Update app.py

app.py

@@ -64,7 +64,7 @@ def chat_inf(system_prompt,prompt,history,client_choice,seed,temp,tokens,top_p,r
         hist_len=len(history)
         print(hist_len)
     in_len=len(system_prompt+prompt)+hist_len
-    print(in_len)
+    print("\n#########"+in_len)
     if (in_len+tokens) > 8000:
         yield [(prompt,"Wait. I need to compress our Chat history...")]
         history=compress_history(history,client_choice,seed,temp,tokens,top_p,rep_p)
@@ -137,7 +137,7 @@ with gr.Blocks() as app:
                 with gr.Group():
                     rand = gr.Checkbox(label="Random Seed", value=True)
                     seed=gr.Slider(label="Seed", minimum=1, maximum=1111111111111111,step=1, value=rand_val)
-                    tokens = gr.Slider(label="Max new tokens",value=6400,minimum=0,maximum=8000,step=64,interactive=True, visible=True,info="The maximum number of tokens")
+                    tokens = gr.Slider(label="Max new tokens",value=3840,minimum=0,maximum=8000,step=64,interactive=True, visible=True,info="The maximum number of tokens")
                     temp=gr.Slider(label="Temperature",step=0.01, minimum=0.01, maximum=1.0, value=0.9)
                     top_p=gr.Slider(label="Top-P",step=0.01, minimum=0.01, maximum=1.0, value=0.9)
                     rep_p=gr.Slider(label="Repetition Penalty",step=0.1, minimum=0.1, maximum=2.0, value=1.0)