Error when checking input: expected lstm_49_input to have shape (10, 1) but got array with shape (14, 1)

class LSTMForecast(object):
    def __init__(self):
        self.batch_size = 24
        self.num_lstm_layers = 2
        self.lstm_units = 10
        self.reset_states = False
        self.time_steps = 1
        self.look_back = 1
        self.tmodel = None
        self.pmodel = None
        self.num_epochs = 10
        self.lstm_model()

    def lstm_model(self):
        from keras.layers import Flatten
        self.tmodel = Sequential()
        self.pmodel = Sequential()

        if(self.num_lstm_layers == 1):
            self.tmodel.add(LSTM(self.lstm_units, batch_input_shape=(self.batch_size, self.look_back, no_of_medicines), stateful=True))
            self.pmodel.add(LSTM(self.lstm_units, input_shape=(self.look_back, no_of_medicines), stateful=False))
        elif(self.num_lstm_layers == 2):
            self.tmodel.add(LSTM(self.lstm_units, batch_input_shape=(self.batch_size, self.lstm_units, self.time_steps), stateful=True, return_sequences=True))
            self.tmodel.add(LSTM(self.lstm_units, batch_input_shape=(self.batch_size, self.lstm_units, self.time_steps), stateful=True))
            self.pmodel.add(LSTM(self.lstm_units, input_shape=(self.lstm_units, self.time_steps), stateful=False, return_sequences=True))
            self.pmodel.add(LSTM(self.lstm_units, input_shape=(self.lstm_units, self.time_steps), stateful=False))

        self.tmodel.add(Dense(1))
        self.tmodel.compile(loss='mean_squared_error', optimizer='adam')
        self.pmodel.add(Dense(1))
        self.pmodel.compile(loss='mean_squared_error', optimizer='adam')

        self.losses = np.zeros(self.num_epochs)
        self.train_scores = np.zeros(self.num_epochs)
        self.test_scores = np.zeros(self.num_epochs)

        print(self.tmodel.summary())
        print(self.tmodel.weights)

    def lstm_model_train(self, xtr, ytr):
        #Reshape input to be [samples, time steps, features]
        print('***',xtr.shape)
        xtr = xtr.drop(["PRODUCT_ID", "week_id"], axis=1)
        xtr = xtr.values
        print('*****',xtr.shape)
        xtr = np.reshape(xtr, (xtr.shape[0], xtr.shape[1], 1))
        print(xtr.shape)
        print(ytr.shape)

        if(self.reset_states==True):
            for i in range(self.num_epochs):
                self.tmodel.fit(xtr, ytr, epochs=1, batch_size=self.batch_size, verbose=2, shuffle=False)
                self.losses[i] = self.tmodel.history["loss"][0]

                trainPredict = self.tmodel.predict(xtr, batch_size=self.batch_size)
                self.tmodel.reset_states()
                print("\nEpoch="+str(i))
        else:
            self.tmodel.fit(xtr, ytr, epochs=self.num_epochs, batch_size=self.batch_size, verbose=2, shuffle=False)
            self.losses = self.tmodel.history["loss"]

        # We maintain two models tmodel and pmodel. pmodel is used for prediction with batch_size = 1. See this:
        # https://stackoverflow.com/questions/43702481/why-does-keras-lstm-batch-size-used-for-prediction-have-to-be-the-same-as-fittin/44228505#44228505
        self.tmodel.save_weights('DataSet/lstm_model.h5')
        self.pmodel.load_weights('DataSet/lstm_model.h5')

    def lstm_model_predict(self, xts):
        xts = xts.drop(["PRODUCT_ID", "week_id"], axis=1)
        xts = xts.values
        print('$',xts.shape[0])
        xts = np.reshape(xts, (xts.shape[0], xts.shape[1], 1))
        return self.pmodel.predict(xts)