I found that the problem was because in

`data_generator(train_descriptions, train_features, wordtoix, max_length, number_pics_per_bath)` the train_features was blank because no images assigned.

I tried to check where they were from, and found out the train_img from this part since the embedding took nearly no time.

# Call the funtion to encode all the train images

# This will take a while on CPU - Execute this only once

start = time()

encoding_train = {}

for img in train_img:

encoding_train[img[len(images):]] = encode(img)

print("Time taken in seconds =", time()-start)

this was the problem and fixed by fixing the directory

# Below path contains all the images

images = 'Flickr_Data/Images'

# Create a list of all image names in the directory

img = glob.glob(images + '*.jpg')

to

# Below path contains all the images

images = 'Flickr_Data/Images/'

# Create a list of all image names in the directory

img = glob.glob(images + '*.jpg')