fix(dataset-tools): unable to merge local datasets in cache due to a …

[suessmann]

What this does

When using local root, the lerobot-edit-dataset with --operation.type=merge returned the FileNotFoundError. This was due to the fact that when specifying root to load a dataset in the generator [LeRobotDataset(repo_id, root=Path(cfg.root) for repo_id in cfg.operation.repo_ids], the root to the exact dataset was not actually passed.

This PR fixes that by adding [...] root=Path(cfg.root) / repo_id) for repo_id in [...].

How it was tested

Just run the merge and it worked, unlike before.

How to checkout & try? (for the reviewer)

Try merging with local root, such as

lerobot-edit-dataset \
    --repo_id ${HF_USER}/merged_dataset \
    --operation.type merge \
    --operation.repo_ids "['user/local_dataset_1', 'user/local_dataset_1']" \
    --root `root/to/local/cache`

[s1lent4gnt]

Hey @suessmann, thanks for the fix! One thing — Path(cfg.root) will raise TypeError when cfg.root is None. Small fix:

datasets = [
    LeRobotDataset(repo_id, root=Path(cfg.root) / repo_id if cfg.root else None)
    for repo_id in cfg.operation.repo_ids
]

[s1lent4gnt]

Superseded by #2369, Thanks @suessmann for the bug report and the fix!