Created by: luotao1

增强维度类报错信息的示例：

涉及维度的所有报错（包括比较共多少维、比较其中一维等），都需要打出具体的维度信息，方便用户查看。

x=fluid.layers.data(name='x', append_batch_size = False, shape=[2, 5], dtype='float64')
y=fluid.layers.data(name='y', append_batch_size = False, shape=[5, 3], dtype='float64')
mul=fluid.layers.mul(x=x, y=y, x_num_col_dims=2)

• 修改前报错: 仅有一句话，没说X的维度是什么， x_num_col_dims是什么

The input tensor X's rank of MulOp should be larger than x_num_col_dims.

• 修改后报错：打出具体X共2维，shape是多少，x_num_col_dims是多少。

ShapeError: The input tensor X's dimensions of MulOp should be larger than x_num_col_dims. But 
received X's dimensions = 2, X's shape = [2, 5], x_num_col_dims = 2.

如果维度做了一些变化，需要打出原始维度和变化后维度的具体信息，方便用户查看。

x=fluid.layers.data(name='x', append_batch_size = False, shape=[2, 5, 4, 3], dtype='float64')
y=fluid.layers.data(name='y', append_batch_size = False, shape=[2, 4, 4, 3], dtype='float64')
mul=fluid.layers.mul(x=x, y=y, x_num_col_dims=2, y_num_col_dims=2)

• 修改前报错: 只说修改后第一个矩阵的宽对应补上第二个矩阵的高，宽是12，高是8。对用户来说，无法和输入的shape联系上。

First matrix's width must be equal with second matrix's height. 12, 8. 

• 修改后报错：打出做了flatten前后X和Y的shape变化，指出具体哪一维出错。

ShapeError: After flatten the input tensor X and Y to 2-D dimensions matrix X1 and Y1. The matrix
 X1's width must be equal with matrix Y1's height. But received X's shape = [2, 5, 4, 3], X1's shape 
= [10, 12], X1's width = 12, Y's shape = [2, 4, 4, 3], Y1's shape = [8, 12], Y1's height = 8.

Shape报错信息格式规范

ShapeError: 检查内容. But received 现状。

• 统一以ShapeError: 开头，先报出是类型错误。
• 其次打印检查的内容，即哪里失败了。
• 最后用But received开头，报出现有的类型信息。
• 尽量用简单的词汇，如dimensions比rank更容易懂。