NVIDIA-NeMo / NVIDIA-NeMo/Curator
Text readers hang instead of failing fast when using `RayDataExecutor`
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 1.8k
- Forks
- 328
- Avg merge
- 4d 5h
- Merged PRs (30d)
- 30
Description
I noticed that if I am using the RayDataExecutor and I accidentally set the wrong input data path or the wrong file reader (e.g., try to use JsonlReader on a directory of Parquet files), it doesn't fail. It just runs forever with 0 tasks, 0 actors, 0 resources, 0B object store...
Minimal reproducer:
from nemo_curator.backends.experimental.ray_data import RayDataExecutor
from nemo_curator.stages.text.io.reader import JsonlReader
from nemo_curator.pipeline import Pipeline
pipeline = Pipeline(name="test")
# Incorrectly use JsonlReader on a directory of Parquet files
pipeline.add_stage(JsonlReader("/path/to/parquet/dir"))
executor = RayDataExecutor()
result = pipeline.run(executor=executor)
Note that the above example fails fast when using the XennaExecutor, which is the desired behavior.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by running the minimal reproducer with RayDataExecutor, JsonlReader, and Pipeline.run, then compare its behavior with XennaExecutor. Trace the RayDataExecutor path to identify why an invalid input path or reader leaves zero tasks running; done means the same cases fail fast instead of hanging.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- data-engineering, distributed-systems
- Issue type
- Bug
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Quiet
- Clarity
- Mostly clear
- Newbie friendliness
- 55/100