Skip to content

Avoid deprecated jit_unspill in dask tests#5456

Merged
rapids-bot[bot] merged 2 commits intorapidsai:mainfrom
TomAugspurger:tom/jit-unspill-depr
Mar 10, 2026
Merged

Avoid deprecated jit_unspill in dask tests#5456
rapids-bot[bot] merged 2 commits intorapidsai:mainfrom
TomAugspurger:tom/jit-unspill-depr

Conversation

@TomAugspurger
Copy link
Contributor

@TomAugspurger TomAugspurger commented Mar 6, 2026

rapidsai/dask-cuda#1631 deprecated jit-unspilling in dask-cuda in favor of cudf's native spilling.

For now we match dask-cuda's default of None, which will avoid any warnings.

Next release, when the deprecation is enforced, we'll need to remove they keyword.

rapidsai/dask-cuda#1631 deprecated
jit-unspilling in dask-cuda in favor of cudf's native spilling.

For now we match dask-cuda's default of None.

Next release, when the deprecation is enforced, we'll need to remove
they keyword.
@TomAugspurger TomAugspurger requested a review from a team as a code owner March 6, 2026 16:41
rmm_pool_size=None,
dask_worker_devices=None,
jit_unspill=False,
jit_unspill=None,
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I wasn't 100% sure if cugraph.testing was part of the public API. If our tests are the only ones using this then we can just remove it.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm not aware of it being used elsewhere (@rlratzel is probably the main person who can say definitively), so yeah it's probably safer to keep it as is for now. Perhaps leave a breadcrumb note such as:

Suggested change
jit_unspill=None,
jit_unspill=None, # Unused (deprecated). Safe to remove if not used elsewhere

Also, I would suggest to similarly update jit_unspill in the docstring.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks, Tom and Erik. It is indeed safe to remove so I just did that.

@eriknw
Copy link
Contributor

eriknw commented Mar 6, 2026

Thanks @TomAugspurger. Would you recommend setting (or being able to set) enable_cudf_spill=True?

Also, it looks like the mention of JIT-unspill in the dask-cudf best practices doc could also be updated: https://docs.rapids.ai/api/dask-cudf/stable/best_practices/#enable-cudf-spilling

@TomAugspurger
Copy link
Contributor Author

Thanks for the link to the outdated docs. I'll get that fixed.

Would you recommend setting (or being able to set) enable_cudf_spill=True?

I personally don't have a recommendation, so using whatever the default is is probably for the best. API-design wise, I'd say that unless cugraph has a strong reason to prefer otherwise (because it knows something about the problem / domain that dask-cuda doesn't), I'd leave cluster creation up to the user / dask-cuda. But I'm not familiar with the history of cugraph so maybe there's a reason for the way things are now.

@jameslamb jameslamb added improvement Improvement / enhancement to an existing function non-breaking Non-breaking change labels Mar 6, 2026
Signed-off-by: rlratzel <rratzel@nvidia.com>
rmm_pool_size=None,
dask_worker_devices=None,
jit_unspill=False,
jit_unspill=None,
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks, Tom and Erik. It is indeed safe to remove so I just did that.

@rlratzel
Copy link
Contributor

/merge

@rapids-bot rapids-bot bot merged commit c5e7c01 into rapidsai:main Mar 10, 2026
77 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

improvement Improvement / enhancement to an existing function non-breaking Non-breaking change

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants