def lstm_basic_api (xtrain, xtest, ytrain, ytest, nodesx, acty_first, lrx, name_pkl):
    ''' LSTM implemented with attention and the original architecture'''
    dim1 = xtrain.shape[1]
    dim2 = xtrain.shape[2]
    batch_size =  32
    #Imput dimensions of the model
    inputs1 = Input(shape=(dim1, dim2))
    ##Create the mask
    model_input= Masking(name='mask')(inputs1)
    #LSTM bidirectional
    lstm_l = Bidirectional(LSTM(nodesx, activation=acty_first, return_sequences=True, name= 'lstm_1'))(model_input)
    ##Create attention layer
    att = Attention( name='attention_weight')(lstm_l)
    ##create output layer
    dense_end = Dense(units=1, activation='sigmoid') (att)
    ##create the model
    model = Model(inputs=inputs1, outputs=dense_end)
    print(model.summary())
    #plot_model(model, to_file='multilayer_perceptron_graph.png')
    model.compile(optimizer=Adam(lrx), loss = 'mse', metrics=['accuracy', mcc_metric])
    output_dir = 'task_add_two_numbers'
    if not os.path.exists(output_dir):
        os.makedirs(output_dir)
    ##check point
    #filepath="weights-improvement-{epoch:02d}-{val_accuracy:.2f}.hdf5"
    #checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True, mode='max')
    #callbacks_list = [checkpoint,history]
    max_epoch = 100
    filepath="weights-improvement-{epoch:02d}-{val_accuracy:.2f}.hdf5"
    checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True, mode='max')
    #checkpoint = ModelCheckpoint("best_model.hdf5", monitor='loss', verbose=1, save_best_only=True, mode='auto', period=1)
    class VisualiseAttentionMap(Callback):

        def on_epoch_end(self, epoch, logs=None):
            attention_map = get_activations(model, xtrain, layer_names='attention_weight')['attention_weight']

            # top is attention map.
            # bottom is ground truth.
            plt.imshow(attention_map, cmap='coolwarm')
            plt.colorbar()
            iteration_no = str(epoch).zfill(3)
            #plt.axis('off')
            plt.title(f'Iteration {iteration_no} / {max_epoch}')
            plt.xlabel('Windows')
            plt.ylabel('Sample')
            plt.savefig(f'{output_dir}/epoch_{iteration_no}.png')
            plt.close()
            plt.clf()
            np.save(f'epoch_{iteration_no}',attention_map)
    #modelh5 = load_model('test_model.h5',  custom_objects={'Attention': Attention, 'mcc_metric': mcc_metric})
    fit_outcome =  model.fit(xtrain, ytrain, batch_size=batch_size, epochs=100, shuffle=True, validation_data=(xtest, ytest), callbacks=[VisualiseAttentionMap(), checkpoint])
    model.save('lstm_model.h5') 
    evaluate_outcome = model.evaluate(xtest, ytest, batch_size=batch_size), 
    fit_history = fit_outcome.history
    y_pred = model.predict(xtest)
    y_pred_c = (y_pred > 0.5).astype(int)
    y_trainP = model.predict(xtrain)
    y_trainP_C =(y_trainP > 0.5).astype(int)
    ##Save the history to a file
    pkl_creation = create_pickle(name_pkl, fit_history)
    ##Make sure that the MCC was correctly calculated
    mcc_sk = matthews_corrcoef(ytest, y_pred_c)
    mcc_skT =  matthews_corrcoef(ytrain, y_trainP_C)
    #mcc_val= fit_history.get('val_accuracy')
    #mcc_last = (acc_val[-1])
    #print ("MCC value calculated:", mcc_last, 'MCC Sklearn:', mcc_sk)
    #print ("MCC value calculated:", mse_last, 'MSE Sklearn:', mse_sk)

    model.reset_states()
    return fit_history, evaluate_outcome, y_pred, y_pred_c, mcc_sk, pkl_creation, mcc_skT