Created by: hohdiy

具体是这样的：解析这个pool层时，新版本的parse_config，少了一个start = 0的字段的设置，这个字段的缺失会导致paddle train失败。

    Layer(
        name="pool",
        type="pool",
        inputs=Input("conv",
                     pool=Pool(pool_type="max-projection",
                               channels=16,
                               size_x=32,
                               start=0,
                               stride=8))

解析后的proto的debug string如下：（右侧是有问题的config，少了start = 0）

layers {                                                              |   39   layers {                                                      
   41     name: "pool"                                                |   40     name: "pool"                                                
   42     type: "pool"                                                |   41     type: "pool"                                                
   43     size: 1936                                                  |   42     size: 1936                                                  
   44     active_type: ""                                             |   43     active_type: ""                                             
   45     inputs {                                                    |   44     inputs {                                                    
   46       input_layer_name: "conv"                                  |   45       input_layer_name: "conv"                                  
   47       pool_conf {                                               |   46       pool_conf {                                               
   48         pool_type: "max-projection"                             |   47         pool_type: "max-projection"                             
   49         channels: 16                                            |   48         channels: 16                                            
   50         size_x: 32                                              |   49         size_x: 32                                              
   51         start: 0                                                |      ----------------------------------------------------------------
   52         stride: 8                                               |   50         stride: 8                                               
   53         output_x: 11                                            |   51         output_x: 11                                            
   54         img_size: 110                                           |   52         img_size: 110                                           
   55         padding: 0                                              |   53         padding: 0                                              
   56         size_y: 32                                              |   54         size_y: 32                                              
   57         stride_y: 8                                             |   55         stride_y: 8                                             
   58         output_y: 11                                            |   56         output_y: 11                                            
   59         img_size_y: 110                                         |   57         img_size_y: 110                                         
      ----------------------------------------------------------------|   58         padding_y: 0                                            
   60       }                                                         |   59       }                                                         
   61     }                                                           |   60     }