我是Tensorflow的新手，我希望我能在这里得到一些帮助，我试图创建一个获得四个输入和一个输出的神经网络，它在Tensorflow中看起来像一个简单的任务，但是我是仍然会收到此错误：
InvalidArgumentError：断言失败：[标签ID必须> = 0] [条件x> = 0不符合元素：] [x（dnn / head / ToFloat：0） =] [[0] [0] [0] ...] [[节点：dnn / head / assert_range / assert_non_negative / assert_less_equal / Assert / AssertGuard / Assert = Assert [T = [DT_STRING，DT_STRING，DT_STRING，DT_FLOAT]， summarize = 3，_device =＆＃34; / job：localhost / replica：0 / task：0 / device：CPU：0＆＃34;]（dnn / head / assert_range / assert_non_negative / assert_less_equal / Assert / AssertGuard / Assert / Switch ， DNN /头/ assert_range / assert_non_negative / assert_less_equal /断言/ AssertGuard /断言/ Data_0的， DNN /头/ assert_range / assert_non_negative / assert_less_equal /断言/ AssertGuard /断言/ _1， DNN /头/ assert_range / assert_non_negative / assert_less_equal /断言/ AssertGuard /断言/ _2， DNN /头/ assert_range / assert_non_negative / assert_less_equal /断言/ AssertGuard /断言/交换机_1）]]
这是我的代码，提前谢谢。
F1 = tf.feature_column.numeric_column('F1')
F2 = tf.feature_column.numeric_column('F2')
F3 =tf.feature_column.numeric_column('F3')
F4 = tf.feature_column.numeric_column('F4')
F5 =tf.feature_column.numeric_column('F5')
F6 = tf.feature_column.numeric_column('F6')
F7 = tf.feature_column.numeric_column('F7')
F8 = tf.feature_column.numeric_column('F8')
F9 = tf.feature_column.numeric_column('F9')
F10 = tf.feature_column.numeric_column('F10')
F11 = tf.feature_column.numeric_column('F11')
F12 = tf.feature_column.numeric_column('F12')
F13 = tf.feature_column.numeric_column('F13')
deep_columns = [F2, F6, F11,F5]
COLUMNS = ["F1", "F2", "F3",
"F4", "F5", "F6", "F7",
"F8", "F9", "F10", "F11",
"F12"]
df_train = pd.read_csv("C:/data_4.csv",
names=COLUMNS, sep = "\t",skiprows=1,skipinitialspace=True)
df_train = df_train.fillna(0)
Xtrain = df_train.drop("F7", axis=1)
y_train = pd.DataFrame(df_train.F7)
ytrain['F7'] = df_train.F7
ytrain = ytrain.astype(int)
estimator = tf.estimator.DNNClassifier(
feature_columns=deep_columns,
hidden_units=[7, 10, 20])
inp = tf.estimator.inputs.pandas_input_fn(Xtrain,ytrain,
batch_size= 128,
num_epochs=1,
shuffle=True,
queue_capacity=1000,
num_threads=1,
target_column='F7')
estimator.train(input_fn=inp,steps=200)