Dataset Viewer (First 5GB)
Auto-converted to Parquet Duplicate
task
stringclasses
1 value
env_id
stringclasses
1 value
env_kwargs
dict
seed
int64
2.03B
2.03B
episode_seed
int64
3.03B
3.03B
episode
int64
0
27.7k
extra_state
dict
timestamp
stringlengths
19
26
history
listlengths
2
30
stats
dict
init_state
dict
init_state_hash
stringlengths
64
64
solve_kwargs
dict
version
stringclasses
1 value
pusht_ppo_1hz_success_30step_v0
pusht/easy
{ "max_steps": 30, "success_threshold": 0.9, "init_distribution": "maniskill_like", "control_hz": 1, "easy_mode": false, "observation_width": 96, "observation_height": 96, "visualization_width": 96, "visualization_height": 96, "action_codec": "norm2" }
2,026,042,400
3,026,042,400
0
{ "checkpoint_root": "/home/raychai/VisGym_3d/visgym_training/ppo_pusht/runs/pusht_ppo_cpu256_10m_1hz_msinit_delta64_norm2_20260502", "checkpoint_path": "/home/raychai/VisGym_3d/visgym_training/ppo_pusht/runs/pusht_ppo_cpu256_10m_1hz_msinit_delta64_norm2_20260502/best_model.pt", "observation_kind": "pusht_ppo_low...
2026-05-03T00:06:49.853064
[ { "step": 0, "prompt": "You are controlling a 2D PushT environment from a top-down image.\n\nVisual elements:\n- The blue circle is the controllable pusher/agent. It moves through the plane and physically pushes the object.\n- The gray T-shaped block is the object that must be moved.\n- The green T-shaped o...
{ "step": 9, "reward": 4.84470040136501, "terminated": true, "truncated": false, "success": true, "coverage": 0.9099659439366508 }
{ "agent_position": [ 157, 486.4 ], "agent_velocity": [ 0, 0 ], "block_position": [ 315.73188294875774, 348.29903260723984 ], "block_velocity": [ 0, 0 ], "block_angle": 1.2220005582148996, "block_angvel": 0, "goal_pose": [ 256, 256, 0.7853981633974483 ...
e7c491aef37fe63529cc188336fe61c9b3646d912ad96ed1cc24eafa7be4e436
{ "strategy": "ppo", "checkpoint": "best_model.pt", "max_total_steps": 30 }
visgym_pusht_ppo_success_30step_parallel_v1
pusht_ppo_1hz_success_30step_v0
pusht/easy
{"max_steps":30,"success_threshold":0.9,"init_distribution":"maniskill_like","control_hz":1,"easy_mo(...TRUNCATED)
2,026,042,400
3,026,042,656
1
{"checkpoint_root":"/home/raychai/VisGym_3d/visgym_training/ppo_pusht/runs/pusht_ppo_cpu256_10m_1hz_(...TRUNCATED)
2026-05-03T00:06:51.179599
[{"step":0,"prompt":"You are controlling a 2D PushT environment from a top-down image.\n\nVisual ele(...TRUNCATED)
{"step":28,"reward":16.516485852745646,"terminated":true,"truncated":false,"success":true,"coverage"(...TRUNCATED)
{"agent_position":[157.0,486.4],"agent_velocity":[0.0,0.0],"block_position":[198.35090677978536,259.(...TRUNCATED)
d0fc4670456eb555a18cebae6edfe6e4cec7fabe6d92832377bdf27d7d41aa55
{ "strategy": "ppo", "checkpoint": "best_model.pt", "max_total_steps": 30 }
visgym_pusht_ppo_success_30step_parallel_v1
pusht_ppo_1hz_success_30step_v0
pusht/easy
{"max_steps":30,"success_threshold":0.9,"init_distribution":"maniskill_like","control_hz":1,"easy_mo(...TRUNCATED)
2,026,042,400
3,026,042,720
2
{"checkpoint_root":"/home/raychai/VisGym_3d/visgym_training/ppo_pusht/runs/pusht_ppo_cpu256_10m_1hz_(...TRUNCATED)
2026-05-03T00:06:51.489006
[{"step":0,"prompt":"You are controlling a 2D PushT environment from a top-down image.\n\nVisual ele(...TRUNCATED)
{"step":14,"reward":7.47706213755413,"terminated":true,"truncated":false,"success":true,"coverage":0(...TRUNCATED)
{"agent_position":[157.0,486.4],"agent_velocity":[0.0,0.0],"block_position":[322.97902048632386,378.(...TRUNCATED)
9e8206d36cd7d69d338b6c92e2c39510b94640058b9e436f91cfbf7530efb2d9
{ "strategy": "ppo", "checkpoint": "best_model.pt", "max_total_steps": 30 }
visgym_pusht_ppo_success_30step_parallel_v1
pusht_ppo_1hz_success_30step_v0
pusht/easy
{"max_steps":30,"success_threshold":0.9,"init_distribution":"maniskill_like","control_hz":1,"easy_mo(...TRUNCATED)
2,026,042,400
3,026,043,040
3
{"checkpoint_root":"/home/raychai/VisGym_3d/visgym_training/ppo_pusht/runs/pusht_ppo_cpu256_10m_1hz_(...TRUNCATED)
2026-05-03T00:06:52.977068
[{"step":0,"prompt":"You are controlling a 2D PushT environment from a top-down image.\n\nVisual ele(...TRUNCATED)
{"step":23,"reward":11.084032144022013,"terminated":true,"truncated":false,"success":true,"coverage"(...TRUNCATED)
{"agent_position":[157.0,486.4],"agent_velocity":[0.0,0.0],"block_position":[243.15319963408626,305.(...TRUNCATED)
5aed2572c7c42c6f2a1afa76be98b63e47fd3eda2ca7f67854093a3b2aa0844e
{ "strategy": "ppo", "checkpoint": "best_model.pt", "max_total_steps": 30 }
visgym_pusht_ppo_success_30step_parallel_v1
pusht_ppo_1hz_success_30step_v0
pusht/easy
{"max_steps":30,"success_threshold":0.9,"init_distribution":"maniskill_like","control_hz":1,"easy_mo(...TRUNCATED)
2,026,042,400
3,026,043,104
4
{"checkpoint_root":"/home/raychai/VisGym_3d/visgym_training/ppo_pusht/runs/pusht_ppo_cpu256_10m_1hz_(...TRUNCATED)
2026-05-03T00:06:53.372236
[{"step":0,"prompt":"You are controlling a 2D PushT environment from a top-down image.\n\nVisual ele(...TRUNCATED)
{"step":19,"reward":13.55792356334266,"terminated":true,"truncated":false,"success":true,"coverage":(...TRUNCATED)
{"agent_position":[157.0,486.4],"agent_velocity":[0.0,0.0],"block_position":[327.9534014370455,395.9(...TRUNCATED)
1d11a838c1b91a649b98a44cc173ca52b0f8a99cd567fb3bf685942786ae0be7
{ "strategy": "ppo", "checkpoint": "best_model.pt", "max_total_steps": 30 }
visgym_pusht_ppo_success_30step_parallel_v1
pusht_ppo_1hz_success_30step_v0
pusht/easy
{"max_steps":30,"success_threshold":0.9,"init_distribution":"maniskill_like","control_hz":1,"easy_mo(...TRUNCATED)
2,026,042,400
3,026,043,168
5
{"checkpoint_root":"/home/raychai/VisGym_3d/visgym_training/ppo_pusht/runs/pusht_ppo_cpu256_10m_1hz_(...TRUNCATED)
2026-05-03T00:06:53.703349
[{"step":0,"prompt":"You are controlling a 2D PushT environment from a top-down image.\n\nVisual ele(...TRUNCATED)
{"step":13,"reward":5.905015010520333,"terminated":true,"truncated":false,"success":true,"coverage":(...TRUNCATED)
{"agent_position":[157.0,486.4],"agent_velocity":[0.0,0.0],"block_position":[254.90630518488106,249.(...TRUNCATED)
059e6fa6f51fa40a22df1f06eaa55e481960303f0b0486c8bd2c049fdbc3debb
{ "strategy": "ppo", "checkpoint": "best_model.pt", "max_total_steps": 30 }
visgym_pusht_ppo_success_30step_parallel_v1
pusht_ppo_1hz_success_30step_v0
pusht/easy
{"max_steps":30,"success_threshold":0.9,"init_distribution":"maniskill_like","control_hz":1,"easy_mo(...TRUNCATED)
2,026,042,400
3,026,043,232
6
{"checkpoint_root":"/home/raychai/VisGym_3d/visgym_training/ppo_pusht/runs/pusht_ppo_cpu256_10m_1hz_(...TRUNCATED)
2026-05-03T00:06:54.218732
[{"step":0,"prompt":"You are controlling a 2D PushT environment from a top-down image.\n\nVisual ele(...TRUNCATED)
{"step":22,"reward":11.696519684915412,"terminated":true,"truncated":false,"success":true,"coverage"(...TRUNCATED)
{"agent_position":[157.0,486.4],"agent_velocity":[0.0,0.0],"block_position":[250.27820849451356,343.(...TRUNCATED)
296ac72acced3d578b0f4ba4c9485a12947b6b4b67be80b2b47eb494088d168a
{ "strategy": "ppo", "checkpoint": "best_model.pt", "max_total_steps": 30 }
visgym_pusht_ppo_success_30step_parallel_v1
pusht_ppo_1hz_success_30step_v0
pusht/easy
{"max_steps":30,"success_threshold":0.9,"init_distribution":"maniskill_like","control_hz":1,"easy_mo(...TRUNCATED)
2,026,042,400
3,026,043,296
7
{"checkpoint_root":"/home/raychai/VisGym_3d/visgym_training/ppo_pusht/runs/pusht_ppo_cpu256_10m_1hz_(...TRUNCATED)
2026-05-03T00:06:54.591617
[{"step":0,"prompt":"You are controlling a 2D PushT environment from a top-down image.\n\nVisual ele(...TRUNCATED)
{"step":17,"reward":7.08774555623277,"terminated":true,"truncated":false,"success":true,"coverage":0(...TRUNCATED)
{"agent_position":[157.0,486.4],"agent_velocity":[0.0,0.0],"block_position":[332.9995112447327,313.6(...TRUNCATED)
245089ade118535828ebc4ead12db078d2b2f47f1203ea0babad051e30680fe7
{ "strategy": "ppo", "checkpoint": "best_model.pt", "max_total_steps": 30 }
visgym_pusht_ppo_success_30step_parallel_v1
pusht_ppo_1hz_success_30step_v0
pusht/easy
{"max_steps":30,"success_threshold":0.9,"init_distribution":"maniskill_like","control_hz":1,"easy_mo(...TRUNCATED)
2,026,042,400
3,026,043,488
8
{"checkpoint_root":"/home/raychai/VisGym_3d/visgym_training/ppo_pusht/runs/pusht_ppo_cpu256_10m_1hz_(...TRUNCATED)
2026-05-03T00:06:55.490970
[{"step":0,"prompt":"You are controlling a 2D PushT environment from a top-down image.\n\nVisual ele(...TRUNCATED)
{"step":18,"reward":10.993222263537474,"terminated":true,"truncated":false,"success":true,"coverage"(...TRUNCATED)
{"agent_position":[157.0,486.4],"agent_velocity":[0.0,0.0],"block_position":[283.0499424679924,391.7(...TRUNCATED)
d9e930373e3001a2e77a7542db13fb7f46a03ad36a5b83a99b325bf3091d074a
{ "strategy": "ppo", "checkpoint": "best_model.pt", "max_total_steps": 30 }
visgym_pusht_ppo_success_30step_parallel_v1
pusht_ppo_1hz_success_30step_v0
pusht/easy
{"max_steps":30,"success_threshold":0.9,"init_distribution":"maniskill_like","control_hz":1,"easy_mo(...TRUNCATED)
2,026,042,400
3,026,043,616
9
{"checkpoint_root":"/home/raychai/VisGym_3d/visgym_training/ppo_pusht/runs/pusht_ppo_cpu256_10m_1hz_(...TRUNCATED)
2026-05-03T00:06:56.646228
[{"step":0,"prompt":"You are controlling a 2D PushT environment from a top-down image.\n\nVisual ele(...TRUNCATED)
{"step":27,"reward":16.362086390695406,"terminated":true,"truncated":false,"success":true,"coverage"(...TRUNCATED)
{"agent_position":[157.0,486.4],"agent_velocity":[0.0,0.0],"block_position":[180.46587835371585,364.(...TRUNCATED)
5ecef39cd1ea7f3d4c72eec3247b5d3d4c87f9b50c3de0fe7866833ba5c0d05d
{ "strategy": "ppo", "checkpoint": "best_model.pt", "max_total_steps": 30 }
visgym_pusht_ppo_success_30step_parallel_v1
End of preview. Expand in Data Studio

pusht_96_norm2

96px PushT PPO successful trajectory dataset.

The trajectories are generated by a 1Hz PPO PushT solver with action codec norm2, rendered images in history[*].image, image_prev, and image_next at 96x96 and JPEG quality 85.

Splits

split records format
train 500,000 gzip-compressed JSONL
test 1,000 gzip-compressed JSONL

Train/test initial states are filtered to be disjoint by init_state_hash; see metadata/.

Coordinates in move actions use normalized [0, 1] coordinates rounded to exactly two decimal places; execution decodes them back to PushT 512x512 physical coordinates. Rendered images are 96px.

Target Hub repo: https://huggingface.co/datasets/novastar112/pusht_96_norm2

Downloads last month
37