d4rl_adroit_hammer

  • 説明:

D4RL は、オフライン強化学習のオープンソース ベンチマークです。トレーニングおよびベンチマーク アルゴリズム用の標準化された環境とデータセットを提供します。

データセットはRLDS 形式に従って、ステップとエピソードを表します。

@misc{fu2020d4rl,
    title={D4RL: Datasets for Deep Data-Driven Reinforcement Learning},
    author={Justin Fu and Aviral Kumar and Ofir Nachum and George Tucker and Sergey Levine},
    year={2020},
    eprint={2004.07219},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

d4rl_adroit_hammer/v0-human (デフォルト設定)

  • ダウンロードサイズ: 5.33 MiB

  • データセットのサイズ: 6.10 MiB

  • 自動キャッシュ(ドキュメント): はい

  • スプリット:

スプリット
'train' 70
  • 機能構造:
FeaturesDict({
    'steps': Dataset({
        'action': Tensor(shape=(26,), dtype=float32),
        'discount': float32,
        'infos': FeaturesDict({
            'qpos': Tensor(shape=(33,), dtype=float32),
            'qvel': Tensor(shape=(33,), dtype=float32),
        }),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': Tensor(shape=(46,), dtype=float32),
        'reward': float32,
    }),
})
  • 機能のドキュメント:
特徴クラスDtype説明
特徴辞書
ステップデータセット
ステップ/アクションテンソル(26) float32
ステップ/割引テンソルfloat32
手順/情報特徴辞書
ステップ/情報/qposテンソル(33) float32
ステップ/情報/qvelテンソル(33) float32
ステップ/is_firstテンソルブール
ステップ/is_lastテンソルブール
ステップ/is_terminalテンソルブール
ステップ/観察テンソル(46) float32
ステップ/報酬テンソルfloat32

d4rl_adroit_hammer/v0-クローン

  • ダウンロードサイズ: 644.69 MiB

  • データセットのサイズ: 538.97 MiB

  • 自動キャッシュ(ドキュメント): いいえ

  • スプリット:

スプリット
'train' 5,594
  • 機能構造:
FeaturesDict({
    'steps': Dataset({
        'action': Tensor(shape=(26,), dtype=float32),
        'discount': float64,
        'infos': FeaturesDict({
            'qpos': Tensor(shape=(33,), dtype=float64),
            'qvel': Tensor(shape=(33,), dtype=float64),
        }),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': Tensor(shape=(46,), dtype=float64),
        'reward': float64,
    }),
})
  • 機能のドキュメント:
特徴クラスDtype説明
特徴辞書
ステップデータセット
ステップ/アクションテンソル(26) float32
ステップ/割引テンソルfloat64
手順/情報特徴辞書
ステップ/情報/qposテンソル(33) float64
ステップ/情報/qvelテンソル(33) float64
ステップ/is_firstテンソルブール
ステップ/is_lastテンソルブール
ステップ/is_terminalテンソルブール
ステップ/観察テンソル(46) float64
ステップ/報酬テンソルfloat64

d4rl_adroit_hammer/v0-専門家

  • ダウンロードサイズ: 529.91 MiB

  • データセットのサイズ: 737.00 MiB

  • 自動キャッシュ(ドキュメント): いいえ

  • スプリット:

スプリット
'train' 5,000
  • 機能構造:
FeaturesDict({
    'steps': Dataset({
        'action': Tensor(shape=(26,), dtype=float32),
        'discount': float32,
        'infos': FeaturesDict({
            'action_logstd': Tensor(shape=(26,), dtype=float32),
            'action_mean': Tensor(shape=(26,), dtype=float32),
            'qpos': Tensor(shape=(33,), dtype=float32),
            'qvel': Tensor(shape=(33,), dtype=float32),
        }),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': Tensor(shape=(46,), dtype=float32),
        'reward': float32,
    }),
})
  • 機能のドキュメント:
特徴クラスDtype説明
特徴辞書
ステップデータセット
ステップ/アクションテンソル(26) float32
ステップ/割引テンソルfloat32
手順/情報特徴辞書
ステップ/情報/action_logstdテンソル(26) float32
ステップ/情報/action_meanテンソル(26) float32
ステップ/情報/qposテンソル(33) float32
ステップ/情報/qvelテンソル(33) float32
ステップ/is_firstテンソルブール
ステップ/is_lastテンソルブール
ステップ/is_terminalテンソルブール
ステップ/観察テンソル(46) float32
ステップ/報酬テンソルfloat32

d4rl_adroit_hammer/v1-人間

  • ダウンロードサイズ: 5.35 MiB

  • データセットのサイズ: 6.34 MiB

  • 自動キャッシュ(ドキュメント): はい

  • スプリット:

スプリット
'train' 25
  • 機能構造:
FeaturesDict({
    'steps': Dataset({
        'action': Tensor(shape=(26,), dtype=float32),
        'discount': float32,
        'infos': FeaturesDict({
            'board_pos': Tensor(shape=(3,), dtype=float32),
            'qpos': Tensor(shape=(33,), dtype=float32),
            'qvel': Tensor(shape=(33,), dtype=float32),
            'target_pos': Tensor(shape=(3,), dtype=float32),
        }),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': Tensor(shape=(46,), dtype=float32),
        'reward': float32,
    }),
})
  • 機能のドキュメント:
特徴クラスDtype説明
特徴辞書
ステップデータセット
ステップ/アクションテンソル(26) float32
ステップ/割引テンソルfloat32
手順/情報特徴辞書
ステップ/情報/board_posテンソル(3) float32
ステップ/情報/qposテンソル(33) float32
ステップ/情報/qvelテンソル(33) float32
ステップ/情報/target_posテンソル(3) float32
ステップ/is_firstテンソルブール
ステップ/is_lastテンソルブール
ステップ/is_terminalテンソルブール
ステップ/観察テンソル(46) float32
ステップ/報酬テンソルfloat32

d4rl_adroit_hammer/v1-クローン

  • ダウンロードサイズ: 425.93 MiB

  • データセットサイズ: 1.68 GiB

  • 自動キャッシュ(ドキュメント): いいえ

  • スプリット:

スプリット
'train' 3,606
  • 機能構造:
FeaturesDict({
    'algorithm': string,
    'policy': FeaturesDict({
        'fc0': FeaturesDict({
            'bias': Tensor(shape=(256,), dtype=float32),
            'weight': Tensor(shape=(46, 256), dtype=float32),
        }),
        'fc1': FeaturesDict({
            'bias': Tensor(shape=(256,), dtype=float32),
            'weight': Tensor(shape=(256, 256), dtype=float32),
        }),
        'last_fc': FeaturesDict({
            'bias': Tensor(shape=(26,), dtype=float32),
            'weight': Tensor(shape=(256, 26), dtype=float32),
        }),
        'nonlinearity': string,
        'output_distribution': string,
    }),
    'steps': Dataset({
        'action': Tensor(shape=(26,), dtype=float32),
        'discount': float32,
        'infos': FeaturesDict({
            'board_pos': Tensor(shape=(3,), dtype=float32),
            'qpos': Tensor(shape=(33,), dtype=float32),
            'qvel': Tensor(shape=(33,), dtype=float32),
            'target_pos': Tensor(shape=(3,), dtype=float32),
        }),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': Tensor(shape=(46,), dtype=float32),
        'reward': float32,
    }),
})
  • 機能のドキュメント:
特徴クラスDtype説明
特徴辞書
アルゴリズムテンソルストリング
ポリシー特徴辞書
ポリシー/fc0特徴辞書
ポリシー/fc0/バイアステンソル(256,) float32
ポリシー/fc0/重みテンソル(46, 256) float32
ポリシー/fc1特徴辞書
ポリシー/fc1/バイアステンソル(256,) float32
ポリシー/fc1/重量テンソル(256, 256) float32
ポリシー/last_fc特徴辞書
ポリシー/last_fc/バイアステンソル(26) float32
ポリシー/last_fc/重量テンソル(256, 26) float32
ポリシー/非線形性テンソルストリング
ポリシー/出力配布テンソルストリング
ステップデータセット
ステップ/アクションテンソル(26) float32
ステップ/割引テンソルfloat32
手順/情報特徴辞書
ステップ/情報/board_posテンソル(3) float32
ステップ/情報/qposテンソル(33) float32
ステップ/情報/qvelテンソル(33) float32
ステップ/情報/target_posテンソル(3) float32
ステップ/is_firstテンソルブール
ステップ/is_lastテンソルブール
ステップ/is_terminalテンソルブール
ステップ/観察テンソル(46) float32
ステップ/報酬テンソルfloat32

d4rl_adroit_hammer/v1-専門家

  • ダウンロードサイズ: 531.24 MiB

  • データセットのサイズ: 843.54 MiB

  • 自動キャッシュ(ドキュメント): いいえ

  • スプリット:

スプリット
'train' 5,000
  • 機能構造:
FeaturesDict({
    'algorithm': string,
    'policy': FeaturesDict({
        'fc0': FeaturesDict({
            'bias': Tensor(shape=(32,), dtype=float32),
            'weight': Tensor(shape=(32, 46), dtype=float32),
        }),
        'fc1': FeaturesDict({
            'bias': Tensor(shape=(32,), dtype=float32),
            'weight': Tensor(shape=(32, 32), dtype=float32),
        }),
        'last_fc': FeaturesDict({
            'bias': Tensor(shape=(26,), dtype=float32),
            'weight': Tensor(shape=(26, 32), dtype=float32),
        }),
        'last_fc_log_std': FeaturesDict({
            'bias': Tensor(shape=(26,), dtype=float32),
            'weight': Tensor(shape=(26, 32), dtype=float32),
        }),
        'nonlinearity': string,
        'output_distribution': string,
    }),
    'steps': Dataset({
        'action': Tensor(shape=(26,), dtype=float32),
        'discount': float32,
        'infos': FeaturesDict({
            'action_log_std': Tensor(shape=(26,), dtype=float32),
            'action_mean': Tensor(shape=(26,), dtype=float32),
            'board_pos': Tensor(shape=(3,), dtype=float32),
            'qpos': Tensor(shape=(33,), dtype=float32),
            'qvel': Tensor(shape=(33,), dtype=float32),
            'target_pos': Tensor(shape=(3,), dtype=float32),
        }),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': Tensor(shape=(46,), dtype=float32),
        'reward': float32,
    }),
})
  • 機能のドキュメント:
特徴クラスDtype説明
特徴辞書
アルゴリズムテンソルストリング
ポリシー特徴辞書
ポリシー/fc0特徴辞書
ポリシー/fc0/バイアステンソル(32) float32
ポリシー/fc0/重みテンソル(32, 46) float32
ポリシー/fc1特徴辞書
ポリシー/fc1/バイアステンソル(32) float32
ポリシー/fc1/重量テンソル(32, 32) float32
ポリシー/last_fc特徴辞書
ポリシー/last_fc/バイアステンソル(26) float32
ポリシー/last_fc/重量テンソル(26, 32) float32
ポリシー/last_fc_log_std特徴辞書
ポリシー/last_fc_log_std/バイアステンソル(26) float32
ポリシー/last_fc_log_std/重量テンソル(26, 32) float32
ポリシー/非線形性テンソルストリング
ポリシー/出力配布テンソルストリング
ステップデータセット
ステップ/アクションテンソル(26) float32
ステップ/割引テンソルfloat32
手順/情報特徴辞書
ステップ/情報/action_log_stdテンソル(26) float32
ステップ/情報/action_meanテンソル(26) float32
ステップ/情報/board_posテンソル(3) float32
ステップ/情報/qposテンソル(33) float32
ステップ/情報/qvelテンソル(33) float32
ステップ/情報/target_posテンソル(3) float32
ステップ/is_firstテンソルブール
ステップ/is_lastテンソルブール
ステップ/is_terminalテンソルブール
ステップ/観察テンソル(46) float32
ステップ/報酬テンソルfloat32