Missing episodes in raw droid dataset
#1
by
nikonikolov
- opened
I am going through the language annotations and there's a large number of episodes which I can't find on the original droid data in https://console.cloud.google.com/storage/browser/gresearch/robotics/droid_raw/1.0.1 . Example IDs are
IPRL+569c8b2a+2024-03-10-13h-51m-43s
PennPAL+c5f808b7+2023-10-19-00h-20m-19s
TRI+52ca9b6a+2024-02-28-10h-41m-29s
TRI+52ca9b6a+2024-03-13-10h-27m-47s
BVL+be856380+2024-02-01-00h-10m-07s
BVL is missing in the link. For TRI and IPRL there are many episodes with dates after the cutoff date in the link
The delta may come from two sources:
- we PII cleaned the data before release and it's possible some episodes got lost in the process
- there may have been a few episodes that came in after release that haven't made it onto the release bucket.
We will soon update the RLDS version of the dataset with all updated annotations and episodes. If you need access to the raw, pls email me [email protected]
KarlP
changed discussion status to
closed