Why tf.matmul generate numerical error in Tensorflow?

Question

I am trying to make my model to accept batched input, but I found something strange. When I tried to following codes with tf.matmul(),

1. tf.matmul( vector1 (1 x 2