回归后预测NaN

Question

X = train_df.iloc[:,:].values
X = np.reshape(X,(634442,1,134))

K.clear_session()
model=Sequential()
model.add(LSTM(units=5,activation='sigmoid',kernel_initializer='zeros',input_shape=(None,134)))
model.add(Dense(units=1))

from keras.callbacks import ModelCheckpoint, TensorBoard, EarlyStopping

model.compile(optimizer='adam', 
                loss='mean_squared_error', 
                metrics=['accuracy'])

checkpointer = ModelCheckpoint(filepath="model.h1",
                           verbose=0,
                           save_best_only=True)
earlystopper=EarlyStopping(monitor='val_loss',min_delta=0,patience=1,verbose=1,mode='auto')
tensorboard = TensorBoard(log_dir='./logs',
                      histogram_freq=0,
                      write_graph=True,
                      write_images=True)
hist=model.fit(X,target,
           epochs=5,
           batch_size=64,
           verbose=1,
           shuffle=True,
           validation_split=0.2,
          callbacks=[checkpointer, tensorboard]).history




X_test = test_df.iloc[:,:].values
X_test = np.reshape(X_test,(214200,1,134))
from keras.models import load_model
regressor = load_model('model.h1')
val = regressor.predict(X_test)

我不明白为什么我的模型总是预测“nan”值的列表，我尝试了所有我可以在网上找到的方法，比如将优化程序改为“Adam”，降低学习率，增加批量大小，减小训练数据的大小，但仍然没有结果。 培训部分-

^{pr2}$