error line 108 in RNNPP

[jjyock]

        wt = tf.cond(tf.less(tf.reshape(tf.abs(w_t),[]),epilson),lambda:tf.sign(wt)*epilson, lambda:wt) #put in wrong position before 这个对么，我运行这个tf.reshape(tf.abs(w_t),[]) 看到报错了，校友