Update app.py

app.py

@@ -3,40 +3,34 @@ from gradio_client import Client
 from huggingface_hub import InferenceClient
 import random
 
-# Configuración del cliente
 ss_client = Client("https://xilixmeaty40-testing.hf.space/")
 
-# Carga de modelos desde un archivo
 with open("models.txt", "r") as file:
     models = file.read().splitlines()
 
-# Creación de clientes de inferencia para cada modelo
-clients = [InferenceClient(model) for model in models]
+clients = []
+for model in models:
+    try:
+        client = InferenceClient(model)
+        clients.append(client)
+    except Exception as e:
+        print(f"Failed to load model {model}: {e}")
 
-# Variable para controlar la verbosidad del programa
-VERBOSE = False
+if not clients:
+    raise Exception("All models failed to load or no models provided.")
 
-# Función para cargar modelos
 def load_models(inp):
-    if VERBOSE:
-        print(type(inp))
-        print(inp)
-        print(models[inp])
     return gr.update(label=models[inp])
 
-# Función para dar formato al prompt de entrada
 def format_prompt(message, history, cust_p):
     prompt = ""
     if history:
         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:
-                print(prompt)
     prompt += cust_p.replace("USER_INPUT", message)
     return prompt
 
-# Función para la interacción de chat
 def chat_inf(system_prompt, prompt, history, memory, client_choice, seed, temp, tokens, top_p, rep_p, chat_mem, cust_p):
     hist_len = 0
     client = clients[int(client_choice) - 1]
@@ -65,17 +59,20 @@ def chat_inf(system_prompt, prompt, history, memory, client_choice, seed, temp,
             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)
-        stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=True)
-        output = ""
-        for response in stream:
-            output += response.token.text
-            yield [(prompt, output)], memory
-        history.append((prompt, output))
-        memory.append((prompt, output))
-        yield history, memory
-
-# Función para obtener capturas de pantalla
-def get_screenshot(chat: list, height=5000, width=600, chatblock=[], theme="light", wait=3000, header=True):
+        try:
+            stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=True)
+            output = ""
+            for response in stream:
+                output += response.token.text
+                yield [(prompt, output)], memory
+            history.append((prompt, output))
+            memory.append((prompt, output))
+            yield history, memory
+        except Exception as e:
+            print(f"Error during model inference: {e}")
+            yield [("Error", "The model failed to respond, possibly due to a timeout. Please try again.")], memory
+
+def get_screenshot(chat, height=5000, width=600, chatblock=[], theme="light", wait=3000, header=True):
     tog = 0
     if chatblock:
         tog = 3
@@ -83,21 +80,17 @@ def get_screenshot(chat: list, height=5000, width=600, chatblock=[], theme="ligh
     out = f'https://xilixmeaty40-testing.hf.space/file={result[tog]}'
     return out
 
-# Función para limpiar los campos
 def clear_fn():
     return None, None, None, None
 
-# Generación de un valor aleatorio para la semilla
 rand_val = random.randint(1, 1111111111111111)
 
-# Función para verificar y establecer la semilla aleatoria
 def check_rand(inp, val):
     if inp:
         return gr.Slider(label="Seed", minimum=1, maximum=1111111111111111, value=random.randint(1, 1111111111111111))
     else:
         return gr.Slider(label="Seed", minimum=1, maximum=1111111111111111, value=int(val))
 
-# Creación de la aplicación con Gradio
 with gr.Blocks() as app:
     memory = gr.State()
     gr.HTML("""<center><h1 style='font-size:xx-large;'>Google Gemma Models</h1><br><h3>running on Huggingface Inference Client</h3><br><h7>EXPERIMENTAL""")
@@ -107,49 +100,35 @@ with gr.Blocks() as app:
             with gr.Column(scale=3):
                 inp = gr.Textbox(label="Prompt")
                 sys_inp = gr.Textbox(label="System Prompt (optional)")
-                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=3, value="<start_of_turn>userUSER_INPUT<end_of_turn><start_of_turn>model")
+                custom_prompt = gr.Textbox(label="Modify Prompt Format", lines=3, value="<start_of_turn>userUSER_INPUT<end_of_turn><start_of_turn>model")
                 with gr.Row():
                     with gr.Column(scale=2):
                         btn = gr.Button("Chat")
                     with gr.Column(scale=1):
-                        with gr.Group():
-                            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)
+                        stop_btn = gr.Button("Stop")
+                        clear_btn = gr.Button("Clear")
+                client_choice = gr.Dropdown(label="Models", type='index', choices=[c for c in models if c in [client.model_id for client in clients]], value=models[0], interactive=True)
             with gr.Column(scale=1):
-                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=300000, minimum=0, maximum=800000, 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)
+                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=300000, minimum=0, maximum=800000, step=64)
+                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", value=4)
         with gr.Accordion(label="Screenshot", open=False):
-            with gr.Row():
-                with gr.Column(scale=3):
-                    im_btn = gr.Button("Screenshot")
-                    img = gr.Image(type='filepath')
-                with gr.Column(scale=1):
-                    with gr.Row():
-                        im_height = gr.Number(label="Height", value=5000)
-                        im_width = gr.Number(label="Width", value=500)
-                    wait_time = gr.Number(label="Wait Time", value=3000)
-                    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)
-
-    # Eventos de la aplicación
-    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])
-
-    stop_btn.click(None, None, None, cancels=[go, im_go, chat_sub])
-    clear_btn.click(clear_fn, None, [inp, sys_inp, chat_b, memory])
-
-# Lanzamiento de la aplicación
-app.queue(default_concurrency_limit=10).launch()
+            im_btn = gr.Button("Screenshot")
+            img = gr.Image(type='filepath')
+            im_height = gr.Number(label="Height", value=5000)
+            im_width = gr.Number(label="Width", value=500)
+            wait_time = gr.Number(label="Wait Time", value=3000)
+            theme = gr.Radio(label="Theme", choices=["light", "dark"], value="light")
+            chatblock = gr.Dropdown(label="Chatblocks", choices=[c for c in range(0, 21)], value=0, type="index")
+            header = gr.Checkbox(label="Include header?", value=True)
+    rand.change(check_rand, inputs=[rand, seed], outputs=seed)
+    btn.click(chat_inf, inputs=[sys_inp, inp, chat_b, memory, client_choice, seed, temp, tokens, top_p, rep_p, chat_mem, custom_prompt], outputs=[chat_b, memory])
+    stop_btn.click(lambda: None, [], chat_b.stop)
+    clear_btn.click(clear_fn, [], [inp, sys_inp, chat_b, memory])
+    im_btn.click(get_screenshot, inputs=[chat_b, im_height, im_width, chatblock, theme, wait_time, header], outputs=img)
+
+app.launch()