Tensorflow--创建模型，自定义loss

1、定义一个简单结构的函数模型，并自定义损失函数

 1 from tensorflow.keras import Model, Input, layers, optimizers, losses
 2 
 3 
 4 x = Input((5,))
 5 y = Input((1,))
 6 
 7 dense1 = layers.Dense(20, 'sigmoid')(x)
 8 dense2 = layers.Dense(1)(dense1)
 9 model1 = Model(x, dense2)
10 # loss = layers.Lambda(losses.binary_crossentropy, output_shape=(1,),
11 #                      arguments={'from_logits': True})([y, dense2])
12 loss = losses.binary_crossentropy(y, model1.output)
13 
14 ####################  可以自定义损失函数
15 # model_loss = Lambda(yolo_loss, output_shape=(1,), name='yolo_loss',
16 #                     arguments={'anchors': anchors, 'num_classes': num_classes, 'ignore_thresh': 0.5})(
17 #     [*model_body.output, *y_true])
18 
19 model = Model([model1.input, y], loss)
20 
21 model.compile(
22     optimizer=optimizers.Adam(lr=1e-3),
23     loss='binary_crossentropy',
24     metrics=['accuracy']
25 )
26 
27 model.summary()

输出：

 1 Model: "model_1"
 2 __________________________________________________________________________________________________
 3 Layer (type)                    Output Shape         Param #     Connected to                     
 4 ==================================================================================================
 5 input_1 (InputLayer)            [(None, 5)]          0                                            
 6 __________________________________________________________________________________________________
 7 dense (Dense)                   (None, 20)           120         input_1[0][0]                    
 8 __________________________________________________________________________________________________
 9 dense_1 (Dense)                 (None, 1)            21          dense[0][0]                      
10 __________________________________________________________________________________________________
11 tf_op_layer_clip_by_value_2/Min [(None, 1)]          0           dense_1[0][0]                    
12 __________________________________________________________________________________________________
13 tf_op_layer_clip_by_value_2 (Te [(None, 1)]          0           tf_op_layer_clip_by_value_2/Minim
14 __________________________________________________________________________________________________
15 tf_op_layer_sub_5 (TensorFlowOp [(None, 1)]          0           tf_op_layer_clip_by_value_2[0][0]
16 __________________________________________________________________________________________________
17 input_2 (InputLayer)            [(None, 1)]          0                                            
18 __________________________________________________________________________________________________
19 tf_op_layer_add_6 (TensorFlowOp [(None, 1)]          0           tf_op_layer_clip_by_value_2[0][0]
20 __________________________________________________________________________________________________
21 tf_op_layer_add_7 (TensorFlowOp [(None, 1)]          0           tf_op_layer_sub_5[0][0]          
22 __________________________________________________________________________________________________
23 tf_op_layer_Log_4 (TensorFlowOp [(None, 1)]          0           tf_op_layer_add_6[0][0]          
24 __________________________________________________________________________________________________
25 tf_op_layer_sub_4 (TensorFlowOp [(None, 1)]          0           input_2[0][0]                    
26 __________________________________________________________________________________________________
27 tf_op_layer_Log_5 (TensorFlowOp [(None, 1)]          0           tf_op_layer_add_7[0][0]          
28 __________________________________________________________________________________________________
29 tf_op_layer_mul_4 (TensorFlowOp [(None, 1)]          0           input_2[0][0]                    
30                                                                  tf_op_layer_Log_4[0][0]          
31 __________________________________________________________________________________________________
32 tf_op_layer_mul_5 (TensorFlowOp [(None, 1)]          0           tf_op_layer_sub_4[0][0]          
33                                                                  tf_op_layer_Log_5[0][0]          
34 __________________________________________________________________________________________________
35 tf_op_layer_add_8 (TensorFlowOp [(None, 1)]          0           tf_op_layer_mul_4[0][0]          
36                                                                  tf_op_layer_mul_5[0][0]          
37 __________________________________________________________________________________________________
38 tf_op_layer_Neg_2 (TensorFlowOp [(None, 1)]          0           tf_op_layer_add_8[0][0]          
39 __________________________________________________________________________________________________
40 tf_op_layer_Mean_2 (TensorFlowO [(None,)]            0           tf_op_layer_Neg_2[0][0]          
41 ==================================================================================================
42 Total params: 141
43 Trainable params: 141
44 Non-trainable params: 0
45 __________________________________________________________________________________________________
posted @ 2019-12-27 15:15  巴蜀秀才  阅读(1172)  评论(0)  编辑  收藏  举报