tfrecordmodel.py

with tf.Session() as sess:
    tf.keras.backend.set_session(sess)
    sess.run(ds_tr_init) # initialize the generator
  
    # Rewire network to tie it into the generator
    image_input = tf.keras.Input(tensor=x)

    # Model definition
    ...

    # Specify the model
    model = tf.keras.Model(inputs=image_input, outputs=[weather_output, ground_output])
    
    # Supply the outputs of the generator as target tensors
    model.compile(optimizer='adam', loss='binary_crossentropy', target_tensors=[w,g])
    model.fit(steps_per_epoch=len(df_train) // batch_size, verbose=1, epochs=1)