How to combine Cross Validation and Early Stopping Round in LSTM model?

Question

I would like to combine Cross Validation and Early Stopping Round in LSTM model in Python. I would like to make cross validation on 3 folds and stop learning my LSTM model if for 3 consecutive epchos there is no improvement in the mean AUC of the 3 folds from the cross-validation.

In this way, I would like to build a model with the highest possible mean AUC of the 3 folds from cross-validation, which has not improved for 3 epchos in a row.

My current code is:

import numpy as np
from keras.models import Sequential
from keras.layers import LSTM, Dense
from sklearn.model_selection import KFold
from sklearn.metrics import roc_auc_score
from keras.callbacks import EarlyStopping

model = Sequential()
model.add(LSTM(units=32, input_shape=(X_train.shape[1], X_train.shape[2]), dropout=0.3))
model.add(Dense(1, activation='sigmoid'))

model.compile(loss='binary_crossentropy', optimizer=Adam(learning_rate=0.01), metrics=['AUC'])

kfold = KFold(n_splits=3, shuffle=True)

early_stopping = EarlyStopping(monitor='val_auc', patience=3, mode='max')

for train_idx, val_idx in kfold.split(X_train):
    X_train_fold, X_val_fold = X_train[train_idx], X_train[val_idx]
    y_train_fold, y_val_fold = y_train[train_idx], y_train[val_idx]

    auc_scores = []

    for epoch in range(100):
        model.fit(X_train_fold, y_train_fold, epochs=1, validation_data=(X_val_fold, y_val_fold), callbacks=[early_stopping], verbose=0)

        y_pred = model.predict(X_val_fold)
        auc = roc_auc_score(y_val_fold, y_pred)
        auc_scores.append(auc)

        print("Epoch:", epoch+1, "Avg AUC from 3 folds:", np.mean(auc_scores))

    print("Model stopped learn after", early_stopping.stopped_epoch)

How can I modify my code to achieve that ?