Update app.py

app.py

@@ -8,37 +8,33 @@ models=[
     "google/gemma-7b",
     "google/gemma-7b-it",
     "google/gemma-2b",
-    "google/gemma-2b-it",
+    "google/gemma-2b-it"
+]
+clients=[
+InferenceClient(models[0]),
+InferenceClient(models[1]),
+InferenceClient(models[2]),
+InferenceClient(models[3]),
 ]
-client_z=[]
 
 VERBOSE=False
 
-def load_models(inp):
-    if VERBOSE==True:    
-        print(type(inp))
-        print(inp)
-        print(models[inp])
-    client_z.clear()
-    client_z.append(InferenceClient(models[inp]))
-    return gr.update(label=models[inp])
-
-def format_prompt(message, history, cust_p):
+def format_prompt(message, history):
     prompt = ""
     if history:
+        #<start_of_turn>userHow does the brain work?<end_of_turn><start_of_turn>model
         for user_prompt, bot_response in history:
-            prompt += f"<start_of_turn>user{user_prompt}<end_of_turn>"
-            prompt += f"<start_of_turn>model{bot_response}<end_of_turn>"
-            if VERBOSE==True:
-                print(prompt)
-    #prompt += f"<start_of_turn>user\n{message}<end_of_turn>\n<start_of_turn>model\n"
-    prompt+=cust_p.replace("USER_INPUT",message)
+            prompt += f"{user_prompt}\n"
+            #print(prompt)
+            prompt += f"{bot_response}\n"
+            #print(prompt)
+    prompt += f"<start_of_turn>user{message}<end_of_turn><start_of_turn>model"
     return prompt
 
-def chat_inf(system_prompt,prompt,history,memory,client_choice,seed,temp,tokens,top_p,rep_p,chat_mem,cust_p):
+def chat_inf(system_prompt,prompt,history,memory,client_choice,seed,temp,tokens,top_p,rep_p,chat_mem):
     #token max=8192
     hist_len=0
-    client=client_z[0]
+    client=clients[int(client_choice)-1]
     if not history:
         history = []
         hist_len=0
@@ -62,10 +58,8 @@ def chat_inf(system_prompt,prompt,history,memory,client_choice,seed,temp,tokens,
             do_sample=True,
             seed=seed,
         )
-        if system_prompt:
-            formatted_prompt = format_prompt(f"{system_prompt}, {prompt}", memory[0-chat_mem:],cust_p)
-        else:
-            formatted_prompt = format_prompt(prompt, memory[0-chat_mem:],cust_p)
+        formatted_prompt = format_prompt(f"{system_prompt}, {prompt}", memory[0-chat_mem:])
+        #print("\n######### PROMPT "+str(len(formatted_prompt)))
         stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
         output = ""
         for response in stream:
@@ -74,7 +68,6 @@ def chat_inf(system_prompt,prompt,history,memory,client_choice,seed,temp,tokens,
         history.append((prompt,output))
         memory.append((prompt,output))
         yield history,memory
-        
     if VERBOSE==True:
         print("\n######### HIST "+str(in_len))
         print("\n######### TOKENS "+str(tokens))        
@@ -116,8 +109,6 @@ with gr.Blocks() as app:
                             stop_btn=gr.Button("Stop")
                             clear_btn=gr.Button("Clear")                
                 client_choice=gr.Dropdown(label="Models",type='index',choices=[c for c in models],value=models[0],interactive=True)
-                with gr.Accordion("Prompt Format",open=False):
-                    custom_prompt=gr.Textbox(label="Modify Prompt Format", info="For testing purposes. 'USER_INPUT' is where 'SYSTEM_PROMPT, PROMPT' will be placed", lines=5,value="<start_of_turn>userUSER_INPUT<end_of_turn><start_of_turn>model")
             with gr.Column(scale=1):
                 with gr.Group():
                     rand = gr.Checkbox(label="Random Seed", value=True)
@@ -140,15 +131,9 @@ with gr.Blocks() as app:
                     theme=gr.Radio(label="Theme", choices=["light","dark"],value="light")
                     chatblock=gr.Dropdown(label="Chatblocks",info="Choose specific blocks of chat",choices=[c for c in range(1,40)],multiselect=True)
 
-    
-    client_choice.change(load_models,client_choice,[chat_b])
-    app.load(load_models,client_choice,[chat_b])
-    
     im_go=im_btn.click(get_screenshot,[chat_b,im_height,im_width,chatblock,theme,wait_time],img)
-    
-    chat_sub=inp.submit(check_rand,[rand,seed],seed).then(chat_inf,[sys_inp,inp,chat_b,memory,client_choice,seed,temp,tokens,top_p,rep_p,chat_mem,custom_prompt],[chat_b,memory])
-    go=btn.click(check_rand,[rand,seed],seed).then(chat_inf,[sys_inp,inp,chat_b,memory,client_choice,seed,temp,tokens,top_p,rep_p,chat_mem,custom_prompt],[chat_b,memory])
-    
+    chat_sub=inp.submit(check_rand,[rand,seed],seed).then(chat_inf,[sys_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,[sys_inp,inp,chat_b,memory,client_choice,seed,temp,tokens,top_p,rep_p,chat_mem],[chat_b,memory])
     stop_btn.click(None,None,None,cancels=[go,im_go,chat_sub])
     clear_btn.click(clear_fn,None,[inp,sys_inp,chat_b,memory])
 app.queue(default_concurrency_limit=10).launch()