File size: 13,751 Bytes
8a53057
 
e0c4be3
 
 
 
 
 
8a53057
e0c4be3
 
 
 
635fc93
 
e0c4be3
635fc93
 
e0c4be3
635fc93
 
 
 
 
 
 
 
 
 
e0c4be3
 
6a6684e
e0c4be3
635fc93
 
 
 
 
 
6a6684e
635fc93
 
 
 
 
 
 
e0c4be3
 
8a53057
e0c4be3
 
8a53057
e0c4be3
635fc93
e0c4be3
 
 
635fc93
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e0c4be3
 
 
 
 
 
 
 
8a53057
e0c4be3
6a6684e
e0c4be3
 
635fc93
6a6684e
e0c4be3
 
6a6684e
 
e0c4be3
 
6a6684e
 
e0c4be3
 
 
 
 
 
 
8a53057
e0c4be3
635fc93
e0c4be3
 
 
 
 
635fc93
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e0c4be3
 
 
 
 
8a53057
e0c4be3
6a6684e
e0c4be3
6a6684e
 
 
 
e0c4be3
 
 
6a6684e
e0c4be3
 
 
6a6684e
e0c4be3
 
6a6684e
e0c4be3
 
 
6a6684e
e0c4be3
8a53057
e0c4be3
7998696
e0c4be3
 
 
 
 
7998696
e0c4be3
7998696
 
 
 
 
e0c4be3
 
 
 
 
 
 
 
 
 
 
 
 
8a53057
e0c4be3
 
 
 
 
 
7998696
 
 
 
 
e0c4be3
 
7998696
 
 
 
8a53057
e0c4be3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8a53057
e0c4be3
 
 
8a53057
e0c4be3
 
635fc93
e0c4be3
 
 
 
 
 
 
7998696
 
e0c4be3
 
 
 
 
 
 
 
 
 
 
7998696
e0c4be3
 
 
 
 
7998696
e0c4be3
 
635fc93
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e0c4be3
 
8a53057
e0c4be3
 
 
6a6684e
e0c4be3
 
 
 
 
 
 
 
 
 
635fc93
e0c4be3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8a53057
e0c4be3
8a53057
e0c4be3
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
import gradio as gr
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
from diffusers import StableDiffusionPipeline, DiffusionPipeline
import requests
from PIL import Image
import io
import base64

# Configuraci贸n de modelos libres
MODELS = {
    "text": {
        "microsoft/DialoGPT-medium": "Chat conversacional",
        "microsoft/DialoGPT-large": "Chat conversacional avanzado",
        "microsoft/DialoGPT-small": "Chat conversacional r谩pido",
        "gpt2": "Generaci贸n de texto",
        "gpt2-medium": "GPT-2 mediano",
        "gpt2-large": "GPT-2 grande",
        "distilgpt2": "GPT-2 optimizado",
        "EleutherAI/gpt-neo-125M": "GPT-Neo peque帽o",
        "EleutherAI/gpt-neo-1.3B": "GPT-Neo mediano",
        "microsoft/DialoGPT-medium": "Chat conversacional",
        "facebook/opt-125m": "OPT peque帽o",
        "facebook/opt-350m": "OPT mediano",
        "bigscience/bloom-560m": "BLOOM multiling眉e",
        "bigscience/bloom-1b1": "BLOOM grande",
        "microsoft/DialoGPT-medium": "Chat conversacional",
        "Helsinki-NLP/opus-mt-es-en": "Traductor espa帽ol-ingl茅s",
        "Helsinki-NLP/opus-mt-en-es": "Traductor ingl茅s-espa帽ol"
    },
    "image": {
        "CompVis/stable-diffusion-v1-4": "Stable Diffusion v1.4 (B谩sico)",
        "runwayml/stable-diffusion-v1-5": "Stable Diffusion v1.5",
        "stabilityai/stable-diffusion-2-1": "Stable Diffusion 2.1",
        "stabilityai/stable-diffusion-xl-base-1.0": "SDXL Base",
        "stabilityai/stable-diffusion-xl-refiner-1.0": "SDXL Refiner",
        "prompthero/openjourney": "Midjourney style",
        "dreamlike-art/dreamlike-photoreal-2.0": "Fotorealista",
        "nitrosocke/Ghibli-Diffusion": "Estilo Studio Ghibli",
        "nitrosocke/mo-di-diffusion": "Estilo moderno"
    },
    "chat": {
        "microsoft/DialoGPT-medium": "Chat conversacional",
        "microsoft/DialoGPT-large": "Chat conversacional avanzado",
        "microsoft/DialoGPT-small": "Chat conversacional r谩pido",
        "facebook/opt-350m": "OPT conversacional",
        "bigscience/bloom-560m": "BLOOM multiling眉e"
    }
}

# Cache para los modelos
model_cache = {}

def load_text_model(model_name):
    """Cargar modelo de texto con soporte para diferentes tipos"""
    if model_name not in model_cache:
        print(f"Cargando modelo de texto: {model_name}")
        
        # Detectar tipo de modelo
        if "opus-mt" in model_name.lower():
            # Modelo de traducci贸n
            from transformers import MarianMTModel, MarianTokenizer
            tokenizer = MarianTokenizer.from_pretrained(model_name)
            model = MarianMTModel.from_pretrained(model_name)
        else:
            # Modelo de generaci贸n de texto
            tokenizer = AutoTokenizer.from_pretrained(model_name)
            model = AutoModelForCausalLM.from_pretrained(model_name)
            
            # Configurar para chat si es DialoGPT
            if "dialogpt" in model_name.lower():
                tokenizer.pad_token = tokenizer.eos_token
                model.config.pad_token_id = model.config.eos_token_id
        
        model_cache[model_name] = {
            "tokenizer": tokenizer,
            "model": model,
            "type": "text"
        }
    
    return model_cache[model_name]

def load_image_model(model_name):
    """Cargar modelo de imagen - versi贸n simplificada"""
    if model_name not in model_cache:
        print(f"Cargando modelo de imagen: {model_name}")
        
        # Configuraci贸n b谩sica sin optimizaciones complejas
        pipe = StableDiffusionPipeline.from_pretrained(
            model_name,
            torch_dtype=torch.float32,
            safety_checker=None
        )
        
        # Solo optimizaci贸n b谩sica de memoria
        pipe.enable_attention_slicing()
        
        model_cache[model_name] = {
            "pipeline": pipe,
            "type": "image"
        }
    
    return model_cache[model_name]

def generate_text(prompt, model_name, max_length=100):
    """Generar texto con el modelo seleccionado - mejorado para diferentes tipos"""
    try:
        model_data = load_text_model(model_name)
        tokenizer = model_data["tokenizer"]
        model = model_data["model"]
        
        # Detectar si es modelo de traducci贸n
        if "opus-mt" in model_name.lower():
            # Traducci贸n
            inputs = tokenizer.encode(prompt, return_tensors="pt", max_length=512, truncation=True)
            with torch.no_grad():
                outputs = model.generate(inputs, max_length=max_length, num_beams=4, early_stopping=True)
            response = tokenizer.decode(outputs[0], skip_special_tokens=True)
        else:
            # Generaci贸n de texto
            inputs = tokenizer.encode(prompt, return_tensors="pt")
            
            # Generar
            with torch.no_grad():
                outputs = model.generate(
                    inputs,
                    max_length=max_length,
                    num_return_sequences=1,
                    temperature=0.7,
                    do_sample=True,
                    pad_token_id=tokenizer.eos_token_id
                )
            
            # Decodificar respuesta
            response = tokenizer.decode(outputs[0], skip_special_tokens=True)
            
            # Para DialoGPT, extraer solo la respuesta del asistente
            if "dialogpt" in model_name.lower():
                response = response.replace(prompt, "").strip()
        
        return response
    
    except Exception as e:
        return f"Error generando texto: {str(e)}"

def generate_image(prompt, model_name, num_inference_steps=20):
    """Generar imagen con el modelo seleccionado - versi贸n simplificada"""
    try:
        print(f"Generando imagen con modelo: {model_name}")
        print(f"Prompt: {prompt}")
        print(f"Pasos: {num_inference_steps}")
        
        model_data = load_image_model(model_name)
        pipeline = model_data["pipeline"]
        
        # Configuraci贸n b谩sica
        image = pipeline(
            prompt,
            num_inference_steps=num_inference_steps,
            guidance_scale=7.5
        ).images[0]
        
        print("Imagen generada exitosamente")
        return image
    
    except Exception as e:
        print(f"Error generando imagen: {str(e)}")
        return f"Error generando imagen: {str(e)}"

def chat_with_model(message, history, model_name):
    """Funci贸n de chat para DialoGPT con formato de mensajes actualizado"""
    try:
        model_data = load_text_model(model_name)
        tokenizer = model_data["tokenizer"]
        model = model_data["model"]
        
        # Construir historial de conversaci贸n desde el nuevo formato
        conversation = ""
        for msg in history:
            if msg["role"] == "user":
                conversation += f"User: {msg['content']}\n"
            elif msg["role"] == "assistant":
                conversation += f"Assistant: {msg['content']}\n"
        
        conversation += f"User: {message}\nAssistant:"
        
        # Generar respuesta
        inputs = tokenizer.encode(conversation, return_tensors="pt", truncation=True, max_length=512)
        
        with torch.no_grad():
            outputs = model.generate(
                inputs,
                max_length=inputs.shape[1] + 50,
                temperature=0.7,
                do_sample=True,
                pad_token_id=tokenizer.eos_token_id
            )
        
        response = tokenizer.decode(outputs[0], skip_special_tokens=True)
        
        # Extraer solo la respuesta del asistente
        response = response.split("Assistant:")[-1].strip()
        
        # Retornar el historial actualizado con el nuevo formato
        history.append({"role": "user", "content": message})
        history.append({"role": "assistant", "content": response})
        
        return history
    
    except Exception as e:
        error_msg = f"Error en el chat: {str(e)}"
        history.append({"role": "user", "content": message})
        history.append({"role": "assistant", "content": error_msg})
        return history

# Interfaz de Gradio
with gr.Blocks(title="Modelos Libres de IA", theme=gr.themes.Soft()) as demo:
    gr.Markdown("# 馃 Modelos Libres de IA")
    gr.Markdown("### Genera texto e im谩genes sin l铆mites de cuota")
    
    with gr.Tabs():
        # Tab de Generaci贸n de Texto
        with gr.TabItem("馃摑 Generaci贸n de Texto"):
            with gr.Row():
                with gr.Column():
                    text_model = gr.Dropdown(
                        choices=list(MODELS["text"].keys()),
                        value="microsoft/DialoGPT-medium",
                        label="Modelo de Texto"
                    )
                    text_prompt = gr.Textbox(
                        label="Prompt",
                        placeholder="Escribe tu prompt aqu铆...",
                        lines=3
                    )
                    max_length = gr.Slider(
                        minimum=50,
                        maximum=200,
                        value=100,
                        step=10,
                        label="Longitud m谩xima"
                    )
                    text_btn = gr.Button("Generar Texto", variant="primary")
                
                with gr.Column():
                    text_output = gr.Textbox(
                        label="Resultado",
                        lines=10,
                        interactive=False
                    )
            
            text_btn.click(
                generate_text,
                inputs=[text_prompt, text_model, max_length],
                outputs=text_output
            )
        
        # Tab de Chat
        with gr.TabItem("馃挰 Chat"):
            with gr.Row():
                with gr.Column():
                    chat_model = gr.Dropdown(
                        choices=list(MODELS["chat"].keys()),
                        value="microsoft/DialoGPT-medium",
                        label="Modelo de Chat"
                    )
                
                with gr.Column():
                    chatbot = gr.Chatbot(
                        label="Chat",
                        height=400,
                        type="messages"
                    )
                    chat_input = gr.Textbox(
                        label="Mensaje",
                        placeholder="Escribe tu mensaje...",
                        lines=2
                    )
                    chat_btn = gr.Button("Enviar", variant="primary")
            
            chat_btn.click(
                chat_with_model,
                inputs=[chat_input, chatbot, chat_model],
                outputs=[chatbot]
            )
            
            chat_input.submit(
                chat_with_model,
                inputs=[chat_input, chatbot, chat_model],
                outputs=[chatbot]
            )
        
        # Tab de Traducci贸n
        with gr.TabItem("馃寪 Traducci贸n"):
            with gr.Row():
                with gr.Column():
                    translate_model = gr.Dropdown(
                        choices=["Helsinki-NLP/opus-mt-es-en", "Helsinki-NLP/opus-mt-en-es"],
                        value="Helsinki-NLP/opus-mt-es-en",
                        label="Modelo de Traducci贸n"
                    )
                    translate_text = gr.Textbox(
                        label="Texto a traducir",
                        placeholder="Escribe el texto que quieres traducir...",
                        lines=3
                    )
                    translate_btn = gr.Button("Traducir", variant="primary")
                
                with gr.Column():
                    translate_output = gr.Textbox(
                        label="Traducci贸n",
                        lines=3,
                        interactive=False
                    )
            
            translate_btn.click(
                generate_text,
                inputs=[translate_text, translate_model, gr.Slider(value=100, visible=False)],
                outputs=translate_output
            )
        
        # Tab de Generaci贸n de Im谩genes
        with gr.TabItem("馃帹 Generaci贸n de Im谩genes"):
            with gr.Row():
                with gr.Column():
                    image_model = gr.Dropdown(
                        choices=list(MODELS["image"].keys()),
                        value="CompVis/stable-diffusion-v1-4",
                        label="Modelo de Imagen"
                    )
                    image_prompt = gr.Textbox(
                        label="Prompt de Imagen",
                        placeholder="Describe la imagen que quieres generar...",
                        lines=3
                    )
                    steps = gr.Slider(
                        minimum=10,
                        maximum=50,
                        value=15,
                        step=5,
                        label="Pasos de inferencia"
                    )
                    image_btn = gr.Button("Generar Imagen", variant="primary")
                
                with gr.Column():
                    image_output = gr.Image(
                        label="Imagen Generada",
                        type="pil"
                    )
            
            image_btn.click(
                generate_image,
                inputs=[image_prompt, image_model, steps],
                outputs=image_output
            )

# Configuraci贸n para Hugging Face Spaces
if __name__ == "__main__":
    demo.launch(
        server_name="0.0.0.0",
        server_port=7860,
        share=False
    )