new code

src/pdfchatbot.py

@@ -53,6 +53,9 @@ class PDFChatBot:
         docs = text_splitter.split_documents(self.documents)
         self.vectordb = Chroma.from_documents(docs, self.embeddings)
         print("Vector store created")
+    @spaces.GPU
+    def load_tokenizer(self):
+        self.tokenizer = AutoModelForCausalLM.from_pretrained("meta-llama/Meta-Llama-3-8B-Instruct")
 
     @spaces.GPU
     def create_organic_pipeline(self):
@@ -62,6 +65,7 @@ class PDFChatBot:
             model_kwargs={"torch_dtype": torch.bfloat16},
             device="cuda",
         )
+        self.load_tokenizer()
         print("Model pipeline loaded")
 
     def get_organic_context(self, query):
@@ -77,7 +81,7 @@ class PDFChatBot:
             {"role": "system", "content": "From the the contained given below, answer the question of user \n " + self.current_context},
             {"role": "user", "content": query},
         ]
-        prompt = self.pipeline.tokenizer.apply_chat_template(
+        prompt = self.tokenizer.apply_chat_template(
             messages,
             tokenize=False,
             add_generation_prompt=True