Update app.py

app.py

@@ -95,6 +95,21 @@ vector_store = Qdrant(
 )
 retriever = vector_store.as_retriever()
 
+pair_ranker = pipeline(
+    "text-classification",
+    model="llm-blender/PairRM",
+    tokenizer="llm-blender/PairRM",
+    return_all_scores=True
+)
+
+gen_fuser = pipeline(
+    "text-generation",
+    model="llm-blender/gen_fuser_3b",
+    tokenizer="llm-blender/gen_fuser_3b",
+    max_length=2048,
+    do_sample=False
+)
+
 
 
 
@@ -137,9 +152,6 @@ elif "All" in selected_model:
     from groq import Groq
     import google.generativeai as genai
     genai.configure(api_key=st.secrets["GEMINI_API_KEY"])
-    pair_ranker = pipeline("text-classification", model="llm-blender/PairRM")
-    gen_fuser = pipeline("text-generation", model="llm-blender/gen_fuser_3b", max_length=2048, do_sample=False)
-
     def get_all_model_responses(prompt):
         openai_resp = ChatOpenAI(model="gpt-4o", temperature=0.2, api_key=st.secrets["OPENAI_API_KEY"]).invoke(
             [{"role": "system", "content": prompt}]).content
@@ -332,26 +344,6 @@ def rerank_with_cohere(query, documents, top_n=5):
     results = co.rerank(query=query, documents=raw_texts, top_n=min(top_n, len(raw_texts)), model="rerank-v3.5")
     return [documents[result.index] for result in results]
 
-
-
-
-
-pair_ranker = pipeline(
-    "text-classification",
-    model="llm-blender/PairRM",
-    tokenizer="llm-blender/PairRM",
-    return_all_scores=True
-)
-
-gen_fuser = pipeline(
-    "text-generation",
-    model="llm-blender/gen_fuser_3b",
-    tokenizer="llm-blender/gen_fuser_3b",
-    max_length=2048,
-    do_sample=False
-)
-
-
 # Final answer generation using reranked context
 def get_reranked_response(query):
     docs = retriever.get_relevant_documents(query)