vector_dot_product

caretdev · Apr 16, 2024 · e5c9a64 · e5c9a64
1 parent d766737
commit e5c9a64
Showing 1 changed file with 11 additions and 21 deletions.
diff --git a/langchain_iris/vectorstores.py b/langchain_iris/vectorstores.py
@@ -5,9 +5,7 @@
 import enum
 import logging
 import uuid
-from functools import partial
 from typing import (
-    TYPE_CHECKING,
     Any,
     Callable,
     Dict,
@@ -20,24 +18,19 @@
     Type,
 )
 
-import numpy as np
 from sqlalchemy import (
     Connection,
     and_,
     asc,
-    literal,
     VARCHAR,
-    UUID,
     TEXT,
     Column,
-    String,
     Table,
     create_engine,
     insert,
     text,
     delete,
     Row,
-    func,
 )
 from sqlalchemy_iris import IRISListBuild
 from sqlalchemy_iris import IRISVector as IRISVectorType
@@ -53,9 +46,6 @@
 from langchain_core.embeddings import Embeddings
 from langchain_core.vectorstores import VectorStore
 
-from langchain.utils import get_from_dict_or_env
-from langchain.vectorstores.utils import maximal_marginal_relevance
-
 Base = declarative_base()  # type: Any
 
 
@@ -64,7 +54,7 @@ class DistanceStrategy(str, enum.Enum):
 
     EUCLIDEAN = "l2"
     COSINE = "cosine"
-    MAX_INNER_PRODUCT = "inner"
+    DOT_PRODUCT = "dot"
 
 
 DEFAULT_DISTANCE_STRATEGY = DistanceStrategy.COSINE
@@ -77,7 +67,6 @@ class DistanceStrategy(str, enum.Enum):
 class IRISVector(VectorStore):
     _conn = None
     native_vector = False
-    native_vector_cosine_similarity = False
 
     def __init__(
         self,
@@ -201,8 +190,8 @@ def distance_strategy(self) -> str:
         if self.native_vector:
             if self._distance_strategy == DistanceStrategy.COSINE:
                 return self.table.c.embedding.cosine
-            elif self._distance_strategy == DistanceStrategy.MAX_INNER_PRODUCT:
-                return self.table.c.embedding.max_inner_product
+            elif self._distance_strategy == DistanceStrategy.DOT_PRODUCT:
+                return self.table.c.embedding.DOT_product
             # elif self._distance_strategy == DistanceStrategy.EUCLIDEAN:
             #     return "langchain_l2_distance"
             else:
@@ -215,7 +204,7 @@ def distance_strategy(self) -> str:
             return "langchain_l2_distance"
         elif self._distance_strategy == DistanceStrategy.COSINE:
             return "langchain_cosine_distance"
-        elif self._distance_strategy == DistanceStrategy.MAX_INNER_PRODUCT:
+        elif self._distance_strategy == DistanceStrategy.DOT_PRODUCT:
             return "langchain_inner_distance"
         else:
             raise ValueError(
@@ -229,9 +218,6 @@ def connect(self) -> Connection:
         try:
             if conn.dialect.supports_vectors:
                 self.native_vector = True
-                self.native_vector_cosine_similarity = (
-                    conn.dialect.vector_cosine_similarity
-                )
         except:  # noqa
             pass
         return conn
@@ -307,8 +293,8 @@ def _select_relevance_score_fn(self) -> Callable[[float], float]:
             return self._cosine_relevance_score_fn
         elif self._distance_strategy == DistanceStrategy.EUCLIDEAN:
             return self._euclidean_relevance_score_fn
-        elif self._distance_strategy == DistanceStrategy.MAX_INNER_PRODUCT:
-            return self._max_inner_product_relevance_score_fn
+        elif self._distance_strategy == DistanceStrategy.DOT_PRODUCT:
+            return self._DOT_product_relevance_score_fn
         else:
             raise ValueError(
                 "No supported normalization function"
@@ -550,7 +536,11 @@ def similarity_search_with_score_by_vector(
                     page_content=result.document,
                     metadata=json.loads(result.metadata),
                 ),
-                round(float(result.distance), 15) if self.embedding_function is not None else None,
+                (
+                    round(float(result.distance), 15)
+                    if self.embedding_function is not None
+                    else None
+                ),
             )
             for result in results
         ]