import tensorflow as tfa = tf.Variable(tf.random_uniform(shape=[10, 10]))with tf.Session() as sess: #必须调用,否则a没办法初始化值 sess.run(tf.global_variables_initializer()) print sess.run(a)