Fix seed init in clustering (#1156)

microsoft · Sep 18, 2024 · 1091079 · 1091079
1 parent 594084f
commit 1091079
Show file tree

Hide file tree

Showing 2 changed files with 8 additions and 1 deletion.
diff --git a/.semversioner/next-release/patch-20240918223351051336.json b/.semversioner/next-release/patch-20240918223351051336.json
@@ -0,0 +1,4 @@
+{
+  "type": "patch",
+  "description": "Fix seed hardcoded init"
+}
diff --git a/graphrag/index/verbs/graph/clustering/cluster_graph.py b/graphrag/index/verbs/graph/clustering/cluster_graph.py
@@ -71,6 +71,9 @@ def cluster_graph(
 
     num_total = len(output_df)
 
+    # Create a seed for this run (if not provided)
+    seed = strategy.get("seed", Random().randint(0, 0xFFFFFFFF))  # noqa S311
+
     # Go through each of the rows
     graph_level_pairs_column: list[list[tuple[int, str]]] = []
     for _, row in progress_iterable(
@@ -87,7 +90,7 @@ def cluster_graph(
                         cast(str, row[column]),
                         cast(Communities, row[community_map_to]),
                         level,
-                        seed=strategy.get("seed"),
+                        seed=seed,
                     )
                 )
             )