Skip to content

[KMeans] Configure Workspace for Large Batch Sizes #2433

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

Open
tarang-jain wants to merge 2 commits into
base: main
Choose a base branch
from
+15 −2
Open

[KMeans] Configure Workspace for Large Batch Sizes #2433

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
7df2d18
update batching
tarang-jain Aug 11, 2026
3b432ee
update workspace conditional
tarang-jain Aug 11, 2026
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
17 changes: 15 additions & 2 deletions cpp/src/cluster/detail/kmeans.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,6 +24,7 @@
#include <raft/core/pinned_mdarray.hpp>
#include <raft/core/pinned_mdspan.hpp>
#include <raft/core/resource/cuda_stream.hpp>
#include <raft/core/resource/device_memory_resource.hpp>
#include <raft/core/resource/thrust_policy.hpp>
#include <raft/core/resources.hpp>
#include <raft/linalg/map.cuh>
Expand Down Expand Up @@ -696,15 +697,27 @@ void kmeans_fit(

rmm::device_uvector<char> batch_workspace(device_buffer_samples, stream);

auto batch_memory = raft::resource::get_workspace_resource_ref(handle);
if constexpr (!data_on_device) {
size_t batch_staging_bytes =
static_cast<size_t>(device_buffer_samples) * static_cast<size_t>(n_features) * sizeof(DataT);
if (weight_ptr != nullptr) {
batch_staging_bytes += static_cast<size_t>(device_buffer_samples) * sizeof(DataT);
}
if (batch_staging_bytes > raft::resource::get_workspace_free_bytes(handle)) {

@cjnolet cjnolet Aug 13, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

Why not just always use the large workspace? I think doing this conditionally creates an additional challenge for user debugging that we could avoid if we just use the same workspace resources all the time. Will let @achirkin comment here too.

@tarang-jain tarang-jain Aug 13, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

Thats a good point, especially since we do expect batches to be quite large ( > 20 GB or so per batch).

@achirkin achirkin Aug 17, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

The user may set up the large workspace to use a slower memory than the normal workspace (e.g. managed memory vs device pool - a setup we recommend and also set in benchmarks).
Therefore, please view this as an optimization: if access to arrays allocated via this resource is the bottleneck, we should keep it; otherwise, it's ok to use the large workspace by default.
In this case, we're talking about batching, so my understanding is having small enough batches is a normal behavior, whereas the switch to the large workspace is an edge case to make the algorithm not fail if there's not enough memory.

batch_memory = raft::resource::get_large_workspace_resource_ref(handle);
}
}

auto data_batches = cuvs::spatial::knn::detail::utils::make_batch_load_iterator<DataT>(
handle, X.data_handle(), n_samples, n_features, device_buffer_samples, stream);
handle, X.data_handle(), n_samples, n_features, device_buffer_samples, stream, batch_memory);
// Host-path weight batches: only materialized when weights are provided and
// the data resides on host
std::optional<cuvs::spatial::knn::detail::utils::batch_load_iterator_dyn<DataT>> weight_batches;
if constexpr (!data_on_device) {
if (weight_ptr != nullptr) {
weight_batches = cuvs::spatial::knn::detail::utils::make_batch_load_iterator<DataT>(
handle, weight_ptr, n_samples, IndexT{1}, device_buffer_samples, stream);
handle, weight_ptr, n_samples, IndexT{1}, device_buffer_samples, stream, batch_memory);
} else {
raft::matrix::fill(handle, batch_weights_buf.view(), DataT{1});
}
Expand Down
Loading