Automatic Partition Count Derivation for ACE #1603
Merged
+897
−153
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
- Added `cuvsHnswAceParams` structure for ACE configuration. - Implemented `cuvsHnswBuild` function to facilitate index construction using ACE. - Updated HNSW index parameters to include ACE settings. - Created new tests for HNSW index building and searching using ACE. - Updated documentation to reflect the new ACE parameters and usage.
- Add heuristic to automatically derive the number of partitions based on host and device memory requirements. - Increase the user-profided `npartitions` if it does not fit memory. - Introduced `max_host_memory_gb` and `max_gpu_memory_gb` fields to `cuvsAceParams` and `cuvsHnswAceParams` structures for controlling memory usage during ACE builds. - Added tests to verify that small memory limits trigger disk mode correctly for both CAGRA and HNSW index builds.
github-project-automation
bot
moved this to Todo
in Vector Search, ML, & Data Mining Release Board
|
Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually. Contributors can view more details about this message here. |
- Renamed parameter `m` to `M` in HNSW structures and related functions for consistency. - Removed `ef_construction` from `cuvsHnswAceParams` and related classes, as it is no longer needed. - Load the HNSW index from file before search if needed.
KyleFromNVIDIA
approved these changes
Dec 3, 2025
Member
KyleFromNVIDIA
left a comment
KyleFromNVIDIA
left a comment
There was a problem hiding this comment.
Approved CMake changes
tfeher
requested changes
Jan 15, 2026
Contributor
tfeher
left a comment
tfeher
left a comment
There was a problem hiding this comment.
Thanks Julian for the PR, it is great to have the n_partition parameter automatically determined. I have reviewed C/C++/Python changes. I have suggestions for the formulas used for memory estimation, but in general the PR is in a good shape.
tfeher
reviewed
Jan 15, 2026
- Update this once we have enough data to recommend this again.
Contributor
|
Thanks Julian for the update, I had one more comment for the workspace size estimate, otherwise the code looks good. |
tfeher
requested changes
Jan 19, 2026
Contributor
tfeher
left a comment
tfeher
left a comment
There was a problem hiding this comment.
Thanks Julian for the updates! Although the memory size estimates could be still improved, I believe it is better to tweak them in a follow up PR. The rest of the PR look good to me.
tfeher
approved these changes
Jan 19, 2026
Contributor
tfeher
left a comment
tfeher
left a comment
There was a problem hiding this comment.
My last review meant to approve the PR.
benfred
approved these changes
Jan 20, 2026
Member
benfred
left a comment
benfred
left a comment
There was a problem hiding this comment.
python/java changes look good!
| npartitions=1, | ||
| npartitions=0, | ||
| ef_construction=120, | ||
| build_dir="/tmp/ace_build", |
Member
There was a problem hiding this comment.
not related to this PR: but I don't think we should be hardcoding the build_dir like this, and instead should be using something like tempfile.TemporaryDirectory . The user might not have /tmp be a valid directory, and this could cause issues with concurrency when two programs both try to access the same directory.
Contributor
|
/merge |
github-project-automation
bot
moved this from In Progress
to Done
in Vector Search, ML, & Data Mining Release Board
lowener
pushed a commit
to lowener/cuvs
that referenced
this pull request
Jan 23, 2026
This PR adds automatic partition count derivation for ACE (Augmented Core Extraction) graph builds based on available system memory. Previously, users had to manually calculate and specify the number of partitions based on their dataset size and available memory. This was error-prone and required understanding the internal memory requirements of the ACE algorithm. With auto-derivation, users get optimal partitioning out of the box while still having the option to override if needed. **Changes** - When `npartitions` is set to `0` (new default), ACE automatically calculates the optimal number of partitions based on available host and GPU memory. - When `npartitions` is set to a positive value, the specified count is used but may be automatically increased if it would exceed memory limits. - Added `max_host_memory_gb` and `max_gpu_memory_gb` parameters to allow users to constrain memory usage (useful for shared systems or testing). This builds on top of PR rapidsai#1597, which should be merged first. Authors: - Julian Miller (https://github.com/julianmi) - Tamas Bela Feher (https://github.com/tfeher) Approvers: - Kyle Edwards (https://github.com/KyleFromNVIDIA) - Tamas Bela Feher (https://github.com/tfeher) - Ben Frederickson (https://github.com/benfred) URL: rapidsai#1603
lowener
added a commit
to lowener/cuvs
that referenced
this pull request
Jan 23, 2026
This reverts commit efe7c97.
lowener
added a commit
to lowener/cuvs
that referenced
this pull request
Jan 23, 2026
This reverts commit efe7c97.
lowener
pushed a commit
to lowener/cuvs
that referenced
this pull request
Jan 26, 2026
This PR adds automatic partition count derivation for ACE (Augmented Core Extraction) graph builds based on available system memory. Previously, users had to manually calculate and specify the number of partitions based on their dataset size and available memory. This was error-prone and required understanding the internal memory requirements of the ACE algorithm. With auto-derivation, users get optimal partitioning out of the box while still having the option to override if needed. **Changes** - When `npartitions` is set to `0` (new default), ACE automatically calculates the optimal number of partitions based on available host and GPU memory. - When `npartitions` is set to a positive value, the specified count is used but may be automatically increased if it would exceed memory limits. - Added `max_host_memory_gb` and `max_gpu_memory_gb` parameters to allow users to constrain memory usage (useful for shared systems or testing). This builds on top of PR rapidsai#1597, which should be merged first. Authors: - Julian Miller (https://github.com/julianmi) - Tamas Bela Feher (https://github.com/tfeher) Approvers: - Kyle Edwards (https://github.com/KyleFromNVIDIA) - Tamas Bela Feher (https://github.com/tfeher) - Ben Frederickson (https://github.com/benfred) URL: rapidsai#1603
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR adds automatic partition count derivation for ACE (Augmented Core Extraction) graph builds based on available system memory.
Previously, users had to manually calculate and specify the number of partitions based on their dataset size and available memory. This was error-prone and required understanding the internal memory requirements of the ACE algorithm. With auto-derivation, users get optimal partitioning out of the box while still having the option to override if needed.
Changes
npartitionsis set to0(new default), ACE automatically calculates the optimal number of partitions based on available host and GPU memory.npartitionsis set to a positive value, the specified count is used but may be automatically increased if it would exceed memory limits.max_host_memory_gbandmax_gpu_memory_gbparameters to allow users to constrain memory usage (useful for shared systems or testing).This builds on top of PR #1597, which should be merged first.