- תיאור :
D4RL הוא אמת מידה בקוד פתוח ללמידת חיזוק לא מקוון. הוא מספק סביבות סטנדרטיות ומערכי נתונים עבור אלגוריתמי הדרכה ומידוד.
מערכי הנתונים פועלים לפי פורמט RLDS כדי לייצג שלבים ופרקים.
תיאור תצורה : ראה פרטים נוספים על המשימה וגרסאותיה ב- https://github.com/rail-berkeley/d4rl/wiki/Tasks#adroit
דף הבית : https://sites.google.com/view/d4rl/home
גרסאות :
-
1.0.0
: שחרור ראשוני. -
1.1.0
(ברירת מחדל): נוסף is_last.
-
מפתחות בפיקוח (ראה
as_supervised
doc ):None
איור ( tfds.show_examples ): לא נתמך.
ציטוט :
@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 (תצורת ברירת המחדל)
גודל הורדה :
5.33 MiB
גודל מערך נתונים :
6.10 MiB
שמור אוטומטי במטמון ( תיעוד ): כן
פיצולים :
לְפַצֵל | דוגמאות |
---|---|
'train' | 70 |
- מבנה תכונה :
FeaturesDict({
'steps': Dataset({
'action': Tensor(shape=(26,), dtype=tf.float32),
'discount': tf.float32,
'infos': FeaturesDict({
'qpos': Tensor(shape=(33,), dtype=tf.float32),
'qvel': Tensor(shape=(33,), dtype=tf.float32),
}),
'is_first': tf.bool,
'is_last': tf.bool,
'is_terminal': tf.bool,
'observation': Tensor(shape=(46,), dtype=tf.float32),
'reward': tf.float32,
}),
})
- תיעוד תכונה :
תכונה | מעמד | צוּרָה | Dtype | תיאור |
---|---|---|---|---|
FeaturesDict | ||||
צעדים | מערך נתונים | |||
צעדים/פעולה | מוֹתֵחַ | (26,) | tf.float32 | |
צעדים/הנחה | מוֹתֵחַ | tf.float32 | ||
שלבים/מידע | FeaturesDict | |||
צעדים/מידע/qpos | מוֹתֵחַ | (33,) | tf.float32 | |
צעדים/מידע/qvel | מוֹתֵחַ | (33,) | tf.float32 | |
צעדים/הוא_ראשון | מוֹתֵחַ | tf.bool | ||
צעדים/הוא_אחרון | מוֹתֵחַ | tf.bool | ||
steps/is_terminal | מוֹתֵחַ | tf.bool | ||
צעדים/תצפית | מוֹתֵחַ | (46,) | tf.float32 | |
צעדים/פרס | מוֹתֵחַ | tf.float32 |
- דוגמאות ( tfds.as_dataframe ):
d4rl_adroit_hammer/v0-cloned
גודל הורדה :
644.69 MiB
גודל מערך נתונים :
538.97 MiB
שמור אוטומטי במטמון ( תיעוד ): לא
פיצולים :
לְפַצֵל | דוגמאות |
---|---|
'train' | 5,594 |
- מבנה תכונה :
FeaturesDict({
'steps': Dataset({
'action': Tensor(shape=(26,), dtype=tf.float32),
'discount': tf.float64,
'infos': FeaturesDict({
'qpos': Tensor(shape=(33,), dtype=tf.float64),
'qvel': Tensor(shape=(33,), dtype=tf.float64),
}),
'is_first': tf.bool,
'is_last': tf.bool,
'is_terminal': tf.bool,
'observation': Tensor(shape=(46,), dtype=tf.float64),
'reward': tf.float64,
}),
})
- תיעוד תכונה :
תכונה | מעמד | צוּרָה | Dtype | תיאור |
---|---|---|---|---|
FeaturesDict | ||||
צעדים | מערך נתונים | |||
צעדים/פעולה | מוֹתֵחַ | (26,) | tf.float32 | |
צעדים/הנחה | מוֹתֵחַ | tf.float64 | ||
שלבים/מידע | FeaturesDict | |||
צעדים/מידע/qpos | מוֹתֵחַ | (33,) | tf.float64 | |
צעדים/מידע/qvel | מוֹתֵחַ | (33,) | tf.float64 | |
צעדים/הוא_ראשון | מוֹתֵחַ | tf.bool | ||
צעדים/הוא_אחרון | מוֹתֵחַ | tf.bool | ||
steps/is_terminal | מוֹתֵחַ | tf.bool | ||
צעדים/תצפית | מוֹתֵחַ | (46,) | tf.float64 | |
צעדים/פרס | מוֹתֵחַ | tf.float64 |
- דוגמאות ( tfds.as_dataframe ):
d4rl_adroit_hammer/v0-expert
גודל הורדה :
529.91 MiB
גודל מערך נתונים :
737.00 MiB
שמור אוטומטי במטמון ( תיעוד ): לא
פיצולים :
לְפַצֵל | דוגמאות |
---|---|
'train' | 5,000 |
- מבנה תכונה :
FeaturesDict({
'steps': Dataset({
'action': Tensor(shape=(26,), dtype=tf.float32),
'discount': tf.float32,
'infos': FeaturesDict({
'action_logstd': Tensor(shape=(26,), dtype=tf.float32),
'action_mean': Tensor(shape=(26,), dtype=tf.float32),
'qpos': Tensor(shape=(33,), dtype=tf.float32),
'qvel': Tensor(shape=(33,), dtype=tf.float32),
}),
'is_first': tf.bool,
'is_last': tf.bool,
'is_terminal': tf.bool,
'observation': Tensor(shape=(46,), dtype=tf.float32),
'reward': tf.float32,
}),
})
- תיעוד תכונה :
תכונה | מעמד | צוּרָה | Dtype | תיאור |
---|---|---|---|---|
FeaturesDict | ||||
צעדים | מערך נתונים | |||
צעדים/פעולה | מוֹתֵחַ | (26,) | tf.float32 | |
צעדים/הנחה | מוֹתֵחַ | tf.float32 | ||
שלבים/מידע | FeaturesDict | |||
steps/infos/action_logstd | מוֹתֵחַ | (26,) | tf.float32 | |
steps/infos/action_mean | מוֹתֵחַ | (26,) | tf.float32 | |
צעדים/מידע/qpos | מוֹתֵחַ | (33,) | tf.float32 | |
צעדים/מידע/qvel | מוֹתֵחַ | (33,) | tf.float32 | |
צעדים/הוא_ראשון | מוֹתֵחַ | tf.bool | ||
צעדים/הוא_אחרון | מוֹתֵחַ | tf.bool | ||
steps/is_terminal | מוֹתֵחַ | tf.bool | ||
צעדים/תצפית | מוֹתֵחַ | (46,) | tf.float32 | |
צעדים/פרס | מוֹתֵחַ | tf.float32 |
- דוגמאות ( tfds.as_dataframe ):
d4rl_adroit_hammer/v1-human
גודל הורדה :
5.35 MiB
גודל ערכת נתונים :
6.34 MiB
שמור אוטומטי במטמון ( תיעוד ): כן
פיצולים :
לְפַצֵל | דוגמאות |
---|---|
'train' | 25 |
- מבנה תכונה :
FeaturesDict({
'steps': Dataset({
'action': Tensor(shape=(26,), dtype=tf.float32),
'discount': tf.float32,
'infos': FeaturesDict({
'board_pos': Tensor(shape=(3,), dtype=tf.float32),
'qpos': Tensor(shape=(33,), dtype=tf.float32),
'qvel': Tensor(shape=(33,), dtype=tf.float32),
'target_pos': Tensor(shape=(3,), dtype=tf.float32),
}),
'is_first': tf.bool,
'is_last': tf.bool,
'is_terminal': tf.bool,
'observation': Tensor(shape=(46,), dtype=tf.float32),
'reward': tf.float32,
}),
})
- תיעוד תכונה :
תכונה | מעמד | צוּרָה | Dtype | תיאור |
---|---|---|---|---|
FeaturesDict | ||||
צעדים | מערך נתונים | |||
צעדים/פעולה | מוֹתֵחַ | (26,) | tf.float32 | |
צעדים/הנחה | מוֹתֵחַ | tf.float32 | ||
שלבים/מידע | FeaturesDict | |||
steps/infos/board_pos | מוֹתֵחַ | (3,) | tf.float32 | |
צעדים/מידע/qpos | מוֹתֵחַ | (33,) | tf.float32 | |
צעדים/מידע/qvel | מוֹתֵחַ | (33,) | tf.float32 | |
steps/infos/target_pos | מוֹתֵחַ | (3,) | tf.float32 | |
צעדים/הוא_ראשון | מוֹתֵחַ | tf.bool | ||
צעדים/הוא_אחרון | מוֹתֵחַ | tf.bool | ||
steps/is_terminal | מוֹתֵחַ | tf.bool | ||
צעדים/תצפית | מוֹתֵחַ | (46,) | tf.float32 | |
צעדים/פרס | מוֹתֵחַ | tf.float32 |
- דוגמאות ( tfds.as_dataframe ):
d4rl_adroit_hammer/v1-cloned
גודל הורדה :
425.93 MiB
גודל מערך נתונים :
1.68 GiB
שמור אוטומטי במטמון ( תיעוד ): לא
פיצולים :
לְפַצֵל | דוגמאות |
---|---|
'train' | 3,606 |
- מבנה תכונה :
FeaturesDict({
'algorithm': tf.string,
'policy': FeaturesDict({
'fc0': FeaturesDict({
'bias': Tensor(shape=(256,), dtype=tf.float32),
'weight': Tensor(shape=(46, 256), dtype=tf.float32),
}),
'fc1': FeaturesDict({
'bias': Tensor(shape=(256,), dtype=tf.float32),
'weight': Tensor(shape=(256, 256), dtype=tf.float32),
}),
'last_fc': FeaturesDict({
'bias': Tensor(shape=(26,), dtype=tf.float32),
'weight': Tensor(shape=(256, 26), dtype=tf.float32),
}),
'nonlinearity': tf.string,
'output_distribution': tf.string,
}),
'steps': Dataset({
'action': Tensor(shape=(26,), dtype=tf.float32),
'discount': tf.float32,
'infos': FeaturesDict({
'board_pos': Tensor(shape=(3,), dtype=tf.float32),
'qpos': Tensor(shape=(33,), dtype=tf.float32),
'qvel': Tensor(shape=(33,), dtype=tf.float32),
'target_pos': Tensor(shape=(3,), dtype=tf.float32),
}),
'is_first': tf.bool,
'is_last': tf.bool,
'is_terminal': tf.bool,
'observation': Tensor(shape=(46,), dtype=tf.float32),
'reward': tf.float32,
}),
})
- תיעוד תכונה :
תכונה | מעמד | צוּרָה | Dtype | תיאור |
---|---|---|---|---|
FeaturesDict | ||||
אַלגוֹרִיתְם | מוֹתֵחַ | tf.string | ||
מְדִינִיוּת | FeaturesDict | |||
מדיניות/fc0 | FeaturesDict | |||
policy/fc0/bias | מוֹתֵחַ | (256,) | tf.float32 | |
policy/fc0/weight | מוֹתֵחַ | (46, 256) | tf.float32 | |
מדיניות/fc1 | FeaturesDict | |||
policy/fc1/bias | מוֹתֵחַ | (256,) | tf.float32 | |
policy/fc1/weight | מוֹתֵחַ | (256, 256) | tf.float32 | |
policy/last_fc | FeaturesDict | |||
policy/last_fc/bias | מוֹתֵחַ | (26,) | tf.float32 | |
policy/last_fc/weight | מוֹתֵחַ | (256, 26) | tf.float32 | |
מדיניות/אי-לינאריות | מוֹתֵחַ | tf.string | ||
מדיניות/הפצה_תפוקה | מוֹתֵחַ | tf.string | ||
צעדים | מערך נתונים | |||
צעדים/פעולה | מוֹתֵחַ | (26,) | tf.float32 | |
צעדים/הנחה | מוֹתֵחַ | tf.float32 | ||
שלבים/מידע | FeaturesDict | |||
steps/infos/board_pos | מוֹתֵחַ | (3,) | tf.float32 | |
צעדים/מידע/qpos | מוֹתֵחַ | (33,) | tf.float32 | |
צעדים/מידע/qvel | מוֹתֵחַ | (33,) | tf.float32 | |
steps/infos/target_pos | מוֹתֵחַ | (3,) | tf.float32 | |
צעדים/הוא_ראשון | מוֹתֵחַ | tf.bool | ||
צעדים/הוא_אחרון | מוֹתֵחַ | tf.bool | ||
steps/is_terminal | מוֹתֵחַ | tf.bool | ||
צעדים/תצפית | מוֹתֵחַ | (46,) | tf.float32 | |
צעדים/פרס | מוֹתֵחַ | tf.float32 |
- דוגמאות ( tfds.as_dataframe ):
d4rl_adroit_hammer/v1-expert
גודל הורדה :
531.24 MiB
גודל ערכת נתונים:
843.54 MiB
שמור אוטומטי במטמון ( תיעוד ): לא
פיצולים :
לְפַצֵל | דוגמאות |
---|---|
'train' | 5,000 |
- מבנה תכונה :
FeaturesDict({
'algorithm': tf.string,
'policy': FeaturesDict({
'fc0': FeaturesDict({
'bias': Tensor(shape=(32,), dtype=tf.float32),
'weight': Tensor(shape=(32, 46), dtype=tf.float32),
}),
'fc1': FeaturesDict({
'bias': Tensor(shape=(32,), dtype=tf.float32),
'weight': Tensor(shape=(32, 32), dtype=tf.float32),
}),
'last_fc': FeaturesDict({
'bias': Tensor(shape=(26,), dtype=tf.float32),
'weight': Tensor(shape=(26, 32), dtype=tf.float32),
}),
'last_fc_log_std': FeaturesDict({
'bias': Tensor(shape=(26,), dtype=tf.float32),
'weight': Tensor(shape=(26, 32), dtype=tf.float32),
}),
'nonlinearity': tf.string,
'output_distribution': tf.string,
}),
'steps': Dataset({
'action': Tensor(shape=(26,), dtype=tf.float32),
'discount': tf.float32,
'infos': FeaturesDict({
'action_log_std': Tensor(shape=(26,), dtype=tf.float32),
'action_mean': Tensor(shape=(26,), dtype=tf.float32),
'board_pos': Tensor(shape=(3,), dtype=tf.float32),
'qpos': Tensor(shape=(33,), dtype=tf.float32),
'qvel': Tensor(shape=(33,), dtype=tf.float32),
'target_pos': Tensor(shape=(3,), dtype=tf.float32),
}),
'is_first': tf.bool,
'is_last': tf.bool,
'is_terminal': tf.bool,
'observation': Tensor(shape=(46,), dtype=tf.float32),
'reward': tf.float32,
}),
})
- תיעוד תכונה :
תכונה | מעמד | צוּרָה | Dtype | תיאור |
---|---|---|---|---|
FeaturesDict | ||||
אַלגוֹרִיתְם | מוֹתֵחַ | tf.string | ||
מְדִינִיוּת | FeaturesDict | |||
מדיניות/fc0 | FeaturesDict | |||
policy/fc0/bias | מוֹתֵחַ | (32,) | tf.float32 | |
policy/fc0/weight | מוֹתֵחַ | (32, 46) | tf.float32 | |
מדיניות/fc1 | FeaturesDict | |||
policy/fc1/bias | מוֹתֵחַ | (32,) | tf.float32 | |
policy/fc1/weight | מוֹתֵחַ | (32, 32) | tf.float32 | |
policy/last_fc | FeaturesDict | |||
policy/last_fc/bias | מוֹתֵחַ | (26,) | tf.float32 | |
policy/last_fc/weight | מוֹתֵחַ | (26, 32) | tf.float32 | |
policy/last_fc_log_std | FeaturesDict | |||
policy/last_fc_log_std/bias | מוֹתֵחַ | (26,) | tf.float32 | |
policy/last_fc_log_std/weight | מוֹתֵחַ | (26, 32) | tf.float32 | |
מדיניות/אי-לינאריות | מוֹתֵחַ | tf.string | ||
מדיניות/הפצה_תפוקה | מוֹתֵחַ | tf.string | ||
צעדים | מערך נתונים | |||
צעדים/פעולה | מוֹתֵחַ | (26,) | tf.float32 | |
צעדים/הנחה | מוֹתֵחַ | tf.float32 | ||
שלבים/מידע | FeaturesDict | |||
steps/infos/action_log_std | מוֹתֵחַ | (26,) | tf.float32 | |
steps/infos/action_mean | מוֹתֵחַ | (26,) | tf.float32 | |
steps/infos/board_pos | מוֹתֵחַ | (3,) | tf.float32 | |
צעדים/מידע/qpos | מוֹתֵחַ | (33,) | tf.float32 | |
צעדים/מידע/qvel | מוֹתֵחַ | (33,) | tf.float32 | |
steps/infos/target_pos | מוֹתֵחַ | (3,) | tf.float32 | |
צעדים/הוא_ראשון | מוֹתֵחַ | tf.bool | ||
צעדים/הוא_אחרון | מוֹתֵחַ | tf.bool | ||
steps/is_terminal | מוֹתֵחַ | tf.bool | ||
צעדים/תצפית | מוֹתֵחַ | (46,) | tf.float32 | |
צעדים/פרס | מוֹתֵחַ | tf.float32 |
- דוגמאות ( tfds.as_dataframe ):