Update app.py

app.py

@@ -7,21 +7,33 @@ import streamlit as st
 from huggingface_hub import InferenceClient, AsyncInferenceClient
 from gradio_client import Client, handle_file
 import asyncio
+from concurrent.futures import ThreadPoolExecutor
 
 MAX_SEED = np.iinfo(np.int32).max
 HF_TOKEN = os.environ.get("HF_TOKEN")
 HF_TOKEN_UPSCALER = os.environ.get("HF_TOKEN_UPSCALER")
 client = AsyncInferenceClient()
+llm_client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
 DATA_PATH = Path("./data")
 DATA_PATH.mkdir(exist_ok=True)
 
-async def generate_image(prompt, model, width, height, scales, steps, seed):
+def run_async(func):
+    loop = asyncio.new_event_loop()
+    asyncio.set_event_loop(loop)
+    executor = ThreadPoolExecutor(max_workers=1)
+    result = loop.run_in_executor(executor, func)
+    return loop.run_until_complete(result)
+
+def enable_lora(lora_add, basemodel):
+    return lora_add if lora_add else basemodel
+
+async def generate_image(combined_prompt, model, width, height, scales, steps, seed):
     try:
         if seed == -1:
             seed = random.randint(0, MAX_SEED)
         seed = int(seed)
         image = await client.text_to_image(
-            prompt=prompt, height=height, width=width, guidance_scale=scales,
+            prompt=combined_prompt, height=height, width=width, guidance_scale=scales,
             num_inference_steps=steps, model=model
         )
         return image, seed
@@ -38,6 +50,63 @@ def get_upscale_finegrain(prompt, img_path, upscale_factor):
     except Exception as e:
         return None
 
+def save_prompt(prompt_text, seed):
+    try:
+        prompt_file_path = DATA_PATH / f"prompt_{seed}.txt"
+        with open(prompt_file_path, "w") as prompt_file:
+            prompt_file.write(prompt_text)
+        return prompt_file_path
+    except Exception as e:
+        st.error(f"Error al guardar el prompt: {e}")
+        return None
+
+async def gen(prompt, basemodel, width, height, scales, steps, seed, upscale_factor, process_upscale, lora_model, process_lora, process_enhancer):
+    model = enable_lora(lora_model, basemodel) if process_lora else basemodel
+    combined_prompt = prompt  
+
+    if process_enhancer:
+        improved_prompt = await improve_prompt(prompt)
+        combined_prompt = f"{prompt} {improved_prompt}"
+    
+    if seed == -1:
+        seed = random.randint(0, MAX_SEED)
+    seed = int(seed)
+    progress_bar = st.progress(0)
+    image, seed = await generate_image(combined_prompt, model, width, height, scales, steps, seed)
+    progress_bar.progress(50)
+
+    if isinstance(image, str) and image.startswith("Error"):
+        progress_bar.empty()
+        return [image, None, combined_prompt]
+
+    image_path = save_image(image, seed)
+    prompt_file_path = save_prompt(combined_prompt, seed)
+
+    if process_upscale:
+        upscale_image_path = get_upscale_finegrain(combined_prompt, image_path, upscale_factor)
+        if upscale_image_path:
+            upscale_image = Image.open(upscale_image_path)
+            upscale_image.save(DATA_PATH / f"upscale_image_{seed}.jpg", format="JPEG")
+            progress_bar.progress(100)
+            image_path.unlink()  
+            return [str(DATA_PATH / f"upscale_image_{seed}.jpg"), str(prompt_file_path)]
+        else:
+            progress_bar.empty()
+            return [str(image_path), str(prompt_file_path)]
+    else:
+        progress_bar.progress(100)
+        return [str(image_path), str(prompt_file_path)]
+
+async def improve_prompt(prompt):
+    try:
+        instruction = ("With this idea, describe in English a detailed txt2img prompt in 500 characters at most, add ilumination, admosphere, cinematic and characters...")
+        formatted_prompt = f"{prompt}: {instruction}"
+        response = llm_client.text_generation(formatted_prompt, max_new_tokens=300)
+        improved_text = response['generated_text'].strip() if 'generated_text' in response else response.strip()
+        return improved_text[:300] if len(improved_text) > 300 else improved_text
+    except Exception as e:
+        return f"Error mejorando el prompt: {e}"
+
 def save_image(image, seed):
     try:
         image_path = DATA_PATH / f"image_{seed}.jpg"
@@ -50,7 +119,8 @@ def save_image(image, seed):
 def get_storage():
     files = [file for file in DATA_PATH.glob("*.jpg") if file.is_file()]
     files.sort(key=lambda x: x.stat().st_mtime, reverse=True)
-    return [str(file.resolve()) for file in files]
+    usage = sum([file.stat().st_size for file in files])
+    return [str(file.resolve()) for file in files], f"Uso total: {usage/(1024.0 ** 3):.3f}GB"
 
 def get_prompts():
     prompt_files = [file for file in DATA_PATH.glob("*.txt") if file.is_file()]
@@ -68,18 +138,19 @@ def delete_image(image_path):
 
 def main():
     st.set_page_config(layout="wide")
-    st.title("Generación de Imágenes")
+    st.title("FLUX with prompt enhancer and upscaler with LORA model training")
     
-    prompt = st.text_input("Descripción de la imagen", max_chars=200)
-
-    with st.expander("Opciones avanzadas", expanded=False):
-        basemodel = st.selectbox("Modelo Base", ["black-forest-labs/FLUX.1-schnell", "black-forest-labs/FLUX.1-DEV"])
-        format_option = st.selectbox("Formato", ["9:16", "16:9"])
-        process_upscale = st.checkbox("Procesar Escalador", value=True)
-        upscale_factor = st.selectbox("Factor de Escala", [2, 4, 8], index=0)
-        scales = st.slider("Escalado", 1, 20, 10)
-        steps = st.slider("Pasos", 1, 100, 20)
-        seed = st.number_input("Semilla", value=-1)
+    prompt = st.sidebar.text_input("Descripción de la imagen", max_chars=200)
+    process_enhancer = st.sidebar.checkbox("Mejorar Prompt", value=True)  # Nuevo checkbox
+    basemodel = st.sidebar.selectbox("Modelo Base", ["black-forest-labs/FLUX.1-schnell", "black-forest-labs/FLUX.1-DEV"])
+    lora_model = st.sidebar.selectbox("LORA Realismo", ["Shakker-Labs/FLUX.1-dev-LoRA-add-details", "XLabs-AI/flux-RealismLora"])
+    format_option = st.sidebar.selectbox("Formato", ["9:16", "16:9"])
+    process_lora = st.sidebar.checkbox("Procesar LORA", value=True)
+    process_upscale = st.sidebar.checkbox("Procesar Escalador", value=True)
+    upscale_factor = st.sidebar.selectbox("Factor de Escala", [2, 4, 8], index=0)
+    scales = st.sidebar.slider("Escalado", 1, 20, 10)
+    steps = st.sidebar.slider("Pasos", 1, 100, 20)
+    seed = st.sidebar.number_input("Semilla", value=-1)
 
     if format_option == "9:16":
         width = 720
@@ -88,35 +159,48 @@ def main():
         width = 1280
         height = 720
 
-    if st.button("Generar Imagen"):
-        with st.spinner("Generando imagen..."):
-            image, seed = await generate_image(prompt, basemodel, width, height, scales, steps, seed)
-            
-            if isinstance(image, str) and image.startswith("Error"):
-                st.error(image)
+    if st.sidebar.button("Generar Imagen"):
+        with st.spinner("Mejorando y generando imagen..."):
+            result = asyncio.run(gen(prompt, basemodel, width, height, scales, steps, seed, upscale_factor, process_upscale, lora_model, process_lora, process_enhancer))
+            image_paths = result[0]
+            prompt_file = result[1]
+
+        st.write(f"Image paths: {image_paths}")
+
+        if image_paths:
+            if Path(image_paths).exists():
+                st.image(image_paths, caption="Imagen Generada")
             else:
-                image_path = save_image(image, seed)
-                if image_path:
-                    st.image(image_path, caption="Imagen Generada")
-                    st.success("Imagen generada y guardada.")
-    
-    # Mostrar galería de imágenes
-    files = get_storage()
-    prompts = get_prompts()
+                st.error("El archivo de imagen no existe.")
 
-    st.subheader("Galería de Imágenes")
-    cols = st.columns(3)
+            if prompt_file and Path(prompt_file).exists():
+                prompt_text = Path(prompt_file).read_text()
+                st.write(f"Prompt utilizado: {prompt_text}")
+            else:
+                st.write("El archivo del prompt no está disponible.")
+
+    files, usage = get_storage()
+    st.text(usage)
+    cols = st.columns(6)
+    prompts = get_prompts()
 
     for idx, file in enumerate(files):
-        with cols[idx % 3]:
+        with cols[idx % 6]:
             image = Image.open(file)
-            prompt_text = prompts.get(Path(file).stem.replace("image_", ""), "No disponible")
+            prompt_file = prompts.get(Path(file).stem.replace("image_", ""), None)
+            prompt_text = Path(prompt_file).read_text() if prompt_file else "No disponible"
             
-            st.image(image, caption=f"Imagen {idx + 1}")
+            st.image(image, caption=f"Imagen {idx+1}")
             st.write(f"Prompt: {prompt_text}")
             
-            if st.button(f"Borrar Imagen {idx + 1}", key=f"delete_{idx}"):
-                delete_image(file)
+            if st.button(f"Borrar Imagen {idx+1}", key=f"delete_{idx}"):
+                try:
+                    os.remove(file)  
+                    if prompt_file:
+                        os.remove(prompt_file)
+                    st.success(f"Imagen {idx+1} y su prompt fueron borrados.")
+                except Exception as e:
+                    st.error(f"Error al borrar la imagen o prompt: {e}")
 
 if __name__ == "__main__":
-    main()
+    main()