在noneager模式下:
def gradients:
U_pred=tf.gradients(u,x)
return U_pred #returns 100x200 list (expected behaviour)
在TF2急切模式下:
def gradients:
with tf.GradientTape() as g:
g.watch(x)
U_pred=g.gradient(u,x)
return U_pred #return 100X1 list (expected 100x200)
如何在急切模式下获得100x200的列表?有什么办法吗?