Update app.py

app.py

@@ -7,6 +7,8 @@ import numpy as np
 import matplotlib.pyplot as plt
 import matplotlib.dates as mdates
 from sklearn.metrics import mean_absolute_error, mean_squared_error
+import tempfile
+
 def get_popular_tickers():
     return [
         "AAPL", "MSFT", "GOOGL", "AMZN", "META", "TSLA", "NVDA", "JPM",
@@ -15,6 +17,10 @@ def get_popular_tickers():
 
 def predict_stock(ticker, train_data_points, prediction_days):
     try:
+        # Asegurar que los parámetros sean enteros
+        train_data_points = int(train_data_points)
+        prediction_days = int(prediction_days)
+        
         # Configurar el pipeline
         pipeline = ChronosPipeline.from_pretrained(
             "amazon/chronos-t5-mini",
@@ -84,7 +90,6 @@ def predict_stock(ticker, train_data_points, prediction_days):
         
         # Calcular métricas si hay datos reales para comparar
         overlap_end_index = train_data_points + prediction_days
-        validation_data = None
         if overlap_end_index <= total_points:
             real_future_dates = df['Date'][train_data_points:overlap_end_index]
             real_future_data = df[f'{ticker}_Close'][train_data_points:overlap_end_index].values
@@ -129,7 +134,8 @@ def predict_stock(ticker, train_data_points, prediction_days):
         
         plt.tight_layout()
         
-        # Crear DataFrame para descarga
+        # Crear un archivo temporal para el CSV
+        temp_csv = tempfile.NamedTemporaryFile(delete=False, suffix='.csv')
         prediction_df = pd.DataFrame({
             'Date': prediction_dates,
             'Predicted_Price': median,
@@ -145,10 +151,12 @@ def predict_stock(ticker, train_data_points, prediction_days):
             prediction_df = prediction_df[prediction_df['Date'].isin(matching_dates)]
             prediction_df['Real_Price'] = real_future_data[:len(prediction_df)]
         
-        csv_path = f"predictions_{ticker}.csv"
-        prediction_df.to_csv(csv_path, index=False)
+        # Guardar el DataFrame en el archivo temporal
+        prediction_df.to_csv(temp_csv.name, index=False)
+        temp_csv.close()
         
-        return plt, csv_path
+        # Retornar el gráfico y la ruta del archivo CSV
+        return plt, temp_csv.name
             
     except Exception as e:
         print(f"Error: {str(e)}")
@@ -166,7 +174,7 @@ with gr.Blocks() as demo:
             )
             train_data_points = gr.Slider(
                 minimum=50,
-                maximum=5000,  # Puedes ajustar este valor si lo deseas
+                maximum=5000,
                 value=1000,
                 step=1,
                 label="Número de Datos para Entrenamiento"
@@ -179,7 +187,7 @@ with gr.Blocks() as demo:
                 label="Número de Días a Predecir"
             )
             predict_btn = gr.Button("Predecir")
-    
+        
     with gr.Column():
         plot_output = gr.Plot(label="Gráfico de Predicción")
         download_btn = gr.File(label="Descargar Predicciones")
@@ -190,7 +198,7 @@ with gr.Blocks() as demo:
         hist = stock.history(period="max")
         total_points = len(hist)
         # Actualizar el deslizador para reflejar el número total de puntos disponibles
-        return gr.update(maximum=total_points, value=min(1000, total_points))
+        return gr.Slider.update(maximum=total_points, value=min(1000, total_points))
     
     ticker.change(
         fn=update_train_data_points,
@@ -204,4 +212,4 @@ with gr.Blocks() as demo:
         outputs=[plot_output, download_btn]
     )
 
-demo.launch(debug=True)
+demo.launch()