Adding rearrange_pick_dataset_v1#1664
Conversation
|
Instead of adding a new entry, you should just replace the links and bump the version on the existing entry. The downloader will handle the versioning for users. |
|
After we bump the version then the tests on Habitat Lab will fail though @aclegg3. How do we resolve this dependency? We need the new dataset for the tests here to pass: facebookresearch/habitat-lab#774. |
I see, this new version doesn't support the current testing scenario in main. One option would be to leave this PR un-merged and edit the CI config in that Habitat-lab branch to grab this PR for the data download step before pulling and installing main. When we are ready to merge that PR, we'll also merge this and remove that extra step. |
|
I think we want the data to be on a different path than the previous version. I start to think it would be better to have 2 different entries for these 2 datasets (like in the last commit) |
aclegg3
left a comment
There was a problem hiding this comment.
I understand the reasoning here. We'll add a new entry now to support testing on existing main and the dev branch. Once everything is merged and it becomes irrelevant, we can remove the old entry.
Co-authored-by: Alexander Clegg <alexclegg@fb.com>
Co-authored-by: Alexander Clegg <alexclegg@fb.com>
Motivation and Context
Adding a v1 of the data set for the pick task using the new data generator schema.
How Has This Been Tested
Types of changes
Checklist