Improve the logic to generate vectors.

Also adds some warning messages for easy of use.
weaviate · Nov 7, 2024 · f11fcca · f11fcca
1 parent a9f7c4a
commit f11fcca
Show file tree

Hide file tree

Showing 3 changed files with 20 additions and 2 deletions.
diff --git a/test/unittests/test_managers/test_data_manager.py b/test/unittests/test_managers/test_data_manager.py
@@ -24,7 +24,7 @@ def test_ingest_data(mock_client):
         consistency_level="quorum",
         randomize=True,
         auto_tenants=0,
-        vector_dimensions=None,
+        vector_dimensions=1536,
     )
 
     mock_client.collections.get.assert_called_once_with("TestCollection")

diff --git a/weaviate_cli/commands/create.py b/weaviate_cli/commands/create.py
@@ -254,6 +254,12 @@ def create_data_cli(
 ):
     """Ingest data into a collection in Weaviate."""
 
+    if vector_dimensions != 1536 and not randomize:
+        click.echo(
+            "Error: --vector_dimensions has no effect unless --randomize is enabled."
+        )
+        sys.exit(1)
+
     client = None
     try:
         client = get_client_from_context(ctx)

diff --git a/weaviate_cli/managers/data_manager.py b/weaviate_cli/managers/data_manager.py
@@ -132,14 +132,26 @@ def __ingest_data(
             cl_collection = collection.with_consistency_level(cl)
             vectorizer = cl_collection.config.get().vectorizer
             if vectorizer == "text2vec-contextionary":
+                (
+                    print("Warning: Using vector dimensions: 300")
+                    if vector_dimensions != 1536
+                    else None
+                )
                 vector_dimensions = 300
             elif vectorizer == "text2vec-transformers":
+                (
+                    print("Warning: Using vector dimensions: 768")
+                    if vector_dimensions != 1536
+                    else None
+                )
                 vector_dimensions = 768
             with cl_collection.batch.dynamic() as batch:
                 for obj in data_objects:
                     batch.add_object(
                         properties=obj,
-                        vector=np.random.rand(1, vector_dimensions)[0].tolist(),
+                        vector=(
+                            2 * np.random.rand(1, vector_dimensions)[0] - 1
+                        ).tolist(),
                     )
                     counter += 1