Merge branch 'fea/cuvs_c_standalone_ci' into pull-request/1438

Author: robertmaynard

.github/workflows/build.yaml

@@ -3,7 +3,7 @@ name: build
 on:
   push:
     branches:
-      - "branch-*"
+      - "main"
     tags:
       - v[0-9][0-9].[0-9][0-9].[0-9][0-9]
   workflow_dispatch:
@@ -34,7 +34,7 @@ concurrency:
 jobs:
   cpp-build:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-build.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-build.yaml@main
     with:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}
@@ -65,7 +65,7 @@ jobs:
   rust-build:
     needs: cpp-build
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@main
     # Artifacts are not published from these jobs, so it's safe to run for multiple CUDA versions.
     # If these jobs start producing artifacts, the names will have to differentiate between CUDA versions.
     strategy:
@@ -86,7 +86,7 @@ jobs:
   go-build:
     needs: cpp-build
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@main
     # Artifacts are not published from these jobs, so it's safe to run for multiple CUDA versions.
     # If these jobs start producing artifacts, the names will have to differentiate between CUDA versions.
     strategy:
@@ -107,7 +107,7 @@ jobs:
   java-build:
     needs: cpp-build
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@main
     # Artifacts are not published from these jobs, so it's safe to run for multiple CUDA versions.
     # If these jobs start producing artifacts, the names will have to differentiate between CUDA versions.
     strategy:
@@ -129,7 +129,7 @@ jobs:
   python-build:
     needs: [cpp-build]
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-python-build.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/conda-python-build.yaml@main
     with:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}
@@ -139,7 +139,7 @@ jobs:
   upload-conda:
     needs: [cpp-build, python-build]
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-upload-packages.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/conda-upload-packages.yaml@main
     with:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}
@@ -154,7 +154,7 @@ jobs:
     if: github.ref_type == 'branch'
     needs: python-build
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@main
     with:
       arch: "amd64"
       branch: ${{ inputs.branch }}
@@ -166,7 +166,7 @@ jobs:
       sha: ${{ inputs.sha }}
   wheel-build-libcuvs:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@main
     with:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}
@@ -180,7 +180,7 @@ jobs:
   wheel-publish-libcuvs:
     needs: wheel-build-libcuvs
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/wheels-publish.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/wheels-publish.yaml@main
     with:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}
@@ -191,7 +191,7 @@ jobs:
   wheel-build-cuvs:
     needs: wheel-build-libcuvs
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@main
     with:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}
@@ -203,7 +203,7 @@ jobs:
   wheel-publish-cuvs:
     needs: wheel-build-cuvs
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/wheels-publish.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/wheels-publish.yaml@main
     with:
       build_type: ${{ inputs.build_type || 'branch' }}
       branch: ${{ inputs.branch }}

.github/workflows/pr.yaml

@@ -31,7 +31,7 @@ jobs:
       - devcontainer
       - telemetry-setup
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/pr-builder.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/pr-builder.yaml@main
     if: always()
     with:
       needs: ${{ toJSON(needs) }}
@@ -58,7 +58,7 @@ jobs:
   changed-files:
     needs: telemetry-setup
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/changed-files.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/changed-files.yaml@main
     with:
       files_yaml: |
         test_cpp:
@@ -134,44 +134,44 @@ jobs:
   checks:
     needs: telemetry-setup
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/checks.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/checks.yaml@main
     with:
       enable_check_generated_files: false
       ignored_pr_jobs: "telemetry-summarize"
   conda-cpp-build:
     needs: checks
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-build.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-build.yaml@main
     with:
       build_type: pull-request
       node_type: cpu16
       script: ci/build_cpp.sh
   conda-cpp-tests:
     needs: [conda-cpp-build, changed-files]
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-tests.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-tests.yaml@main
     if: fromJSON(needs.changed-files.outputs.changed_file_groups).test_cpp
     with:
       build_type: pull-request
       script: ci/test_cpp.sh
   conda-cpp-checks:
     needs: conda-cpp-build
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-post-build-checks.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-post-build-checks.yaml@main
     with:
       build_type: pull-request
       symbol_exclusions: (void (thrust::|cub::))
   conda-python-build:
     needs: conda-cpp-build
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-python-build.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/conda-python-build.yaml@main
     with:
       build_type: pull-request
       script: ci/build_python.sh
   conda-python-tests:
     needs: [conda-python-build, changed-files]
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-python-tests.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/conda-python-tests.yaml@main
     if: fromJSON(needs.changed-files.outputs.changed_file_groups).test_python
     with:
       build_type: pull-request
@@ -262,7 +262,7 @@ jobs:
   conda-java-tests:
     needs: [conda-cpp-build, changed-files]
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@main
     if: fromJSON(needs.changed-files.outputs.changed_file_groups).test_java || fromJSON(needs.changed-files.outputs.changed_file_groups).test_cpp
     # Artifacts are not published from these jobs, so it's safe to run for multiple CUDA versions.
     # If these jobs start producing artifacts, the names will have to differentiate between CUDA versions.
@@ -283,7 +283,7 @@ jobs:
   docs-build:
     needs: conda-python-build
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@main
     with:
       build_type: pull-request
       node_type: "gpu-l4-latest-1"
@@ -293,7 +293,7 @@ jobs:
   rust-build:
     needs: [conda-cpp-build, changed-files]
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@main
     if: fromJSON(needs.changed-files.outputs.changed_file_groups).test_rust || fromJSON(needs.changed-files.outputs.changed_file_groups).test_cpp
     # Artifacts are not published from these jobs, so it's safe to run for multiple CUDA versions.
     # If these jobs start producing artifacts, the names will have to differentiate between CUDA versions.
@@ -331,7 +331,7 @@ jobs:
   wheel-build-libcuvs:
     needs: checks
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@main
     with:
       build_type: pull-request
       script: ci/build_wheel_libcuvs.sh
@@ -342,7 +342,7 @@ jobs:
   wheel-build-cuvs:
     needs: wheel-build-libcuvs
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/wheels-build.yaml@main
     with:
       build_type: pull-request
       script: ci/build_wheel_cuvs.sh
@@ -351,15 +351,15 @@ jobs:
   wheel-tests-cuvs:
     needs: [wheel-build-cuvs, changed-files]
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/wheels-test.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/wheels-test.yaml@main
     if: fromJSON(needs.changed-files.outputs.changed_file_groups).test_python
     with:
       build_type: pull-request
       script: ci/test_wheel_cuvs.sh
   devcontainer:
     secrets: inherit
     needs: telemetry-setup
-    uses: rapidsai/shared-workflows/.github/workflows/build-in-devcontainer.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/build-in-devcontainer.yaml@main
     with:
       arch: '["amd64", "arm64"]'
       cuda: '["13.0"]'

.github/workflows/test.yaml

@@ -25,7 +25,7 @@ on:
 jobs:
   conda-cpp-checks:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-post-build-checks.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-post-build-checks.yaml@main
     with:
       build_type: ${{ inputs.build_type }}
       branch: ${{ inputs.branch }}
@@ -34,7 +34,7 @@ jobs:
       symbol_exclusions: (void (thrust::|cub::))
   conda-cpp-tests:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-tests.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/conda-cpp-tests.yaml@main
     with:
       build_type: ${{ inputs.build_type }}
       branch: ${{ inputs.branch }}
@@ -43,7 +43,7 @@ jobs:
       sha: ${{ inputs.sha }}
   conda-python-tests:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/conda-python-tests.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/conda-python-tests.yaml@main
     with:
       build_type: ${{ inputs.build_type }}
       branch: ${{ inputs.branch }}
@@ -52,7 +52,7 @@ jobs:
       sha: ${{ inputs.sha }}
   conda-java-tests:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/custom-job.yaml@main
     # Artifacts are not published from these jobs, so it's safe to run for multiple CUDA versions.
     # If these jobs start producing artifacts, the names will have to differentiate between CUDA versions.
     strategy:
@@ -72,7 +72,7 @@ jobs:
       script: "ci/test_java.sh"
   wheel-tests-cuvs:
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/wheels-test.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/wheels-test.yaml@main
     with:
       build_type: ${{ inputs.build_type }}
       branch: ${{ inputs.branch }}

.github/workflows/trigger-breaking-change-alert.yaml

@@ -12,7 +12,7 @@ jobs:
   trigger-notifier:
     if: contains(github.event.pull_request.labels.*.name, 'breaking')
     secrets: inherit
-    uses: rapidsai/shared-workflows/.github/workflows/breaking-change-alert.yaml@branch-25.12
+    uses: rapidsai/shared-workflows/.github/workflows/breaking-change-alert.yaml@main
     with:
       sender_login: ${{ github.event.sender.login }}
       sender_avatar: ${{ github.event.sender.avatar_url }}

RAPIDS_BRANCH

@@ -1 +1 @@
-branch-25.12
+main

ci/build_standalone_c.sh

@@ -8,11 +8,9 @@ CMAKE_VERSION=3.31.8
 CMAKE_ARCH=x86_64
 
 dnf install -y \
-      gcc-toolset-${TOOLSET_VERSION} \
       patch \
       tar \
-      make \
-      wget
+      make
 
 # Fetch and install CMake.
 pushd /usr/local
@@ -40,7 +38,6 @@ RAPIDS_ARTIFACTS_DIR=${RAPIDS_ARTIFACTS_DIR:-"${PWD}/artifacts"}
 mkdir -p "${RAPIDS_ARTIFACTS_DIR}"
 export RAPIDS_ARTIFACTS_DIR
 
-
 scl enable gcc-toolset-${TOOLSET_VERSION} -- \
       cmake -S cpp -B cpp/build/temp/ \
             -DCMAKE_CUDA_HOST_COMPILER=/opt/rh/gcc-toolset-${TOOLSET_VERSION}/root/usr/bin/gcc \

cpp/include/cuvs/neighbors/scann.hpp

@@ -30,6 +30,7 @@
 #include <raft/util/integer_utils.hpp>
 #include <rmm/cuda_stream_view.hpp>
 
+#include <cmath>
 #include <optional>
 #include <variant>
 
@@ -74,9 +75,20 @@ struct index_params : cuvs::neighbors::index_params {
   uint32_t pq_train_iters = 10;
 
   /** whether to apply bf16 quantization of dataset vectors **/
-  bool bf16_enabled = false;
-
-  // TODO - add other scann build params
+  bool reordering_bf16 = false;
+
+  /** Threshold T for computing AVQ eta = (dim - 1) ( T^2 / || x ||^2) / ( 1 - T^2 / || x ||^2)
+   *
+   * When quantizing a vector x to x_q, AVQ minimizes the loss function
+   * L(x, x_q) = eta * || r_para ||^2 + || r_perp ||^2, where
+   * r = x - x_q, r_para = <r, x> * x / || x ||^2, r_perp = r - r_para
+   *
+   * Compared to L2 loss, This produces an x_q which better approximates
+   * the dot product of a query vector with x
+   *
+   * If the threshold is NAN, AVQ is not performed during bfloat16 quant
+   */
+  float reordering_noise_shaping_threshold = NAN;
 };
 
 /**
@@ -136,7 +148,7 @@ struct index : cuvs::neighbors::index {
         IdxT dim,
         uint32_t pq_clusters,
         uint32_t pq_num_subspaces,
-        bool bf16_enabled)
+        bool reordering_bf16)
     : cuvs::neighbors::index(),
       metric_(metric),
       pq_dim_(pq_dim),
@@ -154,7 +166,7 @@ struct index : cuvs::neighbors::index {
       n_rows_(n_rows),
       dim_(dim),
       bf16_dataset_(raft::make_host_matrix<int16_t, IdxT, raft::row_major>(
-        bf16_enabled ? n_rows : 0, bf16_enabled ? dim : 0))
+        reordering_bf16 ? n_rows : 0, reordering_bf16 ? dim : 0))
 
   {
   }
@@ -169,7 +181,7 @@ struct index : cuvs::neighbors::index {
             dim,
             1 << params.pq_bits,
             dim / params.pq_dim,
-            params.bf16_enabled)
+            params.reordering_bf16)
   {
     RAFT_EXPECTS(params.pq_bits == 4 || params.pq_bits == 8, "ScaNN only supports 4 or 8 bit PQ");
     RAFT_EXPECTS(dim >= params.pq_dim,
@@ -260,9 +272,21 @@ struct index : cuvs::neighbors::index {
   raft::device_matrix<float, uint32_t, raft::row_major> pq_codebook_;
   raft::host_matrix<uint8_t, IdxT, raft::row_major> quantized_residuals_;
   raft::host_matrix<uint8_t, IdxT, raft::row_major> quantized_soar_residuals_;
+
+  /* Internally, __nv_bfloat16 is used for float <-> __nv_bfloat16 conversion.
+   * The bits of __nv_bfloat16 are stored here reinterpreted as int16_t
+   *
+   * int16_t is used for two reaosns:
+   * * OSS ScaNN expects int16_t, so the serialzed bf16_dataset_ can be consumed
+   *   without any additional post-processing
+   * * For AVQ, we need to find the next bfloat16 number that is larger/smaller than a
+   *   given float. This is equivalent to incrementing/decrementing the mantissa
+   *   in IEEE representation of the bfloat16 number, which in turn is equivalent
+   *   to incrementing/decrementing the int16_t representation
+   */
   raft::host_matrix<int16_t, IdxT, raft::row_major> bf16_dataset_;
-  // TODO - add any data, pointers or structures needed
 };
+
 /**
  * @}
  */