ValueError: Output of generator should be a tuple `(x, y, sample_weight)` or `(x, y)`. Found: [[[[0.08627451 0.10980393 0.10980393]

I am a new learner in tensorflow, when I try to do the transfer learning. I meet an error of Value error. Does anyone know where the bug is? This code is related to the transfer learning of VGG16. Basically I just created my own MLP layers and do the fine-tuning

import os
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Flatten, Dense, Dropout, ZeroPadding2D
from tensorflow.keras.layers import Convolution2D, MaxPooling2D
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.optimizers import Adam
from tensorflow.keras import applications
import numpy as np
def get_length(Path, Pattern):
     # Pattern: name of the subdirectory
     Length = len(os.listdir(os.path.join(Path, Pattern)))
     return Length
train_data_dir = ''
validation_data_dir = ''
img_width, img_height = 224, 224
epochs = 150
batch_size = 8
LR = 0.00001
Len_C1_Train = get_length(train_data_dir,'AFF')
Len_C2_Train = get_length(train_data_dir,'NFF')
Len_C1_Val = get_length(validation_data_dir,'AFF')
Len_C2_Val = get_length(validation_data_dir,'NFF')
model = applications.VGG16(include_top=False, weights='imagenet')
datagen = ImageDataGenerator(rescale=1. / 255)
train_generator = datagen.flow_from_directory(train_data_dir,
                                              target_size=(img_width, img_height),
                                              batch_size=batch_size,
                                              class_mode=None,
                                              shuffle=False)  
# Extracting the features from the loaded images
features_train = model.predict_generator(train_generator,
                                         (Len_C1_Train+Len_C2_Train) // batch_size, 

max_queue_size=1)
val_generator = datagen.flow_from_directory(validation_data_dir,
                                            target_size=(img_width, img_height),
                                            batch_size=batch_size,
                                            class_mode=None,
                                            shuffle=False)  
# Extracting the features from the loaded images
features_val = model.predict_generator(val_generator,
                                      (Len_C1_Val+Len_C2_Val) // batch_size, max_queue_size=1)        
    
train_data = features_train
train_labels = np.array([0] * int(Len_C1_Train) + [1] * int(Len_C2_Train))
validation_data = features_val
validation_labels = np.array([0] * int(Len_C1_Val) + [1] * int(Len_C2_Val))
#  Building the MLP model
model2=Sequential()
model2.add(Dense(128, activation='relu'))
model2.add(Dropout(0.5))
model2.add(Dense(1, activation='softmax'))
model_total = Sequential([model, model2])
model.trainable = False
model_total.compile(loss='binary_crossentropy', optimizer = Adam(lr = LR), metrics=['binary_accuracy'])