'''
Created on 2019年4月11日

@author: User
'''


from keras import layers,models,losses,optimizers
from BiddingKG.dl.common.Utils import *



def getTextCNNModel(vocab,embedding_weights,input_shape=(3,50,100),classes=3):
    input_left = layers.Input(shape=(input_shape[1],))
    input_center = layers.Input(shape=(input_shape[1],))
    input_right = layers.Input(shape=(input_shape[1],))
    
    list_kernel = [3,6]
    out_left = []
    out_center = []
    out_right = []
    
    embedding = layers.Embedding(len(vocab),input_shape[2],weights=[embedding_weights] if embedding_weights is not None else None,trainable=True,name="char_embeding")
    
    for kernel in list_kernel:
        out_left.append(layers.Conv1D(10, kernel, activation="relu",padding='same')(embedding(input_left)))
    concat_left = layers.merge(out_left,mode="concat")
    
    for kernel in list_kernel:
        out_center.append(layers.Conv1D(4,kernel,activation="relu",padding="same")(embedding(input_center)))
    concat_center = layers.merge(out_center,mode="concat")
        
    for kernel in list_kernel:
        out_right.append(layers.Conv1D(10,kernel,activation="relu",padding="same")(embedding(input_right)))
    concat_right = layers.merge(out_right,mode="concat")
    
    matrix_left = layers.Dense(12,activation="relu")(concat_left)
    matrix_center = layers.Dense(12,activation="relu")(concat_center)
    matrix_right = layers.Dense(12,activation="relu")(concat_right)
    #layers.average(inputs)
    concat_matrix = layers.merge([matrix_left,matrix_center,matrix_right],mode="ave")
    
    flattern = layers.Flatten()(concat_matrix)
    
    out = layers.Dense(classes,activation="softmax")(flattern)
    
    model = models.Model([input_left,input_center,input_right],out)
    
    model.compile(optimizer=optimizers.Adam(lr=0.001), loss=losses.categorical_crossentropy, metrics=[precision,recall,f1_score])
    model.summary()
    return model

import tensorflow as tf
if __name__=="__main__":
    with tf.Graph().as_default() as g:
        getTextCNNModel()
        for _vars in tf.global_variables():
            print(_vars.name,_vars)