Paddle报错的解决方案（持续更新）

问题一

 self.output = paddle.layer.fc(
            input=[self.linear_part, self.interaction_part],
            size=1,
            act=paddle.activation.Sigmoid(),
            param_attr=paddle.attr.ParamAttr(
                name="%s_fc_output_linear_part.w" % self.network_name,
                is_static=is_subnet_param_static,
                l2_rate=self.l2_rate,
            ),
            bias_attr=paddle.attr.ParamAttr(
                name="%s_fc_output_linear_part.b" % self.network_name,
                is_static=is_subnet_param_static,
                l2_rate=self.l2_rate,
            )
        )

报错如下：

File "/python/lib/python2.7/site-packages/paddle/trainer/config_parser.py", line 4341, in my_fatal
raise Exception()

原因如下：
When the name field of param_attr is manually specified and the input is a list, the param_attr should also be a list with each item being the param_attr for each input item. If only one named param_attr is provided, all the input items would share this parameter.
但由于fc_layer的更新，使得不能默认的共享参数，导致报错。

解决方法：
若代码设定param_attr，而input是个list，param_attr也设置成list，即可解决。

参考：
https://github.com/PaddlePaddle/models/issues/512

问题二

报错如下：

File "/python