Update app.py

app.py

@@ -47,13 +47,6 @@ def load_data():
 
 embeddings, patent_numbers, metadata, texts = load_data()
 
-# Normalize embeddings for cosine similarity
-embeddings = embeddings / np.linalg.norm(embeddings, axis=1, keepdims=True)
-
-# Create FAISS index for cosine similarity
-index = faiss.IndexFlatIP(embeddings.shape[1])
-index.add(embeddings)
-
 # Load BERT model for encoding search queries
 try:
     bert_model = AutoModel.from_pretrained('anferico/bert-for-patents')
@@ -66,6 +59,20 @@ except Exception as e:
     print("Falling back to a general-purpose model.")
     model = SentenceTransformer('all-MiniLM-L6-v2')
 
+# Check if the embedding dimensions match
+if embeddings.shape[1] != model.get_sentence_embedding_dimension():
+    print("Embedding dimensions do not match. Rebuilding FAISS index.")
+    # Rebuild embeddings using the new model
+    embeddings = np.array([model.encode(text) for text in texts])
+    embeddings = embeddings / np.linalg.norm(embeddings, axis=1, keepdims=True)
+
+# Normalize embeddings for cosine similarity
+embeddings = embeddings / np.linalg.norm(embeddings, axis=1, keepdims=True)
+
+# Create FAISS index for cosine similarity
+index = faiss.IndexFlatIP(embeddings.shape[1])
+index.add(embeddings)
+
 # Create TF-IDF vectorizer
 tfidf_vectorizer = TfidfVectorizer(stop_words='english')
 tfidf_matrix = tfidf_vectorizer.fit_transform(texts)