Open HeegerGao opened 1 month ago
Hi sorry, we're in the process of releasing the data -- feel free to ping this thread again in a week if there is not a link to the data
Hi, any progress on the data?
Hi sorry about the delay -- we should probably get it released this week.
Can you see if you can access the dataset at the gs bucket here?
gs://gresearch/robotics/language_table/captions/
Hi, thanks for your efforts. I can access the dataset now. However, I still have two questions:
gs://gresearch/robotics/language_table/captions/
and gs://gresearch/robotics/language_table/long_horizon/
?dataset_info.json
and features.json
files that can help me to load your data? Or can you provide an example script to load the data?The new dataset should have long horizon dataset annotations while the previous one doesn't.
I mean there are two new folders in gs
website. One is gs://gresearch/robotics/language_table/captions/
and the other one is gs://gresearch/robotics/language_table/long_horizon/
. They both seem to be the long horizon datasets. What's the difference between them?
The captions dataset should be more recent -- the long_horizon dataset was an incomplete upload.
Hi, VLP is a great work! I find that you said VLP uses "10000 long horizon trajectories" of the language-table environment to train. However, I can only find short-horizon data in the official language-table repo. Could you tell me how to get the long-horizon trajectory data for training?
Maybe I didn't find it in the language-table repo, if that's the case, I apologize. Thanks!