hfd downloads only partial files for very large HF repositories
#14
by lumpin - opened
It seems siblings can return truncated/incomplete file lists for very large repositories, so hfd falls back to list_repo_tree. However, recursive listing with list_repo_tree is very slow on first traversal for large repos (~100k+ files).
As a result, hfd may only download part of the dataset without obvious errors.
Would it be possible to improve file list validation or large-repo traversal behavior?