service model 优化

spark/featureEng.py

@@ -881,8 +881,7 @@ if __name__ == '__main__':
     # model columns
     print("model columns to redis...")
     model_columns = user_columns + item_columns
-    not_model_columns = [col for col in model_columns if not col.endswith("_number")]
-    featureColumnsToRedis(not_model_columns)
+    featureColumnsToRedis(model_columns)
 
     print("数据字典save...")
     print("dataVocab:", str(dataVocab.keys()))

test/model_test.py

@@ -197,7 +197,7 @@ def test(df_train):
     # pre_data = pre_data.replace("'",'"')
     print(pre_data)
     timestmp1 = int(round(time.time()*1000))
-    r = requests.post('http://10.27.248.223:80/v1/models/service:predict', data=pre_data)
+    r = requests.post('http://tensorserving:80/v1/models/service:predict', data=pre_data)
     print("测试样本数：{},测试耗时ms:{}".format(10,int(round(time.time()*1000))- timestmp1))
 
 if __name__ == '__main__':

train/train_service.py

@@ -83,10 +83,10 @@ def getTrainColumns(train_columns,data_vocab):
             emb_columns.append(col)
             inputs[feature] = tf.keras.layers.Input(name=feature, shape=(), dtype='string')
 
-        # elif feature.endswith("_number"):
-        #     col = tf.feature_column.numeric_column(feature)
-        #     number_columns.append(col)
-        #     inputs[feature] = tf.keras.layers.Input(name=feature, shape=(), dtype='float32')
+        elif feature.endswith("_number"):
+            col = tf.feature_column.numeric_column(feature)
+            number_columns.append(col)
+            inputs[feature] = tf.keras.layers.Input(name=feature, shape=(), dtype='float32')
 
     return emb_columns,number_columns,inputs