Update app.py

app.py

@@ -41,9 +41,12 @@ class Chatbot:
         self.embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2", model_kwargs={"device": "cuda"})
 
         try:
+            # Initialize FAISS with GPU support
             cpu_index = faiss.read_index('faiss_index_new_model3.index')
-            gpu_resource = faiss.StandardGpuResources()
-            self.vectorstore = faiss.index_cpu_to_gpu(gpu_resource, 0, cpu_index)
+            res = faiss.StandardGpuResources()  # Use this to allocate the GPU resources
+            co = faiss.GpuClonerOptions()
+            co.useFloat16 = True  # Enable float16 for better performance
+            self.vectorstore = faiss.index_cpu_to_gpu(res, 0, cpu_index, co)
             print("Loaded embedding successfully")
         except Exception as e:
             print("FAISS could not be imported or index could not be loaded.")
@@ -60,10 +63,10 @@ class Chatbot:
             return False
 
     def format_prompt(self, query):
-        prompt=f"""
+        prompt = f"""
         You are a knowledgeable assistant with access to a comprehensive database. 
         I need you to answer my question and provide related information in a specific format.
-        I have provided four relatable json files , choose the most suitable chunks for answering the query
+        I have provided four relatable json files, choose the most suitable chunks for answering the query.
         Here's what I need:
         Include a final answer without additional comments, sign-offs, or extra phrases. Be direct and to the point.