Merge pull request #1800 from weaviate/uncompressed_quantizer

Add uncompressed quantitizer factory

Author: dirkkul

.github/workflows/main.yaml

@@ -25,7 +25,7 @@ env:
   WEAVIATE_129: 1.29.9
   WEAVIATE_130: 1.30.12
   WEAVIATE_131: 1.31.5
-  WEAVIATE_132: 1.32.0-rc.1-098c594
+  WEAVIATE_132: 1.32.4-cdf9a3b
 
 jobs:
   lint-and-format:

integration/test_collection_config.py

@@ -1636,3 +1636,22 @@ def test_collection_config_create_with_deprecated_syntax(
     assert config.vectorizer_config is None
     assert config.vector_config is None
     assert config.vectorizer == Vectorizers.NONE
+
+
+def test_uncompressed_quantitizer(collection_factory: CollectionFactory) -> None:
+    dummy = collection_factory("dummy")
+    if dummy._connection._weaviate_version.is_lower_than(1, 32, 4):
+        pytest.skip("uncompressed is not supported in Weaviate versions lower than 1.32.4")
+
+    collection = collection_factory(
+        vector_index_config=Configure.VectorIndex.hnsw(
+            vector_cache_max_objects=5,
+            quantizer=Configure.VectorIndex.Quantizer.none(),
+        ),
+    )
+
+    config = collection.config.get()
+    assert config.vector_index_type == VectorIndexType.HNSW
+    assert config.vector_index_config is not None
+    assert isinstance(config.vector_index_config, _VectorIndexConfigHNSW)
+    assert config.vector_index_config.quantizer is None

weaviate/collections/classes/config_vector_index.py

@@ -80,7 +80,10 @@ def vector_index_type() -> VectorIndexType: ...
     def _to_dict(self) -> Dict[str, Any]:
         ret_dict = super()._to_dict()
         if self.quantizer is not None:
-            ret_dict[self.quantizer.quantizer_name()] = self.quantizer._to_dict()
+            if isinstance(self.quantizer, _UncompressedConfigCreate):
+                ret_dict[self.quantizer.quantizer_name()] = True
+            else:
+                ret_dict[self.quantizer.quantizer_name()] = self.quantizer._to_dict()
         if self.distance is not None:
             ret_dict["distance"] = str(self.distance.value)
         if self.multivector is not None and self.multivector.encoding is not None:
@@ -280,6 +283,12 @@ def quantizer_name() -> str:
         return "rq"
 
 
+class _UncompressedConfigCreate(_QuantizerConfigCreate):
+    @staticmethod
+    def quantizer_name() -> str:
+        return "skipDefaultQuantization"
+
+
 class _PQConfigUpdate(_QuantizerConfigUpdate):
     bitCompression: Optional[bool] = Field(default=None)
     centroids: Optional[int]
@@ -449,6 +458,11 @@ def rq(
             rescoreLimit=rescore_limit,
         )
 
+    @staticmethod
+    def none() -> _UncompressedConfigCreate:
+        """Create a a vector index without compression."""
+        return _UncompressedConfigCreate()
+
 
 class _VectorIndex:
     MultiVector = _VectorIndexMultiVector