Unstructured index distance metric fix

src/marqo/core/models/marqo_index.py

@@ -56,7 +56,7 @@ class DistanceMetric(Enum):
     Euclidean = 'euclidean'
     Angular = 'angular'
     DotProduct = 'dotproduct'
-    PrenormalizedAnguar = 'prenormalized-angular'
+    PrenormalizedAngular = 'prenormalized-angular'
     Geodegrees = 'geodegrees'
     Hamming = 'hamming'
 

src/marqo/core/structured_vespa_index/common.py

@@ -17,19 +17,3 @@
 
 SUMMARY_ALL_NON_VECTOR = 'all-non-vector-summary'
 SUMMARY_ALL_VECTOR = 'all-vector-summary'
-
-_DISTANCE_METRIC_MAP = {
-    DistanceMetric.Euclidean: 'euclidean',
-    DistanceMetric.Angular: 'angular',
-    DistanceMetric.DotProduct: 'dotproduct',
-    DistanceMetric.PrenormalizedAnguar: 'prenormalized-angular',
-    DistanceMetric.Geodegrees: 'geodegrees',
-    DistanceMetric.Hamming: 'hamming'
-}
-
-
-def get_distance_metric(marqo_distance_metric: DistanceMetric) -> str:
-    try:
-        return _DISTANCE_METRIC_MAP[marqo_distance_metric]
-    except KeyError:
-        raise ValueError(f'Unknown Marqo distance metric: {marqo_distance_metric}')

src/marqo/core/structured_vespa_index/structured_vespa_schema.py

@@ -129,7 +129,7 @@ def _generate_document_section(self, schema_name: str) -> (List[str], Structured
             document.append(f'field {embedding_field_name} type tensor<float>(p{{}}, x[{model_dim}]) {{')
             document.append('indexing: attribute | index | summary')
             document.append(
-                f'attribute {{ distance-metric: {common.get_distance_metric(self._index_request.distance_metric)} }}')
+                f'attribute {{ distance-metric: {self._get_distance_metric(self._index_request.distance_metric)} }}')
             document.append('index { hnsw {')
             document.append(f'max-links-per-node: {self._index_request.hnsw_config.m}')
             document.append(f'neighbors-to-explore-at-insert: {self._index_request.hnsw_config.ef_construction}')

src/marqo/core/unstructured_vespa_index/unstructured_vespa_schema.py

@@ -152,7 +152,7 @@ def _generate_unstructured_schema(cls, marqo_index: UnstructuredMarqoIndex) -> s
                     field {cls._EMBEDDINGS} type tensor<float>(p{{}}, x[{dimension}]) {{
                         indexing: attribute | index | summary
                         attribute {{
-                            distance-metric: prenormalized-angular
+                            distance-metric: {cls._get_distance_metric(cls, marqo_index.distance_metric)}
                         }}
                         index {{
                             hnsw {{

src/marqo/core/vespa_schema.py

@@ -5,6 +5,8 @@
 from marqo.core.models.marqo_index_request import MarqoIndexRequest, StructuredMarqoIndexRequest, \
     UnstructuredMarqoIndexRequest
 
+from marqo.core.models.marqo_index import *
+
 
 class VespaSchema(ABC):
     """
@@ -16,6 +18,21 @@ class VespaSchema(ABC):
         '-': '_01',
     }
 
+    _DISTANCE_METRIC_MAP = {
+        DistanceMetric.Euclidean: 'euclidean',
+        DistanceMetric.Angular: 'angular',
+        DistanceMetric.DotProduct: 'dotproduct',
+        DistanceMetric.PrenormalizedAngular: 'prenormalized-angular',
+        DistanceMetric.Geodegrees: 'geodegrees',
+        DistanceMetric.Hamming: 'hamming'
+    }
+
+    def _get_distance_metric(self, marqo_distance_metric: DistanceMetric) -> str:
+        try:
+            return self._DISTANCE_METRIC_MAP[marqo_distance_metric]
+        except KeyError:
+            raise ValueError(f'Unknown Marqo distance metric: {marqo_distance_metric}')
+
     @abstractmethod
     def generate_schema(self) -> (str, MarqoIndex):
         """

src/marqo/tensor_search/models/index_settings.py

@@ -35,7 +35,7 @@ class IndexSettings(StrictBaseModel):
     )
     vectorNumericType: core.VectorNumericType = core.VectorNumericType.Float
     annParameters: AnnParameters = AnnParameters(
-        spaceType=core.DistanceMetric.PrenormalizedAnguar,
+        spaceType=core.DistanceMetric.PrenormalizedAngular,
         parameters=core.HnswConfig(
             efConstruction=512,
             m=16

tests/core/index_management/test_index_management.py

@@ -113,7 +113,7 @@ def test_create_index_settingsSchemaDoesNotExist_successful(self):
         marqo_index_request = self.structured_marqo_index_request(
             name=index_name,
             model=Model(name='ViT-B/32'),
-            distance_metric=DistanceMetric.PrenormalizedAnguar,
+            distance_metric=DistanceMetric.PrenormalizedAngular,
             vector_numeric_type=VectorNumericType.Float,
             hnsw_config=HnswConfig(ef_construction=100, m=16),
             fields=[
@@ -156,7 +156,7 @@ def test_create_index_settingsSchemaExists_successful(self):
         marqo_index_request = self.structured_marqo_index_request(
             name=index_name_1,
             model=Model(name='ViT-B/32'),
-            distance_metric=DistanceMetric.PrenormalizedAnguar,
+            distance_metric=DistanceMetric.PrenormalizedAngular,
             vector_numeric_type=VectorNumericType.Float,
             hnsw_config=HnswConfig(ef_construction=100, m=16),
             fields=[
@@ -203,7 +203,7 @@ def test_create_index_indexExists_fails(self):
         marqo_index_request = self.structured_marqo_index_request(
             name=index_name,
             model=Model(name='ViT-B/32'),
-            distance_metric=DistanceMetric.PrenormalizedAnguar,
+            distance_metric=DistanceMetric.PrenormalizedAngular,
             vector_numeric_type=VectorNumericType.Float,
             hnsw_config=HnswConfig(ef_construction=100, m=16),
             fields=[

tests/core/structured_vespa_index/test_schemas/structured_distance_metric_angular.sd

@@ -0,0 +1,65 @@
+schema marqo__test_00structured_00schema_00distance_00metric {
+document {
+field marqo__id type string {
+indexing: attribute | summary
+attribute: fast-search
+rank: filter
+}
+field title type string {
+indexing: summary
+}
+field description type string {
+indexing: summary
+}
+field marqo__chunks_title type array<string> {
+indexing: attribute | summary
+}
+field marqo__embeddings_title type tensor<float>(p{}, x[512]) {
+indexing: attribute | index | summary
+attribute { distance-metric: angular }
+index { hnsw {
+max-links-per-node: 16
+neighbors-to-explore-at-insert: 100
+}}
+}
+field marqo__chunks_description type array<string> {
+indexing: attribute | summary
+}
+field marqo__embeddings_description type tensor<float>(p{}, x[512]) {
+indexing: attribute | index | summary
+attribute { distance-metric: angular }
+index { hnsw {
+max-links-per-node: 16
+neighbors-to-explore-at-insert: 100
+}}
+}
+field marqo__vector_count type int { indexing: attribute | summary }
+}
+rank-profile embedding_similarity inherits default {
+inputs {
+query(marqo__query_embedding) tensor<float>(x[512])
+query(title): 0
+query(description): 0
+}
+first-phase {
+expression: max(if(query(title) > 0, closeness(field, marqo__embeddings_title), 0), if(query(description) > 0, closeness(field, marqo__embeddings_description), 0))
+}
+match-features: closest(marqo__embeddings_title) closest(marqo__embeddings_description) distance(field, marqo__embeddings_title) distance(field, marqo__embeddings_description)
+}
+document-summary all-non-vector-summary {
+summary marqo__id type string { }
+summary title type string { source: title }
+summary description type string { source: description }
+summary marqo__chunks_title type array<string> { }
+summary marqo__chunks_description type array<string> { }
+}
+document-summary all-vector-summary {
+summary marqo__id type string { }
+summary title type string { source: title }
+summary description type string { source: description }
+summary marqo__chunks_title type array<string> { }
+summary marqo__chunks_description type array<string> { }
+summary marqo__embeddings_title type tensor<float>(p{}, x[512]) { }
+summary marqo__embeddings_description type tensor<float>(p{}, x[512]) { }
+}
+}

tests/core/structured_vespa_index/test_schemas/structured_distance_metric_dotproduct.sd

@@ -0,0 +1,65 @@
+schema marqo__test_00structured_00schema_00distance_00metric {
+document {
+field marqo__id type string {
+indexing: attribute | summary
+attribute: fast-search
+rank: filter
+}
+field title type string {
+indexing: summary
+}
+field description type string {
+indexing: summary
+}
+field marqo__chunks_title type array<string> {
+indexing: attribute | summary
+}
+field marqo__embeddings_title type tensor<float>(p{}, x[512]) {
+indexing: attribute | index | summary
+attribute { distance-metric: dotproduct }
+index { hnsw {
+max-links-per-node: 16
+neighbors-to-explore-at-insert: 100
+}}
+}
+field marqo__chunks_description type array<string> {
+indexing: attribute | summary
+}
+field marqo__embeddings_description type tensor<float>(p{}, x[512]) {
+indexing: attribute | index | summary
+attribute { distance-metric: dotproduct }
+index { hnsw {
+max-links-per-node: 16
+neighbors-to-explore-at-insert: 100
+}}
+}
+field marqo__vector_count type int { indexing: attribute | summary }
+}
+rank-profile embedding_similarity inherits default {
+inputs {
+query(marqo__query_embedding) tensor<float>(x[512])
+query(title): 0
+query(description): 0
+}
+first-phase {
+expression: max(if(query(title) > 0, closeness(field, marqo__embeddings_title), 0), if(query(description) > 0, closeness(field, marqo__embeddings_description), 0))
+}
+match-features: closest(marqo__embeddings_title) closest(marqo__embeddings_description) distance(field, marqo__embeddings_title) distance(field, marqo__embeddings_description)
+}
+document-summary all-non-vector-summary {
+summary marqo__id type string { }
+summary title type string { source: title }
+summary description type string { source: description }
+summary marqo__chunks_title type array<string> { }
+summary marqo__chunks_description type array<string> { }
+}
+document-summary all-vector-summary {
+summary marqo__id type string { }
+summary title type string { source: title }
+summary description type string { source: description }
+summary marqo__chunks_title type array<string> { }
+summary marqo__chunks_description type array<string> { }
+summary marqo__embeddings_title type tensor<float>(p{}, x[512]) { }
+summary marqo__embeddings_description type tensor<float>(p{}, x[512]) { }
+}
+}

tests/core/structured_vespa_index/test_schemas/structured_distance_metric_euclidean.sd

@@ -0,0 +1,65 @@
+schema marqo__test_00structured_00schema_00distance_00metric {
+document {
+field marqo__id type string {
+indexing: attribute | summary
+attribute: fast-search
+rank: filter
+}
+field title type string {
+indexing: summary
+}
+field description type string {
+indexing: summary
+}
+field marqo__chunks_title type array<string> {
+indexing: attribute | summary
+}
+field marqo__embeddings_title type tensor<float>(p{}, x[512]) {
+indexing: attribute | index | summary
+attribute { distance-metric: euclidean }
+index { hnsw {
+max-links-per-node: 16
+neighbors-to-explore-at-insert: 100
+}}
+}
+field marqo__chunks_description type array<string> {
+indexing: attribute | summary
+}
+field marqo__embeddings_description type tensor<float>(p{}, x[512]) {
+indexing: attribute | index | summary
+attribute { distance-metric: euclidean }
+index { hnsw {
+max-links-per-node: 16
+neighbors-to-explore-at-insert: 100
+}}
+}
+field marqo__vector_count type int { indexing: attribute | summary }
+}
+rank-profile embedding_similarity inherits default {
+inputs {
+query(marqo__query_embedding) tensor<float>(x[512])
+query(title): 0
+query(description): 0
+}
+first-phase {
+expression: max(if(query(title) > 0, closeness(field, marqo__embeddings_title), 0), if(query(description) > 0, closeness(field, marqo__embeddings_description), 0))
+}
+match-features: closest(marqo__embeddings_title) closest(marqo__embeddings_description) distance(field, marqo__embeddings_title) distance(field, marqo__embeddings_description)
+}
+document-summary all-non-vector-summary {
+summary marqo__id type string { }
+summary title type string { source: title }
+summary description type string { source: description }
+summary marqo__chunks_title type array<string> { }
+summary marqo__chunks_description type array<string> { }
+}
+document-summary all-vector-summary {
+summary marqo__id type string { }
+summary title type string { source: title }
+summary description type string { source: description }
+summary marqo__chunks_title type array<string> { }
+summary marqo__chunks_description type array<string> { }
+summary marqo__embeddings_title type tensor<float>(p{}, x[512]) { }
+summary marqo__embeddings_description type tensor<float>(p{}, x[512]) { }
+}
+}

tests/core/structured_vespa_index/test_schemas/structured_distance_metric_geodegrees.sd

@@ -0,0 +1,65 @@
+schema marqo__test_00structured_00schema_00distance_00metric {
+document {
+field marqo__id type string {
+indexing: attribute | summary
+attribute: fast-search
+rank: filter
+}
+field title type string {
+indexing: summary
+}
+field description type string {
+indexing: summary
+}
+field marqo__chunks_title type array<string> {
+indexing: attribute | summary
+}
+field marqo__embeddings_title type tensor<float>(p{}, x[512]) {
+indexing: attribute | index | summary
+attribute { distance-metric: geodegrees }
+index { hnsw {
+max-links-per-node: 16
+neighbors-to-explore-at-insert: 100
+}}
+}
+field marqo__chunks_description type array<string> {
+indexing: attribute | summary
+}
+field marqo__embeddings_description type tensor<float>(p{}, x[512]) {
+indexing: attribute | index | summary
+attribute { distance-metric: geodegrees }
+index { hnsw {
+max-links-per-node: 16
+neighbors-to-explore-at-insert: 100
+}}
+}
+field marqo__vector_count type int { indexing: attribute | summary }
+}
+rank-profile embedding_similarity inherits default {
+inputs {
+query(marqo__query_embedding) tensor<float>(x[512])
+query(title): 0
+query(description): 0
+}
+first-phase {
+expression: max(if(query(title) > 0, closeness(field, marqo__embeddings_title), 0), if(query(description) > 0, closeness(field, marqo__embeddings_description), 0))
+}
+match-features: closest(marqo__embeddings_title) closest(marqo__embeddings_description) distance(field, marqo__embeddings_title) distance(field, marqo__embeddings_description)
+}
+document-summary all-non-vector-summary {
+summary marqo__id type string { }
+summary title type string { source: title }
+summary description type string { source: description }
+summary marqo__chunks_title type array<string> { }
+summary marqo__chunks_description type array<string> { }
+}
+document-summary all-vector-summary {
+summary marqo__id type string { }
+summary title type string { source: title }
+summary description type string { source: description }
+summary marqo__chunks_title type array<string> { }
+summary marqo__chunks_description type array<string> { }
+summary marqo__embeddings_title type tensor<float>(p{}, x[512]) { }
+summary marqo__embeddings_description type tensor<float>(p{}, x[512]) { }
+}
+}