{"episode_index": 0, "tasks": ["Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."], "length": 1215, "magicsim/control_steps": 2430, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is behind the rack. The white st booth r is on top of the rack. The white st booth r is in front of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is in front of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front-left, at table height."}
+{"episode_index": 0, "tasks": ["Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."], "length": 1215, "magicsim/control_steps": 2430, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is behind the rack. The white st booth r is on top of the rack. The white st booth r is in front of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is in front of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front-left, at table height."}

N053/go2_arm/dr_on/seed_2125/rec/memorystage/000/lerobot_v21/meta/extra/episode_000000.json

@@ -534,7 +534,7 @@
   },
   "task": {
    "task_type": "MemoryStage",
-   "instruction": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.",
+   "instruction": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.",
    "template_id": -1,
    "slots": {
     "obj": 1.5707963267948966,

N053/go2_arm/dr_on/seed_2125/rec/memorystage/000/lerobot_v21/meta/tasks.jsonl

@@ -1 +1 @@
-{"task_index": 0, "task": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."}
+{"task_index": 0, "task": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."}

N053/go2_arm/dr_on/seed_2126/rec/check.json

@@ -23,7 +23,7 @@
   },
   "task text = MobileBench instruction": {
    "pass": true,
-   "detail": "task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'"
+   "detail": "task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'"
   },
   "rows = recorded control steps / 2 (= episode length at 25 Hz)": {
    "pass": true,

N053/go2_arm/dr_on/seed_2126/rec/check.txt

@@ -7,7 +7,7 @@ dataset root /mnt/scratch/sz850_atomic/n053_go2_on_2126_r1/out/rec/memorystage/0
 --- episode file episode_000000.parquet: 1161 rows, columns ['action', 'action_is_hold', 'observation.state', 'observation.joint_vel', 'observation.eef_pose', 'observation.environment_state', 'action.command', 'action.applied', 'narration', 'collect_info', 'extra.reward', 'extra.action_info', 'next.truncated', 'sim_step_idx', 'timestamp', 'frame_index', 'episode_index', 'index', 'task_index', 'next.done']
 [PASS] observation.state shape: parquet width 26, info shape [26], 26 names first: ['joint_0', 'joint_1', 'joint_2', 'joint_3', 'joint_4', 'joint_5']
 [PASS] action shape: parquet width 12, info shape [12], 12 names first: ['base_action_0', 'base_action_1', 'base_action_2', 'base_action_3', 'eef0.pos_x', 'eef0.pos_y']
-[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'
+[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'
 rows 1161 at dataset fps 25 = 46.44 s
 [PASS] rows = recorded control steps / 2 (= episode length at 25 Hz): 1161 rows vs ceil(2321/2) = 1161
 [PASS] rows are every 2nd control step (sim_step_idx): sim_step_idx 61..2381, step differences [2]

N053/go2_arm/dr_on/seed_2126/rec/memorystage/000/lerobot_v21/meta/episodes.jsonl

@@ -1 +1 @@
-{"episode_index": 0, "tasks": ["Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."], "length": 1161, "magicsim/control_steps": 2321, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is to the right of the rack. The white st booth r is on top of the rack. The white st booth r is to the left of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is to the left of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front-left, at table height."}
+{"episode_index": 0, "tasks": ["Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."], "length": 1161, "magicsim/control_steps": 2321, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is to the right of the rack. The white st booth r is on top of the rack. The white st booth r is to the left of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is to the left of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front-left, at table height."}

N053/go2_arm/dr_on/seed_2126/rec/memorystage/000/lerobot_v21/meta/extra/episode_000000.json

@@ -534,7 +534,7 @@
   },
   "task": {
    "task_type": "MemoryStage",
-   "instruction": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.",
+   "instruction": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.",
    "template_id": -1,
    "slots": {
     "obj": 1.5707963267948966,

N053/go2_arm/dr_on/seed_2126/rec/memorystage/000/lerobot_v21/meta/tasks.jsonl

@@ -1 +1 @@
-{"task_index": 0, "task": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."}
+{"task_index": 0, "task": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."}

N053/go2_arm/dr_on/seed_2127/rec/check.json

@@ -23,7 +23,7 @@
   },
   "task text = MobileBench instruction": {
    "pass": true,
-   "detail": "task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'"
+   "detail": "task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'"
   },
   "rows = recorded control steps / 2 (= episode length at 25 Hz)": {
    "pass": true,

N053/go2_arm/dr_on/seed_2127/rec/check.txt

@@ -7,7 +7,7 @@ dataset root /mnt/scratch/sz850_atomic/n053_go2_on_2127/out/rec/memorystage/000/
 --- episode file episode_000000.parquet: 1162 rows, columns ['action', 'action_is_hold', 'observation.state', 'observation.joint_vel', 'observation.eef_pose', 'observation.environment_state', 'action.command', 'action.applied', 'narration', 'collect_info', 'extra.reward', 'extra.action_info', 'next.truncated', 'sim_step_idx', 'timestamp', 'frame_index', 'episode_index', 'index', 'task_index', 'next.done']
 [PASS] observation.state shape: parquet width 26, info shape [26], 26 names first: ['joint_0', 'joint_1', 'joint_2', 'joint_3', 'joint_4', 'joint_5']
 [PASS] action shape: parquet width 12, info shape [12], 12 names first: ['base_action_0', 'base_action_1', 'base_action_2', 'base_action_3', 'eef0.pos_x', 'eef0.pos_y']
-[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'
+[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'
 rows 1162 at dataset fps 25 = 46.48 s
 [PASS] rows = recorded control steps / 2 (= episode length at 25 Hz): 1162 rows vs ceil(2324/2) = 1162
 [PASS] rows are every 2nd control step (sim_step_idx): sim_step_idx 61..2383, step differences [2]

N053/go2_arm/dr_on/seed_2127/rec/memorystage/000/lerobot_v21/meta/episodes.jsonl

@@ -1 +1 @@
-{"episode_index": 0, "tasks": ["Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."], "length": 1162, "magicsim/control_steps": 2324, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is to the left of the rack. The white st booth r is on top of the rack. The white st booth r is to the right of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is to the right of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front, at table height."}
+{"episode_index": 0, "tasks": ["Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."], "length": 1162, "magicsim/control_steps": 2324, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is to the left of the rack. The white st booth r is on top of the rack. The white st booth r is to the right of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is to the right of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front, at table height."}

N053/go2_arm/dr_on/seed_2127/rec/memorystage/000/lerobot_v21/meta/extra/episode_000000.json

@@ -534,7 +534,7 @@
   },
   "task": {
    "task_type": "MemoryStage",
-   "instruction": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.",
+   "instruction": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.",
    "template_id": -1,
    "slots": {
     "obj": 1.5707963267948966,

N053/go2_arm/dr_on/seed_2127/rec/memorystage/000/lerobot_v21/meta/tasks.jsonl

@@ -1 +1 @@
-{"task_index": 0, "task": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."}
+{"task_index": 0, "task": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."}

N053/go2_arm/dr_on/seed_2128/rec/check.json

@@ -23,7 +23,7 @@
   },
   "task text = MobileBench instruction": {
    "pass": true,
-   "detail": "task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'"
+   "detail": "task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'"
   },
   "rows = recorded control steps / 2 (= episode length at 25 Hz)": {
    "pass": true,

N053/go2_arm/dr_on/seed_2128/rec/check.txt

@@ -7,7 +7,7 @@ dataset root /mnt/scratch/sz850_atomic/n053_go2_on_2128/out/rec/memorystage/000/
 --- episode file episode_000000.parquet: 1140 rows, columns ['action', 'action_is_hold', 'observation.state', 'observation.joint_vel', 'observation.eef_pose', 'observation.environment_state', 'action.command', 'action.applied', 'narration', 'collect_info', 'extra.reward', 'extra.action_info', 'next.truncated', 'sim_step_idx', 'timestamp', 'frame_index', 'episode_index', 'index', 'task_index', 'next.done']
 [PASS] observation.state shape: parquet width 26, info shape [26], 26 names first: ['joint_0', 'joint_1', 'joint_2', 'joint_3', 'joint_4', 'joint_5']
 [PASS] action shape: parquet width 12, info shape [12], 12 names first: ['base_action_0', 'base_action_1', 'base_action_2', 'base_action_3', 'eef0.pos_x', 'eef0.pos_y']
-[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'; expected 'Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.'
+[PASS] task text = MobileBench instruction: task_index 0: 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'; expected 'Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.'
 rows 1140 at dataset fps 25 = 45.60 s
 [PASS] rows = recorded control steps / 2 (= episode length at 25 Hz): 1140 rows vs ceil(2280/2) = 1140
 [PASS] rows are every 2nd control step (sim_step_idx): sim_step_idx 61..2339, step differences [2]

N053/go2_arm/dr_on/seed_2128/rec/memorystage/000/lerobot_v21/meta/episodes.jsonl

@@ -1 +1 @@
-{"episode_index": 0, "tasks": ["Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."], "length": 1140, "magicsim/control_steps": 2280, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is to the left of the rack. The white st booth r is on top of the rack. The white st booth r is to the right of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is to the right of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front, at table height."}
+{"episode_index": 0, "tasks": ["Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."], "length": 1140, "magicsim/control_steps": 2280, "magicsim/subsample_every": 2, "magicsim/env_id": 0, "magicsim/trajectory_id": 0, "magicsim/scene_caption": "An Unitree Go2 EDU quadruped with a Piper arm is positioned at the workspace, facing the rack. There are a large white st wall, a large gray st table a, a large gray st table b, a large white st booth l, a large white st booth r, a medium rack, a small pen s0, a small pen s1, a large gray pad, a large red tag red and a large blue tag blue. The white st wall is on top of the rack. The white st wall is to the left of the rack. The gray st table b is to the right of the pen s1. The white st booth l is on top of the rack. The white st booth l is to the left of the rack. The white st booth r is on top of the rack. The white st booth r is to the right of the rack. The rack is on top of the gray st table a. The pen s0 is to the right of the red tag red. The pen s1 is to the right of the blue tag blue. The gray pad is to the right of the pen s1. The red tag red is to the left of the pen s0. The blue tag blue is to the left of the pen s1. The scene is viewed from front, at table height."}

N053/go2_arm/dr_on/seed_2128/rec/memorystage/000/lerobot_v21/meta/extra/episode_000000.json

@@ -534,7 +534,7 @@
   },
   "task": {
    "task_type": "MemoryStage",
-   "instruction": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it.",
+   "instruction": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad.",
    "template_id": -1,
    "slots": {
     "obj": 1.5707963267948966,

N053/go2_arm/dr_on/seed_2128/rec/memorystage/000/lerobot_v21/meta/tasks.jsonl

@@ -1 +1 @@
-{"task_index": 0, "task": "Watch which pen touches the red zone, then go around the partition and touch the target pad with the pen that did not touch it."}
+{"task_index": 0, "task": "Watch which pen touches the red panel, then go around the partition and put the pen that did not touch it on the target pad."}