Reuse transformer code from Pystow (#485)

Author: cthoyt

pyproject.toml

@@ -68,7 +68,7 @@ dependencies = [
     "humanize",
     "tabulate",
     "cachier",
-    "pystow<=0.7.23",
+    "pystow>=0.7.28",
     "bioversions>=0.8.243",
     "bioregistry>=0.12.30",
     "ssslm>=0.0.13",

src/pyobo/api/embedding.py

@@ -3,13 +3,15 @@
 from __future__ import annotations
 
 import tempfile
+import warnings
 from pathlib import Path
 from typing import TYPE_CHECKING, Literal
 
 import bioregistry
 import curies
 import numpy as np
 import pandas as pd
+from pystow import get_sentence_transformer
 from tqdm import tqdm
 from typing_extensions import Unpack
 
@@ -34,10 +36,12 @@
 
 def get_text_embedding_model() -> sentence_transformers.SentenceTransformer:
     """Get the default text embedding model."""
-    from sentence_transformers import SentenceTransformer
-
-    model = SentenceTransformer("all-MiniLM-L6-v2")
-    return model
+    warnings.warn(
+        "get_text_embedding_model() is deprecated, use pystow.get_sentence_transfomer() directly",
+        DeprecationWarning,
+        stacklevel=2,
+    )
+    return get_sentence_transformer()
 
 
 def _get_text(
@@ -157,7 +161,7 @@ def get_text_embeddings_df(
         luids.append(identifier)
         texts.append(text)
     if model is None:
-        model = get_text_embedding_model()
+        model = get_sentence_transformer()
     res = model.encode(texts, show_progress_bar=True)
     df = pd.DataFrame(res, index=luids)
     df.to_csv(path, sep="\t")  # index is important here!
@@ -199,7 +203,7 @@ def get_text_embedding(
     if text is None:
         return None
     if model is None:
-        model = get_text_embedding_model()
+        model = get_sentence_transformer()
     res = model.encode([text])
     return res[0]
 
@@ -239,7 +243,7 @@ def get_text_embedding_similarity(
         # 0.24702128767967224
     """
     if model is None:
-        model = get_text_embedding_model()
+        model = get_sentence_transformer()
     e1 = get_text_embedding(reference_1, model=model)
     e2 = get_text_embedding(reference_2, model=model)
     if e1 is None or e2 is None: