一年连载半日完,打下省市三个半
归化小民愁白头,主任写到哪一段?

萧主任比较健康,连载永远不够看。为了自己动手丰衣足食,用现有文本训练了个RNN神经网络,吐点东西出来聊以自慰
更多信息请参考关于部分

第65535节 ...

Generate 生成

一切荣耀归元老院

About 关于《临高》生成器

使用的神经网络architecture见下:

Layer (type)                     Output Shape          Param #     Connected to                     
====================================================================================================
embedding_2 (Embedding)          (64, 48, 200)         1137600     embedding_input_2[0][0]          
____________________________________________________________________________________________________
batchnormalization_2 (BatchNormal(64, 48, 200)         400         embedding_2[0][0]                
____________________________________________________________________________________________________
lstm_3 (LSTM)                    (64, 48, 512)         1460224     batchnormalization_2[0][0]       
____________________________________________________________________________________________________
dropout_4 (Dropout)              (64, 48, 512)         0           lstm_3[0][0]                     
____________________________________________________________________________________________________
lstm_4 (LSTM)                    (64, 48, 512)         2099200     dropout_4[0][0]                  
____________________________________________________________________________________________________
dropout_5 (Dropout)              (64, 48, 512)         0           lstm_4[0][0]                     
____________________________________________________________________________________________________
timedistributed_3 (TimeDistribute(64, 48, 512)         262656      dropout_5[0][0]                  
____________________________________________________________________________________________________
dropout_6 (Dropout)              (64, 48, 512)         0           timedistributed_3[0][0]          
____________________________________________________________________________________________________
timedistributed_4 (TimeDistribute(64, 48, 5688)        2917944     dropout_6[0][0]                  
====================================================================================================
Total params: 7878024    
                    

训练于us-oregon AWS p2 instance,Tesla K80 GPU,Theano backend

一切荣耀归元老院