{
  "version": 3,
  "state": "original observation.state stays 30D; action unchanged",
  "separate_fields": {
    "skill_onehot": {
      "dtype": "float32",
      "shape": [
        16
      ],
      "names": [
        "approach",
        "grasp",
        "retreat",
        "transport",
        "pour",
        "place",
        "stack",
        "return",
        "skill_reserved_08",
        "skill_reserved_09",
        "skill_reserved_10",
        "skill_reserved_11",
        "skill_reserved_12",
        "skill_reserved_13",
        "skill_reserved_14",
        "skill_reserved_15"
      ],
      "reserved_ids": [
        8,
        9,
        10,
        11,
        12,
        13,
        14,
        15
      ]
    },
    "pos_onehot": {
      "dtype": "float32",
      "shape": [
        16
      ],
      "names": [
        "none",
        "right1",
        "right2",
        "top1",
        "top2",
        "pos_reserved_05",
        "pos_reserved_06",
        "pos_reserved_07",
        "pos_reserved_08",
        "pos_reserved_09",
        "pos_reserved_10",
        "pos_reserved_11",
        "pos_reserved_12",
        "pos_reserved_13",
        "pos_reserved_14",
        "pos_reserved_15"
      ],
      "reserved_ids": [
        5,
        6,
        7,
        8,
        9,
        10,
        11,
        12,
        13,
        14,
        15
      ]
    },
    "obj_onehot": {
      "dtype": "float32",
      "shape": [
        16
      ],
      "names": [
        "none",
        "bread",
        "kettle",
        "cup",
        "obj_reserved_04",
        "obj_reserved_05",
        "obj_reserved_06",
        "obj_reserved_07",
        "obj_reserved_08",
        "obj_reserved_09",
        "obj_reserved_10",
        "obj_reserved_11",
        "obj_reserved_12",
        "obj_reserved_13",
        "obj_reserved_14",
        "obj_reserved_15"
      ],
      "reserved_ids": [
        4,
        5,
        6,
        7,
        8,
        9,
        10,
        11,
        12,
        13,
        14,
        15
      ]
    },
    "progress_fourier": {
      "dtype": "float32",
      "shape": [
        11
      ],
      "names": [
        "progress_path",
        "progress_sin_1pi",
        "progress_cos_1pi",
        "progress_sin_2pi",
        "progress_cos_2pi",
        "progress_sin_3pi",
        "progress_cos_3pi",
        "progress_sin_4pi",
        "progress_cos_4pi",
        "progress_sin_5pi",
        "progress_cos_5pi"
      ]
    }
  },
  "onehot_vocabularies": {
    "skill": {
      "approach": 0,
      "grasp": 1,
      "retreat": 2,
      "transport": 3,
      "pour": 4,
      "place": 5,
      "stack": 6,
      "return": 7
    },
    "pos": {
      "none": 0,
      "right1": 1,
      "right2": 2,
      "top1": 3,
      "top2": 4
    },
    "obj": {
      "none": 0,
      "bread": 1,
      "kettle": 2,
      "cup": 3
    }
  },
  "onehot_rule": "exactly one 1; none is an explicit category, not all zeros",
  "progress_source": "progress_path; cumulative joint path with existing gripper event weight",
  "fourier_formula": "[p, sin(pi*p), cos(pi*p), sin(2*pi*p), cos(2*pi*p), ..., sin(5*pi*p), cos(5*pi*p)]",
  "fourier_computation_dtype": "float64",
  "storage_dtype": "float32",
  "normalization": "none during encoding",
  "missing_obj": "Missing source obj is explicitly none/0 in encoded Parquet and CSV; obj_was_missing=1 preserves source missingness.",
  "scalar_labels": "source scalar labels retained; absent obj/obj_id normalized to none/0 in encoded Parquet and CSV; obj_was_missing in CSV records provenance",
  "csv": "labels.csv",
  "source_root": "/home/chw/桌面/proskill1017/dataset_encode",
  "dataset_layout": [
    "."
  ],
  "media": "Independent regular file copies; no symlinks or hardlinks",
  "onehot_capacity": 16,
  "extension_policy": "Assign new categories to unused IDs in 0..15; preserve all existing IDs; update codebook and names without resizing fields. Reserved slots are zero until assigned.",
  "merged_task_count": 6
}