tf.truncated_normal

tf.truncated_normal(shape, mean=0.0, stddev=1.0, dtype=tf.float32, seed=None, name=None)

In [9]: a = tf.truncated_normal([2, 4])

In [10]: sess = tf.Session()

In [11]: sess.run(a)
Out[11]: 
array([[ 0.575279  ,  0.45625743, -0.62980247, -1.59388161],
       [-1.83954525, -1.44619739,  0.47894329, -0.96225327]], dtype=float32)