ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

fix: preserve zero-valued dataset slice boundaries in embedding worker

Browse source 
Use explicit None checks instead of falsy `or` for slice_start and
slice_end so that a valid slice_end=0 is not replaced with the full
dataset length.
This commit is contained in:
Roland Tannous 2026-03-10 19:33:10 +00:00
commit c986174c56
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
studio/backend/core/training/worker.py
View file

@ -555,8 +555,8 @@ def _run_embedding_training(event_queue: Any, stop_queue: Any, config: dict) ->
slice_start = config.get("dataset_slice_start")
slice_end = config.get("dataset_slice_end")
if slice_start is not None or slice_end is not None:
start = slice_start or 0
end = slice_end or len(dataset)
start = slice_start if slice_start is not None else 0
end = slice_end if slice_end is not None else len(dataset)
dataset = dataset.select(range(start, min(end + 1, len(dataset))))
logger.info(f"Embedding dataset loaded: {len(dataset)} samples")