got back threading

app.py

@@ -6,7 +6,7 @@ import os
 from threading import Thread
 from openai import OpenAI
 import spaces
-import multiprocessing as mp
+# import multiprocessing as mp
 import sys
 
 # Init ZeroGPU
@@ -117,9 +117,9 @@ def llama_generation(input_text: str,
     # thread.start()
 
     # Multiprocessing to avoid pickle errors
-    process = mp.Process(target=llama_model.generate, kwargs=generate_kwargs)
-    process.start()
-    process.join()
+    thread = Thread(target=llama_model.generate, kwargs=generate_kwargs)
+    thread.start()
+    thread.join()
     return streamer
 
 def check_cuda():
@@ -131,10 +131,10 @@ def check_cuda():
 first_time = True
 llm_mode = ""
 
-async def bot_comms(input_text: str,
-                    history: list,
-                    temperature: float,
-                    max_new_tokens: int):
+def bot_comms(input_text: str,
+              history: list,
+              temperature: float,
+              max_new_tokens: int):
     """
     The connection between gradio and the LLM's
     """