Skip to content

UMAP multi-gpu knn graph build docs #7019

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
rapids-bot merged 12 commits into from
Jul 28, 2025
Merged

UMAP multi-gpu knn graph build docs #7019

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
12 commits
Select commit Hold shift + click to select a range
08fbdad
handle docs
jinsolp Jul 18, 2025
82dd9d5
Merge branch 'rapidsai:branch-25.08' into umap-mg-docs
jinsolp Jul 18, 2025
687d2dc
Merge branch 'branch-25.08' into umap-mg-docs
jinsolp Jul 18, 2025
12e637e
Merge branch 'branch-25.08' into umap-mg-docs
jinsolp Jul 22, 2025
b9c2224
base test
jinsolp Jul 23, 2025
945652b
Merge branch 'branch-25.08' into umap-mg-docs
jinsolp Jul 23, 2025
2b162eb
rm example
jinsolp Jul 25, 2025
87b1451
fix
jinsolp Jul 28, 2025
5a530ca
Merge branch 'branch-25.08' into umap-mg-docs
jinsolp Jul 28, 2025
431a971
fix test
jinsolp Jul 28, 2025
b66d8e5
Merge branch 'umap-mg-docs' of https://github.com/jinsolp/cuml into u…
jinsolp Jul 28, 2025
bb5ea81
Merge branch 'branch-25.08' into umap-mg-docs
jinsolp Jul 28, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 4 additions & 1 deletion python/cuml/cuml/manifold/umap.pyx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -273,13 +273,16 @@ class UMAP(Base,
def on_train_end(self, embeddings):
print(embeddings.copy_to_host())

handle : cuml.Handle
handle : cuml.Handle or pylibraft.common.DeviceResourcesSNMG
Specifies the cuml.handle that holds internal CUDA state for
computations in this model. Most importantly, this specifies the CUDA
stream that will be used for the model's computations, so users can
run different models concurrently in different streams by creating
handles in several streams.
If it is None, a new one is created.
Using `pylibraft.common.DeviceResourcesSNMG` as the handle will run batched knn graph
Comment thread
jinsolp marked this conversation as resolved.
building using multiple GPUs. This will only be valid when `build_algo=nn_descent` and
`nnd_n_clusters > 1`.
verbose : int or boolean, default=False
Sets logging level. It must be one of `cuml.common.logger.level_*`.
See :ref:`verbosity-levels` for more info.
Expand Down
14 changes: 9 additions & 5 deletions python/cuml/cuml/tests/test_base.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -138,12 +138,16 @@ def get_param_doc(param_doc_obj, name: str):

base_item_doc = get_param_doc(base_doc_params, name)

# Ensure the docstring is identical
assert (
found_doc.type == base_item_doc.type
), "Docstring mismatch for {}".format(name)
if not (
found_doc.type.startswith("cuml.Handle")
and klass == cuml.manifold.umap.UMAP
):
# Ensure the docstring is identical
assert (
found_doc.type == base_item_doc.type
), "Docstring mismatch for {}".format(name)
Comment on lines +141 to +148
Copy link
Copy Markdown
Contributor

@csadorf csadorf Jul 28, 2025 •
edited
Loading

Choose a reason for hiding this comment

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

The need for the exceptional handling here hints at an abstraction issue. This test ensures that the base class parameters and its children have the same signature for parameters of the same name. We need the exception because this is no longer the case, the UMAP class has a different signature.

A "correct" fix could be to either actually change the signature of the Base class such that it generally accepts instances of either cuml.Handle or pylibraft.common.DeviceResourcesSNMG type or ensure that cuml.Handle is either dervied from DeviceResourcesSNMG or at least shares a common base class.

We should not make any attempt at fixing or addressing this in this PR since it was originally introduced in #6654.

I suggest that we maintain this state for now and evaluate independently whether a correction is needed.


assert " ".join(found_doc.desc) == " ".join(base_item_doc.desc)
assert " ".join(found_doc.desc) == " ".join(base_item_doc.desc)


@pytest.mark.parametrize("child_class", list(all_base_children.keys()))
Expand Down