assignment.py

import pandas as pd
import numpy as np

import os
import sys
stderr = sys.stderr
sys.stderr = open(os.devnull, 'w')
import tensorflow as tf
sys.stderr = stderr

os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'

from tensorflow.keras.models import Model, Sequential
from tensorflow.keras.layers import *
from tensorflow.keras.callbacks import EarlyStopping
from tensorflow.keras.optimizers import Adam
from tensorflow.keras import regularizers
from sklearn.model_selection import train_test_split
from sklearn.utils import shuffle
from sklearn import preprocessing
from sklearn.metrics import *

def get_model(x_train):
    """
    Return the requested Keras model

    :param x_train:      The NumPy array used for training (for dimension size extraction)


    :return:             The Keras model
    """
    
    # some parameters to control model
    dp_lvl = 0.2
    regularizer_lvl = 0.002
    
    strategy = tf.distribute.MirroredStrategy()
    with strategy.scope():
        # network design
        model = Sequential()
        model.add(LSTM(128, input_shape=(x_train.shape[1], x_train.shape[2]),dropout = dp_lvl,
                       recurrent_dropout = dp_lvl, return_sequences =  True ))
        model.add(LSTM(128, dropout = dp_lvl,recurrent_dropout = dp_lvl, return_sequences =  False ))
        model.add(Dense(256, activation='tanh',activity_regularizer=regularizers.l2(regularizer_lvl)))
        model.add(Dropout (0.2))
        model.add(Dense(128, activation='tanh',activity_regularizer=regularizers.l2(regularizer_lvl)))
        model.add(Dense(x_train.shape[2], activation='relu',activity_regularizer=regularizers.l2(regularizer_lvl)))
        
    return model