我正在嘗試將我的 Keras 神經(jīng)網(wǎng)絡(luò)包裝在一個class對象中。我已經(jīng)在類設(shè)置之外實(shí)現(xiàn)了以下內(nèi)容，但我想讓它更加對象友好。總而言之，我的model調(diào)用函數(shù)sequential_model創(chuàng)建了一個sequential模型。在這compile一步中，我定義了自己的損失函數(shù)weighted_categorical_crossentropy，我希望順序模型實(shí)現(xiàn)它。但是，當(dāng)我運(yùn)行下面的代碼時，出現(xiàn)以下錯誤：ValueError: No gradients provided for any variable:我懷疑問題在于我如何定義該weighted_categorical_crossentropy函數(shù)以供sequential.再次，我能夠以非面向?qū)ο蟮姆绞酵瓿蛇@項(xiàng)工作。任何幫助都感激不盡。from tensorflow.keras import Sequential, backend as Kclass MyNetwork():             def __init__(self, file, n_output=4, n_hidden=20, epochs=3,                 dropout=0.10, batch_size=64, metrics = ['categorical_accuracy'],                 optimizer = 'rmsprop', activation = 'softmax'):    [...] //Other Class attributes     def model(self):        self.model = self.sequential_model(False)        self.model.summary()    def sequential_model(self, val):        K.clear_session()        if val == False:            self.epochs = 3        regressor = Sequential()        #regressor.run_eagerly = True        regressor.add(LSTM(units = self.n_hidden, dropout=self.dropout, return_sequences = True, input_shape = (self.X.shape[1], self.X.shape[2])))        regressor.add(LSTM(units = self.n_hidden, dropout=self.dropout, return_sequences = True))        regressor.add(Dense(units = self.n_output, activation=self.activation))            self.weights = np.array([0.025,0.225,0.78,0.020])        regressor.compile(optimizer = self.optimizer, loss = self.weighted_categorical_crossentropy(self.weights), metrics = [self.metrics])        regressor.fit(self.X, self.Y*1.0,batch_size=self.batch_size, epochs=self.epochs, verbose=1, validation_data=(self.Xval, self.Yval*1.0))        return regressor