Non trainable Parameter and Padding

[lima21bd]

Total params: 4,840,828
Trainable params: 647,428
Non-trainable params: 4,193,400

I am working on your dataset. Would you like to explain why Non-trainable params amount is so high?

def padding(Sentences):
maxlen = 101
for sentence in Sentences:
char = sentence[1]
for x in char:
maxlen = max(maxlen,len(str(x)))
for i,sentence in enumerate(Sentences):
Sentences[i][1] = pad_sequences(Sentences[i][1],101,padding='post')
return Sentences

Moreover, please explain this portion of the code with an example.