How to save my trained seq2seq language translator model?

Question

how do i save my trained seq2seq model. i understand that the weights do not change in this model, and ive tried multiple functions i found online but none of them seem to work well. let me know if you need any info about the model

i tried checkpoints but it did not exactly work. it seems to have 'jumbled' the words.

here is a copy of the relevant part of my code:

def encode_sequences(tokenizer,length,lines):
    seq=tokenizer.texts_to_sequences(lines)
    seq=pad_sequences(seq,maxlen=length,padding='post')
    return seq
from sklearn.model_selection import train_test_split
train,test=train_test_split(fra_eng,test_size=0.2,random_state=12)
trainX= encode_sequences(fra_tokenizer,fra_length,train[:,1])
trainY= encode_sequences(eng_tokenizer,eng_length,train[:,0])

testX= encode_sequences(fra_tokenizer, fra_length,test[:,1])
testY=encode_sequences(eng_tokenizer,eng_length,test[:,0])
def struct_model(in_vocab,out_vocab,a,b,neuron):
    model = Sequential()
    model.add(Embedding(in_vocab,neuron,input_length=a,mask_zero=True))
    model.add(LSTM(neuron))
    model.add(RepeatVector(8))
    model.add(LSTM(neuron,return_sequences=True))
    model.add(Dense(out_vocab,activation='softmax'))
    return model
model=struct_model(fra_vocab_size,eng_vocab_size,fra_length,eng_length,512)
rms=optimizers.RMSprop(learning_rate=0.001)
model.compile(optimizer=rms,loss='sparse_categorical_crossentropy')
**checkpoint_path = "training_1/cp.h5"
checkpoint_dir = os.path.dirname(checkpoint_path)
cp_callback = ModelCheckpoint(filepath=checkpoint_path,
                              save_best_only=True,
                              save_weights_only=False,
                              verbose=1)**
history=model.fit(trainX , trainY.reshape(trainY.shape[0],trainY.shape[1],1),
            epochs=50,batch_size=500,validation_split=0.1,callbacks=[cp_callback])
**# After training is complete
from keras.models import load_model

# Load the saved model
model = load_model("training_1/cp.h5")**

# You can then make predictions with the loaded model or continue training
import numpy as np
pred_probabilities = model.predict(testX.reshape((testX.shape[0], testY.shape[1])))
pred = np.argmax(pred_probabilities, axis=-1)
def gettingword(p,tokenizer):
    for word,index in tokenizer.word_index.items():
        if index ==p:
            return word
preds_text = []
for i in pred:
    temp = []
    for j in range(len(i)):
        t = gettingword(i[j], eng_tokenizer)
        if j > 0:
            if t and (t != gettingword(i[j-1], eng_tokenizer)):
                temp.append(t)
        else:
            if t:
                temp.append(t)
    preds_text.append(' '.join(temp))
**output=pd.DataFrame({'french':test[:,1],'actual':test[:,0],'predicted' : preds_text})
output.sample(15)**