lerobot icon indicating copy to clipboard operation
lerobot copied to clipboard

Add script to remove episodes from dataset

Open Cadene opened this issue 1 year ago • 1 comments

What this does

  • Add lerobot/scripts/edit_dataset.py script with mode "remove"

TODO:

  • [ ] Add tests

How it was tested

Ran

# save
cp -r data/cadene/koch_bimanual_folding_2 \
cadene/koch_bimanual_folding_2/save

python lerobot/scripts/edit_dataset.py remove \
    --root data \
    --repo-id cadene/koch_bimanual_folding_2 \
    --episodes 0 4 7 10 34 54 69

python lerobot/scripts/visualize_dataset_html.py \
  --root data \
  --repo-id cadene/koch_bimanual_folding_2

Visualization look good.

How to checkout & try? (for the reviewer)

Cadene avatar Aug 24 '24 16:08 Cadene

@Cadene Is this still ongoing work?

I know there was recently a large refactor on the dataset format, so not sure if this code needs to change to address this.

At any rate, my team is looking to build out larger datasets, and are looking into editting and deleting episodes from datasets here, so we want this capability.

If this is not being worked on by anyone right now, we may be able to pick this up

brysonjones avatar Mar 02 '25 17:03 brysonjones

Related to: https://github.com/huggingface/lerobot/pull/831

imstevenpmwork avatar Apr 17 '25 14:04 imstevenpmwork

I'm closing this PR as it has been superseded by https://github.com/huggingface/lerobot/pull/2100

imstevenpmwork avatar Oct 07 '25 16:10 imstevenpmwork