修改模型嵌入层

tensnsorflow/train.py

@@ -98,19 +98,18 @@ def model_fn(features, labels, mode, params):
     #------bulid weights------
     Feat_Emb = tf.get_variable(name='embeddings', shape=[feature_size, embedding_size], initializer=tf.glorot_normal_initializer())
 
-    #------build feaure-------
-    #{U-A-X-C不需要特殊处理的特征}
+    feat_ids = features['ids']
 
-    ucity_id    = features['ucity_id']
-    clevel1_id = features['clevel1_id']
-    ccity_name = features['ccity_name']
-    device_type = features['device_type']
-    manufacturer = features['manufacturer']
-    channel = features['channel']
-    top = features['top']
-    level2_ids = features['level2_ids']
-    time = features['time']
-    stat_date = features['stat_date']
+    # ucity_id    = features['ucity_id']
+    # clevel1_id = features['clevel1_id']
+    # ccity_name = features['ccity_name']
+    # device_type = features['device_type']
+    # manufacturer = features['manufacturer']
+    # channel = features['channel']
+    # top = features['top']
+    # level2_ids = features['level2_ids']
+    # time = features['time']
+    # stat_date = features['stat_date']
 
 
     if FLAGS.task_type != "infer":
@@ -119,20 +118,9 @@ def model_fn(features, labels, mode, params):
 
     #------build f(x)------
     with tf.variable_scope("Shared-Embedding-layer"):
-        ucity_id = tf.nn.embedding_lookup(Feat_Emb, ucity_id)
-        clevel1_id = tf.nn.embedding_lookup(Feat_Emb, clevel1_id)
-        ccity_name = tf.nn.embedding_lookup(Feat_Emb, ccity_name)
-        device_type = tf.nn.embedding_lookup(Feat_Emb, device_type)
-        manufacturer = tf.nn.embedding_lookup(Feat_Emb, manufacturer)
-        channel = tf.nn.embedding_lookup(Feat_Emb, channel)
-        top = tf.nn.embedding_lookup(Feat_Emb, top)
-        level2_ids = tf.nn.embedding_lookup(Feat_Emb, level2_ids)
-        time = tf.nn.embedding_lookup(Feat_Emb, time)
-        stat_date = tf.nn.embedding_lookup(Feat_Emb, stat_date)
-
-
-        x_concat = tf.concat([ucity_id,clevel1_id,ccity_name,device_type,manufacturer,
-                              channel,top,level2_ids,time,stat_date],axis=1)    # None * (F * K)
+        embedding_id = tf.nn.embedding_lookup(Feat_Emb,feat_ids)
+
+        x_concat = tf.concat(tf.reshape(embedding_id,shape=[-1, common_dims],axis=1))  # None * (F * K)
 
     with tf.name_scope("CVR_Task"):
         if mode == tf.estimator.ModeKeys.TRAIN: