Update app.py

app.py

@@ -14,7 +14,6 @@ from io import BytesIO
 import boto3
 from botocore.exceptions import NoCredentialsError
 from huggingface_hub import snapshot_download
-import asyncio
 import shutil
 
 # Configuración global
@@ -72,6 +71,12 @@ class S3ModelLoader:
                 print(f"Modelo {model_name} ya existe en S3.")
             except self.s3_client.exceptions.ClientError:
                 print(f"Modelo {model_name} no existe en S3. Descargando desde Hugging Face...")
+
+                # Eliminar caché local de Hugging Face (si existe)
+                local_cache_dir = os.path.join(os.getenv("HOME"), ".cache/huggingface/hub/models--")
+                if os.path.exists(local_cache_dir):
+                    shutil.rmtree(local_cache_dir)
+
                 model_path = snapshot_download(model_name, token=HUGGINGFACE_HUB_TOKEN)
 
                 # Cargar el modelo y tokenizer
@@ -212,25 +217,6 @@ async def generate_image(request: GenerateRequest):
     except Exception as e:
         raise HTTPException(status_code=500, detail=f"Internal server error: {str(e)}")
 
-# Endpoint para generar texto a voz
-@app.post("/generate-text-to-speech")
-async def generate_text_to_speech(request: GenerateRequest):
-    try:
-        validated_body = request
-        device = "cuda" if torch.cuda.is_available() else "cpu"
-        
-        audio_generator = pipeline("text-to-speech", model=validated_body.model_name, device=device)
-        audio = audio_generator(validated_body.input_text)[0]
-
-        audio_byte_arr = BytesIO()
-        audio.save(audio_byte_arr, format="WAV")
-        audio_byte_arr.seek(0)
-
-        return StreamingResponse(audio_byte_arr, media_type="audio/wav")
-
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=f"Internal server error: {str(e)}")
-
 # Ejecutar el servidor FastAPI con Uvicorn
 if __name__ == "__main__":
     import uvicorn