TensorFlow执行一维卷积的问题

Question

以下代码说明生成的网络的形状为（3，），这不正确。有什么想法吗？

n = 100
m = 10
k = 1000
depth = 10
hidden_layers = 20

# ... fill variables with synthetic data ...

x = tf.placeholder(tf.float64,shape = (k,n,1))
y = tf.placeholder(tf.float64,shape = (k,4))

conv1_weights = tf.Variable(tf.truncated_normal([m,1,hidden_layers],stddev = 0.1,seed = 0,dtype = tf.float64))
conv1_biases = tf.Variable(tf.zeros([hidden_layers],tf.float64))

conv = tf.nn.conv1d(x,conv1_weights,stride = 1,padding = 'SAME')
relu = tf.nn.relu(tf.nn.bias_add(conv,conv1_biases))

print(tf.shape(relu))