Update app.py

app.py

@@ -1,6 +1,7 @@
 import gradio as gr
 from gradio_client import Client
 from huggingface_hub import InferenceClient
+import random
 
 ss_client = Client("https://omnibus-html-image-current-tab.hf.space/")
 
@@ -37,7 +38,7 @@ def format_prompt(message, history):
     prompt += message
     return prompt
 
-def chat_inf(prompt,history,memory,client_choice,temp,tokens,top_p,rep_p,chat_mem):
+def chat_inf(system_prompt,prompt,history,memory,client_choice,seed,temp,tokens,top_p,rep_p,chat_mem):
     hist_len=0
     client=clients[int(client_choice)-1]
     if not history:
@@ -49,7 +50,7 @@ def chat_inf(prompt,history,memory,client_choice,temp,tokens,top_p,rep_p,chat_me
     if memory:
         for ea in memory[0-chat_mem:]:
             hist_len+=len(str(ea))
-    in_len=len(prompt)+hist_len
+    in_len=len(system_prompt+prompt)+hist_len
 
     if (in_len+tokens) > 8000:
         history.append((prompt,"Wait, that's too many tokens, please reduce the 'Chat Memory' value, or reduce the 'Max new tokens' value"))
@@ -61,8 +62,9 @@ def chat_inf(prompt,history,memory,client_choice,temp,tokens,top_p,rep_p,chat_me
             top_p=top_p,
             repetition_penalty=rep_p,
             do_sample=True,
+            seed=seed,
         )
-        formatted_prompt = format_prompt(prompt, memory[0-chat_mem:])
+        formatted_prompt = format_prompt(f"{system_prompt}, {prompt}", memory[0-chat_mem:])
         stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=True)
         output = ""
         for response in stream:
@@ -86,7 +88,14 @@ def get_screenshot(chat: list,height=5000,width=600,chatblock=[],theme="light",w
 
 def clear_fn():
     return None,None,None,None
+rand_val=random.randint(1,1111111111111111)
 
+def check_rand(inp,val):
+    if inp==True:
+        return random.randint(1,1111111111111111)
+    else:
+        return int(val)
+    
 with gr.Blocks() as app:
     memory=gr.State()
     chat_b = gr.Chatbot(height=500)
@@ -97,8 +106,10 @@ with gr.Blocks() as app:
                 btn = gr.Button("Chat")
             with gr.Column(scale=1):
                 with gr.Group():
-                    temp=gr.Slider(label="Temperature",step=0.01, minimum=0.01, maximum=1.0, value=0.49)
+                    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=1600,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=0.49)
                     top_p=gr.Slider(label="Top-P",step=0.01, minimum=0.01, maximum=1.0, value=0.49)
                     rep_p=gr.Slider(label="Repetition Penalty",step=0.01, minimum=0.1, maximum=2.0, value=0.99)
                     chat_mem=gr.Number(label="Chat Memory", info="Number of previous chats to retain",value=4)
@@ -107,7 +118,7 @@ with gr.Blocks() as app:
     client_choice.change(load_models,client_choice,[chat_b])
     app.load(load_models,client_choice,[chat_b])
     
-    chat_sub=inp.submit().then(chat_inf,[inp,chat_b,memory,client_choice,temp,tokens,top_p,rep_p,chat_mem],[chat_b,memory])
-    go=btn.click().then(chat_inf,[inp,chat_b,memory,client_choice,temp,tokens,top_p,rep_p,chat_mem],[chat_b,memory])
+    chat_sub=inp.submit(check_rand,[rand,seed],seed).then(chat_inf,[inp,inp,chat_b,memory,client_choice,seed,temp,tokens,top_p,rep_p,chat_mem],[chat_b,memory])
+    go=btn.click(check_rand,[rand,seed],seed).then(chat_inf,[inp,inp,chat_b,memory,client_choice,seed,temp,tokens,top_p,rep_p,chat_mem],[chat_b,memory])
 
     app.queue(default_concurrency_limit=10).launch()