encoder.py

class Encoder(keras.Model):
    def __init__(self,vocab_size=10000,emb_dim=128,units=256,batch_size=64):
        super(Encoder,self).__init__()
        self.units = units
        self.batch = batch_size
        self.emb_layer = keras.layers.Embedding(vocab_size,emb_dim)
        self.lstm = keras.layers.LSTM(self.units,return_sequences=True,return_state=True)
        
    def call(self,x,states):
        emb=self.emb_layer(x)
        output,hidden,carry=self.lstm(emb,initial_state=states)
        return output,hidden,carry
    
    def init_hidden_state(self):
        return tf.zeros((self.batch,self.units)),tf.zeros((self.batch,self.units))