ebm4subjects 0.5.2__tar.gz → 0.5.3__tar.gz

{ebm4subjects-0.5.2 → ebm4subjects-0.5.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ebm4subjects
-Version: 0.5.2
+Version: 0.5.3
 Summary: Embedding Based Matching for Automated Subject Indexing
 Author: Deutsche Nationalbibliothek
 Maintainer-email: Clemens Rietdorf <c.rietdorf@dnb.de>, Maximilian Kähler <m.kaehler@dnb.de>

{ebm4subjects-0.5.2 → ebm4subjects-0.5.3}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "ebm4subjects"
-version = "0.5.2"
+version = "0.5.3"
 description = "Embedding Based Matching for Automated Subject Indexing"
 authors = [
     {name = "Deutsche Nationalbibliothek"},

{ebm4subjects-0.5.2 → ebm4subjects-0.5.3}/src/ebm4subjects/ebm_model.py

@@ -15,7 +15,7 @@ from ebm4subjects.duckdb_client import Duckdb_client
 from ebm4subjects.ebm_logging import EbmLogger, NullLogger, XGBLogging
 from ebm4subjects.embedding_generator import (
     EmbeddingGeneratorHuggingFaceTEI,
-    EmbeddingGeneratorInternal,
+    EmbeddingGeneratorOfflineInference,
     EmbeddingGeneratorMock,
 )
 
@@ -43,7 +43,7 @@ class EbmModel:
         use_altLabels: bool = True,
         hnsw_index_params: dict | str | None = None,
         embedding_model_name: str | None = None,
-        embedding_model_type: str = "internal",
+        embedding_model_type: str = "offline-inference",
         embedding_model_args: dict | str | None = None,
         encode_args_vocab: dict | str | None = None,
         encode_args_documents: dict | str | None = None,
@@ -179,9 +179,9 @@ class EbmModel:
             None
         """
         if self.generator is None:
-            if self.embedding_model_type == "internal":
-                self.logger.info("initializing internal embedding generator")
-                self.generator = EmbeddingGeneratorInternal(
+            if self.embedding_model_type == "offline-inference":
+                self.logger.info("initializing offline-inference embedding generator")
+                self.generator = EmbeddingGeneratorOfflineInference(
                     model_name=self.embedding_model_name,
                     embedding_dimensions=self.embedding_dimensions,
                     **self.embedding_model_args,
@@ -404,7 +404,7 @@ class EbmModel:
             .join(
                 other=gold_standard.with_columns(pl.lit(True).alias("gold")),
                 on=["doc_id", "label_id"],
-                how="outer",
+                how="full",
             )
             # Fill dataframe so that all not suggested labels which are not part of
             # the gold standard and all gold standard labels which where not

{ebm4subjects-0.5.2 → ebm4subjects-0.5.3}/src/ebm4subjects/embedding_generator.py

@@ -102,9 +102,10 @@ class EmbeddingGeneratorHuggingFaceTEI(EmbeddingGeneratorAPI):
         return np.array(embeddings)
 
 
-class EmbeddingGeneratorInternal(EmbeddingGenerator):
+class EmbeddingGeneratorOfflineInference(EmbeddingGenerator):
     """
-    A class for generating embeddings using a given SentenceTransformer model.
+    A class for generating embeddings using a given SentenceTransformer model
+    loaded offline with SentenceTransformer.
 
     Args:
         model_name (str): The name of the SentenceTransformer model.
@@ -118,7 +119,7 @@ class EmbeddingGeneratorInternal(EmbeddingGenerator):
 
     def __init__(self, model_name: str, embedding_dimensions: int, **kwargs) -> None:
         """
-        Initializes the internal EmbeddingGenerator.
+        Initializes the EmbeddingGenerator in offline inference mode.
 
         Sets the model name, embedding dimensions, and creates a
         SentenceTransformer model instance.