Linslab's picture
Add level2_rigid multi-task PI0.5 LoRA (16 level-2 tasks, 50k steps, mean success 0.173)
6418366 verified
Raw History Blame Contribute Delete
2.32 kB
{
"suite": "level2_rigid",
"suite_id": "l2rigid",
"description": "all level-2 rigid-body tasks",
"level": "level_2",
"step": 50000,
"campaign": "pi05_lv2sub_20261029",
"tasks": 16,
"mean_success_rate": 0.17250000000000004,
"episodes_per_task": 50,
"episode_seeds": "1000-1049",
"per_task": [
{
"task": "HangMug-v1",
"success_rate": 0.62,
"successes": 31,
"return_mean": 29.45346
},
{
"task": "PivotBook-v1",
"success_rate": 0.04,
"successes": 2,
"return_mean": 185.57872000000003
},
{
"task": "OpenBoxRetrieval-v1",
"success_rate": 0.02,
"successes": 1,
"return_mean": 180.47969999999998
},
{
"task": "KitAssembly-v1",
"success_rate": 0.06,
"successes": 3,
"return_mean": 342.5
},
{
"task": "GraspTwoBlocks-v1",
"success_rate": 0.56,
"successes": 28,
"return_mean": 222.97526
},
{
"task": "ReorientPen-v1",
"success_rate": 0.4,
"successes": 20,
"return_mean": 98.3238
},
{
"task": "DexOpenDoor-v1",
"success_rate": 0.38,
"successes": 19,
"return_mean": 383.57285999999993
},
{
"task": "HammerNailOut-v1",
"success_rate": 0.08,
"successes": 4,
"return_mean": -31.14895999999999
},
{
"task": "DualArmLiftBall-v1",
"success_rate": 0.2,
"successes": 10,
"return_mean": 25.69026000000001
},
{
"task": "DualArmBookPickPlace-v1",
"success_rate": 0.0,
"successes": 0,
"return_mean": 759.46292
},
{
"task": "DualArmPegInHole-v1",
"success_rate": 0.1,
"successes": 5,
"return_mean": 54.31842000000002
},
{
"task": "DualArmCleanupPotEnv-v1",
"success_rate": 0.0,
"successes": 0,
"return_mean": 348.3684799999999
},
{
"task": "DualArmOpenLaptop-v1",
"success_rate": 0.04,
"successes": 2,
"return_mean": 1606.3234400000001
},
{
"task": "DualArmThrowBlock-v1",
"success_rate": 0.22,
"successes": 11,
"return_mean": 2226.29098
},
{
"task": "DualArmJugglingBall-v1",
"success_rate": 0.04,
"successes": 2,
"return_mean": -4901.14318
},
{
"task": "DualArmOpenBottle-v1",
"success_rate": 0.0,
"successes": 0,
"return_mean": 2164.84324
}
],
"note": "Only the final checkpoint (step 50000) was evaluated; 50 episodes per task, the same seeds for every task and suite."
}