Update app.py

app.py

@@ -21,7 +21,6 @@ temperature=0.9
 top_p=0.95
 repetition_penalty=1.0
 
-
 def compress_history(history,temperature=temperature,top_p=top_p,repetition_penalty=repetition_penalty):
     formatted_prompt=f"{COMPRESS_HISTORY.format(history=history)}"
     generate_kwargs = dict(
@@ -64,8 +63,8 @@ def generate(prompt, history,max_new_tokens,temperature=temperature,top_p=top_p,
             cnt+=len(l.split("\n"))
     print(f'cnt:: {cnt}')
     if cnt > MAX_HISTORY:
-        history = compress_history(str(history), temperature, max_new_tokens, top_p, repetition_penalty)
-    formatted_prompt = format_prompt(f"{GAME_MASTER.format(history=history)}, {prompt}", history)
+        history1 = compress_history(str(history), temperature, max_new_tokens, top_p, repetition_penalty)
+    formatted_prompt = format_prompt(f"{GAME_MASTER.format(history=history1)}, {prompt}", history)
     stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
     output = ""
 
@@ -92,9 +91,9 @@ def generate(prompt, history,max_new_tokens,temperature=temperature,top_p=top_p,
     
     if history:
         history.append((prompt,output))
-        return "",history
+        yield "",history
     else:
-        return "",[(prompt,output)]
+        yield "",[(prompt,output)]
 
 def clear_fn():
     return None,None