saved model's training progress

Dockerfile

@@ -10,5 +10,4 @@ RUN pip install --no-cache-dir -r requirements.txt
 
 COPY . .
 
-#CMD [ "python3", "-m" , "predict_sales_api", "run", "--host=0.0.0.0"]
-CMD ["gunicorn", "-b", "0.0.0.0:7860", "main:app"]
+CMD ["gunicorn", "-b", "0.0.0.0:7860", "main:app"]

forecast_model.h5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:439fe48ffd89a6c30944d66e7310219c7ea4715ed2ee5d13799480ed19fb193d
+size 1535000

main.py

@@ -1,9 +1,8 @@
 import numpy as np
 import pandas as pd
 from flask import Flask, jsonify, request
+from keras.models import load_model
 from sklearn.preprocessing import MinMaxScaler
-from tensorflow import keras
-from tensorflow.keras import layers
 
 app = Flask(__name__)
 
@@ -45,8 +44,7 @@ def predict_sales():
     sales_data = sales_data.set_index('Date').reindex(full_date_range).fillna(0).reset_index()
     sales_data = sales_data.rename(columns={'index': 'Date'})
 
-    # sequence_len = 60
-    sequence_len = 3  # for testing with few datapoints
+    sequence_len = 60
     if len(sales_data) <= sequence_len:
         return jsonify({'code': 400, 'message': 'Insufficient data points for training after adding missing dates.'}), 400
 
@@ -55,26 +53,10 @@ def predict_sales():
 
     scaler = MinMaxScaler(feature_range=(0, 1))
     scaled_data = scaler.fit_transform(values.reshape(-1, 1))
-
-    x_train, y_train = [], []
-
-    for i in range(sequence_len, len(scaled_data)):
-        x_train.append(scaled_data[i - sequence_len:i, 0])
-        y_train.append(scaled_data[i, 0])
-
-    x_train, y_train = np.array(x_train), np.array(y_train)
-    x_train = np.reshape(x_train, (x_train.shape[0], x_train.shape[1], 1))
-
-    forecast_model = keras.Sequential()
-    forecast_model.add(layers.LSTM(100, return_sequences=True, input_shape=(x_train.shape[1], 1)))
-    forecast_model.add(layers.LSTM(100, return_sequences=False))
-    forecast_model.add(layers.Dense(25))
-    forecast_model.add(layers.Dense(1))
-    forecast_model.compile(optimizer='adam', loss='mean_squared_error')
-    forecast_model.fit(x_train, y_train, batch_size=1, epochs=3)
+    
+    forecast_model = load_model("forecast_model.h5")
 
     forecast_len = int(forecast_len_str)
-
     future_dates = pd.date_range(start=sales_data['Date'].iloc[-1] + pd.Timedelta(days=1), periods=forecast_len)
     predicted_sales = []
     recent_historical_sales = scaled_data[-sequence_len:].reshape(1, -1, 1)
@@ -92,5 +74,5 @@ def predict_sales():
     return jsonify({'code': 200, 'message': 'Forecasted sales successfully.', 'data': output_json}), 200
 
 
-#if __name__ == '__main__':
-#    app.run(debug=False, host='0.0.0.0', port=5005)
+# if __name__ == '__main__':
+#     app.run(debug=False, port=5005)