RSN for removing cugraph-ops (#566)

Co-authored-by: Bradley Dice <bdice@bradleydice.com>
Co-authored-by: James Lamb <jaylamb20@gmail.com>

_notices/rsn0041.md

@@ -0,0 +1,35 @@
+---
+layout: notice
+parent: RAPIDS Support Notices
+grand_parent: RAPIDS Notices
+nav_exclude: true
+notice_type: rsn
+# Update meta-data for notice
+notice_id: 41 # should match notice number
+notice_pin: true # set to true to pin to notice page
+
+title: "Removing cugraph-ops from RAPIDS"
+notice_author: RAPIDS Ops
+notice_status: Completed
+notice_status_color: green
+# 'notice_status' and 'notice_status_color' combinations:
+#   "Proposal" - "blue"
+#   "Completed" - "green"
+#   "Review" - "purple"
+#   "In Progress" - "yellow"
+#   "Closed" - "red"
+notice_topic: Platform Support Change
+notice_rapids_version: "v25.02+"
+notice_created: 2025-01-10
+# 'notice_updated' should match 'notice_created' until an update is made
+notice_updated: 2025-01-10
+---
+
+## Overview
+
+RAPIDS will stop releasing `cugraph-ops` in 25.02, including the `libcugraphops` and `pylibcugraphops` packages. The graph sampling code previously used by cuGraph has been directly integrated into cuGraph. The equivariant neural network primitives have been migrated to BioNeMo, with the front-end package now available as [cuEquivariance](https://github.com/NVIDIA/cuEquivariance). The GNN aggregation primitives will be relocated to a new destination soon.
+
+
+## Impact
+
+The GNN convolutional layers in the `cugraph_dgl.nn` and `cugraph_pyg.nn` modules will be temporarily removed. They will be reintroduced at a later time.