- Descrizione :
Set di dati per il documento MT-Opt .
Pagina iniziale : https://karolhausman.github.io/mt-opt/
Codice sorgente :
tfds.robotics.mt_opt.MtOptVersioni :
-
1.0.0(impostazione predefinita): versione iniziale.
-
Dimensioni del download :
Unknown sizeCache automatica ( documentazione ): No
Chiavi supervisionate (Vedi
as_superviseddoc ):NoneFigura ( tfds.show_examples ): non supportato.
Citazione :
@misc{kalashnikov2021mtopt,
title={MT-Opt: Continuous Multi-Task Robotic Reinforcement Learning at Scale},
author={Dmitry Kalashnikov and Jacob Varley and Yevgen Chebotar and Benjamin Swanson and Rico Jonschkowski and Chelsea Finn and Sergey Levine and Karol Hausman},
year={2021},
eprint={2104.08212},
archivePrefix={arXiv},
primaryClass={cs.RO}
}
mt_opt/rlds (configurazione predefinita)
Descrizione della configurazione : questo set di dati contiene episodi di attività raccolti in una flotta di robot reali. Segue il formato RLDS per rappresentare passaggi ed episodi.
Dimensione del set di dati:
4.38 TiBDivisioni :
| Diviso | Esempi |
|---|---|
'train' | 920.165 |
- Struttura delle caratteristiche :
FeaturesDict({
'episode_id': string,
'skill': uint8,
'steps': Dataset({
'action': FeaturesDict({
'close_gripper': bool,
'open_gripper': bool,
'target_pose': Tensor(shape=(7,), dtype=float32),
'terminate': bool,
}),
'is_first': bool,
'is_last': bool,
'is_terminal': bool,
'observation': FeaturesDict({
'gripper_closed': bool,
'height_to_bottom': float32,
'image': Image(shape=(512, 640, 3), dtype=uint8),
'state_dense': Tensor(shape=(7,), dtype=float32),
}),
}),
'task_code': string,
})
- Documentazione delle funzionalità :
| Caratteristica | Classe | Forma | Tipo D | Descrizione |
|---|---|---|---|---|
| CaratteristicheDict | ||||
| episodio_id | Tensore | corda | ||
| abilità | Tensore | uint8 | ||
| passi | Set di dati | |||
| passi/azione | CaratteristicheDict | |||
| passi/azione/close_gripper | Tensore | bool | ||
| passi/azione/open_gripper | Tensore | bool | ||
| passaggi/azione/target_pose | Tensore | (7,) | galleggiante32 | |
| passaggi/azione/termina | Tensore | bool | ||
| passi/è_primo | Tensore | bool | ||
| passi/è_ultimo | Tensore | bool | ||
| passi/è_terminale | Tensore | bool | ||
| passi/osservazione | CaratteristicheDict | |||
| passi/osservazione/gripper_closed | Tensore | bool | ||
| passi/osservazione/altezza_dal_basso | Tensore | galleggiante32 | ||
| passi/osservazione/immagine | Immagine | (512, 640, 3) | uint8 | |
| passi/osservazione/state_dense | Tensore | (7,) | galleggiante32 | |
| task_code | Tensore | corda |
- Esempi ( tfds.as_dataframe ):
mt_opt/sd
Descrizione della configurazione : il set di dati dei rilevatori di successo che contiene definizioni curate dall'uomo del completamento delle attività.
Dimensione del set di dati:
548.56 GiBDivisioni :
| Diviso | Esempi |
|---|---|
'test' | 94.636 |
'train' | 380.234 |
- Struttura delle caratteristiche :
FeaturesDict({
'image_0': Image(shape=(512, 640, 3), dtype=uint8),
'image_1': Image(shape=(480, 640, 3), dtype=uint8),
'image_2': Image(shape=(480, 640, 3), dtype=uint8),
'success': bool,
'task_code': string,
})
- Documentazione delle funzionalità :
| Caratteristica | Classe | Forma | Tipo D | Descrizione |
|---|---|---|---|---|
| CaratteristicheDict | ||||
| immagine_0 | Immagine | (512, 640, 3) | uint8 | |
| immagine_1 | Immagine | (480, 640, 3) | uint8 | |
| immagine_2 | Immagine | (480, 640, 3) | uint8 | |
| successo | Tensore | bool | ||
| task_code | Tensore | corda |
- Esempi ( tfds.as_dataframe ):