Update app.py

app.py

@@ -232,7 +232,7 @@ def main(query: str, client: QdrantClient, collection_name: str, llm, dense_mode
         result_metadatas = "\n\n".join(f'{value}' for value in filtered_metadatas)
 
         prompt = answer_with_context(context, query)
-        answer = gen_text(prompt, max_tokens=300, sampling_params=SamplingParams(temperature=0.6, top_p=0.9))
+        answer = gen_text(prompt, max_tokens=300, sampling_params=SamplingParams(temperature=0.6, top_p=0.9, top_k=10))
         answer = f"{answer}\n\n\nSource(s) :\n\n{result_metadatas}"
             
         if not st.session_state.documents_only:
@@ -244,15 +244,15 @@ def main(query: str, client: QdrantClient, collection_name: str, llm, dense_mode
         print(f'Choice 2: {action}')
         if action == 'General Question':
             prompt = open_query_prompt(past_messages, query)
-            answer = gen_text(prompt, max_tokens=300, sampling_params=SamplingParams(temperature=0.6, top_p=0.9))
+            answer = gen_text(prompt, max_tokens=300, sampling_params=SamplingParams(temperature=0.6, top_p=0.9, top_k=10))
         else:
             if st.session_state.documents_only:
                 prompt = idk(query)
-                answer = gen_text(prompt, max_tokens=128, sampling_params=SamplingParams(temperature=0.6, top_p=0.9))
+                answer = gen_text(prompt, max_tokens=128, sampling_params=SamplingParams(temperature=0.6, top_p=0.9, top_k=10))
                 print(f'TYPE: {type(answer)}')
             else:
                 prompt = self_knowledge(query)
-                answer = gen_text(prompt, max_tokens=300, sampling_params=SamplingParams(temperature=0.6, top_p=0.9))
+                answer = gen_text(prompt, max_tokens=300, sampling_params=SamplingParams(temperature=0.6, top_p=0.9, top_k=10))
                 answer = f'Internal Knowledge :\n\n{answer}'
 
     torch.cuda.empty_cache()
@@ -745,7 +745,7 @@ if __name__ == '__main__':
             if st.session_state.chat_id == 'New Conversation':
                 output = llm.chat(
                     build_prompt_conv(),
-                    SamplingParams(temperature=0.6,top_p=0.9, max_tokens=10)
+                    SamplingParams(temperature=0.6,top_p=0.9, max_tokens=10, top_k=10)
                 )
                 print(f'OUTPUT : {output[0].outputs[0].text}')
                 st.session_state.chat_id = output[0].outputs[0].text.replace('"', '')