llama

app.py

@@ -1,151 +1,149 @@
-import subprocess
 import os
-import torch
 from threading import Thread
-from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer, StoppingCriteriaList, StoppingCriteria
+from typing import Iterator
+
 import gradio as gr
-import time
 import spaces
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
 
-# Decorador para asegurar que la función se ejecute en la GPU
-@spaces.GPU
-def greet(n):
-    zero = torch.tensor(0, device='cuda:0')  # Inicializa el tensor en GPU
-    print(zero.device)  # Confirma que está en 'cuda:0'
-    return f"Hello {zero.item() + n} Tensor"
-
-# Instalar dependencias necesarias sin recompilar CUDA
-subprocess.run(
-    'pip install flash-attn --no-build-isolation',
-    env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"},
-    shell=True
-)
+MAX_MAX_NEW_TOKENS = 2048
+DEFAULT_MAX_NEW_TOKENS = 1024
+MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
 
-# Cargar el token desde las variables de entorno
-HF_TOKEN = os.environ.get("HUGGINGFACE_HUB_TOKEN", None)
-
-# Cargar el modelo y el tokenizador
-MODEL_NAME = "EleutherAI/gpt-j-6B"
-model = AutoModelForCausalLM.from_pretrained(
-    MODEL_NAME,
-    torch_dtype=torch.bfloat16,
-    device_map="auto",
-    trust_remote_code=True,
-    use_auth_token=HF_TOKEN
-).eval()
-
-tokenizer = AutoTokenizer.from_pretrained(
-    MODEL_NAME, 
-    trust_remote_code=True, 
-    use_fast=False, 
-    use_auth_token=HF_TOKEN
-)
+DESCRIPTION = """\
+# Llama-2 13B Chat
+
+This Space demonstrates model [Llama-2-13b-chat](https://huggingface.co/meta-llama/Llama-2-13b-chat) by Meta, a Llama 2 model with 13B parameters fine-tuned for chat instructions. Feel free to play with it, or duplicate to run generations without a queue! If you want to run your own service, you can also [deploy the model on Inference Endpoints](https://huggingface.co/inference-endpoints).
+
+🔎 For more details about the Llama 2 family of models and how to use them with `transformers`, take a look [at our blog post](https://huggingface.co/blog/llama2).
+
+🔨 Looking for an even more powerful model? Check out the large [**70B** model demo](https://huggingface.co/spaces/ysharma/Explore_llamav2_with_TGI).
+🐇 For a smaller model that you can run on many GPUs, check our [7B model demo](https://huggingface.co/spaces/huggingface-projects/llama-2-7b-chat).
 
-# Ejecutar la función greet para prueba inicial
-print(greet(5))  # Ejemplo: Cambia 5 por el valor deseado para probar la función
-
-# Títulos y estilos para la interfaz
-TITLE = "<h1><center>Gemma Model Chat</center></h1>"
-PLACEHOLDER = f'<h3><center>{MODEL_NAME} es un modelo avanzado capaz de generar respuestas detalladas basadas en entradas complejas.</center></h3>'
-CSS = """
-.duplicate-button {
-  margin: auto !important;
-  color: white !important;
-  background: black !important;
-  border-radius: 100vh !important;
-}
 """
 
-# Definición de criterios de parada personalizados
-class StopOnTokens(StoppingCriteria):
-    def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor, **kwargs) -> bool:
-        stop_ids = [tokenizer.eos_token_id]
-        return any(input_ids[0][-1] == stop_id for stop_id in stop_ids)
+LICENSE = """
+<p/>
 
-# Función para generar respuestas del chatbot con streaming
-def stream_chat(message: str, history: list, temperature: float, max_new_tokens: int):
-    print(f'Mensaje: {message}')
-    print(f'Historia: {history}')
+---
+As a derivate work of [Llama-2-13b-chat](https://huggingface.co/meta-llama/Llama-2-13b-chat) by Meta,
+this demo is governed by the original [license](https://huggingface.co/spaces/huggingface-projects/llama-2-13b-chat/blob/main/LICENSE.txt) and [acceptable use policy](https://huggingface.co/spaces/huggingface-projects/llama-2-13b-chat/blob/main/USE_POLICY.md).
+"""
 
-    # Limpieza del historial para evitar pares con 'None'
-    cleaned_history = [[prompt, answer if answer is not None else ""] for prompt, answer in history]
+if not torch.cuda.is_available():
+    DESCRIPTION += "\n<p>Running on CPU 🥶 This demo does not work on CPU.</p>"
 
-    stop = StopOnTokens()
 
-    # Verificar y asignar pad_token_id si es None
-    if tokenizer.pad_token_id is None:
-        tokenizer.pad_token_id = tokenizer.eos_token_id
+if torch.cuda.is_available():
+    model_id = "meta-llama/Llama-2-13b-chat-hf"
+    model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", load_in_4bit=True)
+    tokenizer = AutoTokenizer.from_pretrained(model_id)
+    tokenizer.use_default_system_prompt = False
 
-    # Preparar los input_ids y manejar la máscara de atención
-    input_ids = tokenizer.encode(message, return_tensors='pt').to(next(model.parameters()).device)
-    attention_mask = input_ids.ne(tokenizer.pad_token_id).long()
-    streamer = TextIteratorStreamer(tokenizer, timeout=60.0, skip_prompt=True, skip_special_tokens=True)
 
+@spaces.GPU
+def generate(
+    message: str,
+    chat_history: list[tuple[str, str]],
+    system_prompt: str,
+    max_new_tokens: int = 1024,
+    temperature: float = 0.6,
+    top_p: float = 0.9,
+    top_k: int = 50,
+    repetition_penalty: float = 1.2,
+) -> Iterator[str]:
+    conversation = []
+    if system_prompt:
+        conversation.append({"role": "system", "content": system_prompt})
+    for user, assistant in chat_history:
+        conversation.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])
+    conversation.append({"role": "user", "content": message})
+
+    input_ids = tokenizer.apply_chat_template(conversation, return_tensors="pt")
+    if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
+        input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
+        gr.Warning(f"Trimmed input from conversation as it was longer than {MAX_INPUT_TOKEN_LENGTH} tokens.")
+    input_ids = input_ids.to(model.device)
+
+    streamer = TextIteratorStreamer(tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True)
     generate_kwargs = dict(
-        input_ids=input_ids,
-        attention_mask=attention_mask,
+        {"input_ids": input_ids},
         streamer=streamer,
         max_new_tokens=max_new_tokens,
         do_sample=True,
-        top_k=50,
+        top_p=top_p,
+        top_k=top_k,
         temperature=temperature,
-        repetition_penalty=1.1,
-        stopping_criteria=StoppingCriteriaList([stop]),
-        pad_token_id=tokenizer.pad_token_id
+        num_beams=1,
+        repetition_penalty=repetition_penalty,
     )
+    t = Thread(target=model.generate, kwargs=generate_kwargs)
+    t.start()
+
+    outputs = []
+    for text in streamer:
+        outputs.append(text)
+        yield "".join(outputs)
+
+
+chat_interface = gr.ChatInterface(
+    fn=generate,
+    additional_inputs=[
+        gr.Textbox(label="System prompt", lines=6),
+        gr.Slider(
+            label="Max new tokens",
+            minimum=1,
+            maximum=MAX_MAX_NEW_TOKENS,
+            step=1,
+            value=DEFAULT_MAX_NEW_TOKENS,
+        ),
+        gr.Slider(
+            label="Temperature",
+            minimum=0.1,
+            maximum=4.0,
+            step=0.1,
+            value=0.6,
+        ),
+        gr.Slider(
+            label="Top-p (nucleus sampling)",
+            minimum=0.05,
+            maximum=1.0,
+            step=0.05,
+            value=0.9,
+        ),
+        gr.Slider(
+            label="Top-k",
+            minimum=1,
+            maximum=1000,
+            step=1,
+            value=50,
+        ),
+        gr.Slider(
+            label="Repetition penalty",
+            minimum=1.0,
+            maximum=2.0,
+            step=0.05,
+            value=1.2,
+        ),
+    ],
+    stop_btn=None,
+    examples=[
+        ["Hello there! How are you doing?"],
+        ["Can you explain briefly to me what is the Python programming language?"],
+        ["Explain the plot of Cinderella in a sentence."],
+        ["How many hours does it take a man to eat a Helicopter?"],
+        ["Write a 100-word article on 'Benefits of Open-Source in AI research'"],
+    ],
+    cache_examples=False,
+)
 
-    # Ejecutar la generación de tokens en un hilo separado
-    thread = Thread(target=model.generate, kwargs=generate_kwargs)
-    thread.start()
-    buffer = ""
-
-    # Procesar el streaming de tokens y formatear la respuesta para Gradio
-    for new_token in streamer:
-        if new_token:
-            buffer += new_token
-            # Asegúrate de que solo estás trabajando con texto puro
-            buffer = buffer.strip()  # Eliminar espacios innecesarios
-            # Emitir el texto acumulado en un formato compatible con Gradio: [[Mensaje del usuario, Respuesta del bot]]
-            yield cleaned_history + [[message, buffer]]
-
-
-# Configuración de la interfaz Gradio
-chatbot = gr.Chatbot(height=600, placeholder=PLACEHOLDER)
-
-with gr.Blocks(css=CSS) as demo:
-    gr.HTML(TITLE)
-    gr.DuplicateButton(value="Duplicate Space for private use", elem_classes="duplicate-button")
-    gr.ChatInterface(
-        fn=stream_chat,
-        chatbot=chatbot,
-        fill_height=True,
-        additional_inputs_accordion=gr.Accordion(label="⚙️ Parameters", open=False, render=False),
-        additional_inputs=[
-            gr.Slider(
-                minimum=0,
-                maximum=1,
-                step=0.1,
-                value=0.5,
-                label="Temperature",
-                render=False,
-            ),
-            gr.Slider(
-                minimum=1024,
-                maximum=32768,
-                step=1,
-                value=4096,
-                label="Max New Tokens",
-                render=False,
-            ),
-        ],
-        examples=[
-            ["Help me study vocabulary: write a sentence for me to fill in the blank, and I'll try to pick the correct option."],
-            ["What are 5 creative things I could do with my kids' art? I don't want to throw them away, but it's also so much clutter."],
-            ["Tell me a random fun fact about the Roman Empire."],
-            ["Show me a code snippet of a website's sticky header in CSS and JavaScript."],
-        ],
-        cache_examples=False,
-    )
+with gr.Blocks(css="style.css", fill_height=True) as demo:
+    gr.Markdown(DESCRIPTION)
+    gr.DuplicateButton(value="Duplicate Space for private use", elem_id="duplicate-button")
+    chat_interface.render()
+    gr.Markdown(LICENSE)
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.queue(max_size=20).launch()
+