change test file

tensnsorflow/multi.py

@@ -71,11 +71,11 @@ def feature_engineer():
         .map(lambda x: (app_list_func(x[0], app_list_map),[value_map[x[2]],value_map[x[1]]], x[3],x[4]))
 
     spark.createDataFrame(test).toDF("level2_ids","ids","y","z")\
-        .repartition(1).write.format("tfrecords").save(path=path+"va/", mode="overwrite")
+        .repartition(1).write.format("tfrecords").option("recordType", "SequenceExample").save(path=path+"va/", mode="overwrite")
 
     print("va write done")
     spark.createDataFrame(train).toDF("level2_ids","ids","y","z") \
-        .repartition(1).write.format("tfrecords").save(path=path+"tr/", mode="overwrite")
+        .repartition(1).write.format("tfrecords").option("recordType", "SequenceExample").save(path=path+"tr/", mode="overwrite")
 
     print("done")
     rdd.unpersist()
@@ -85,7 +85,7 @@ def feature_engineer():
 
 def get_predict(date,value_map,app_list_map):
     sql = "select e.y,e.z,e.label,e.ucity_id,feat.level2_ids,e.device_id,e.cid_id from esmm_pre_data e " \
-          "left join diary_feat feat on e.cid_id = feat.diary_id"
+          "left join diary_feat feat on e.cid_id = feat.diary_id limit 50000"
 
     features = ["ucity_id"]
     df = spark.sql(sql)
@@ -103,7 +103,7 @@ def get_predict(date,value_map,app_list_map):
 
     spark.createDataFrame(rdd.filter(lambda x: x[4] == 0).map(lambda x: (x[0],x[5],x[6],x[7]))) \
         .toDF("level2_ids","y","z","ids").repartition(1).write.format("tfrecords") \
-        .save(path=path+"native/", mode="overwrite")
+            .option("recordType", "SequenceExample").save(path=path+"native/", mode="overwrite")
 
     nearby_pre = spark.createDataFrame(rdd.filter(lambda x: x[4] == 1).map(lambda x:(x[1],x[2],x[3]))) \
         .toDF("city", "uid", "cid_id")
@@ -111,7 +111,7 @@ def get_predict(date,value_map,app_list_map):
     nearby_pre.toPandas().to_csv(local_path+"nearby.csv", header=True)
     spark.createDataFrame(rdd.filter(lambda x: x[4] == 1).map(lambda x: (x[0], x[5], x[6], x[7]))) \
         .toDF("level2_ids","y","z","ids").repartition(1).write.format("tfrecords") \
-        .save(path=path+"nearby/", mode="overwrite")
+        .option("recordType", "SequenceExample").save(path=path+"nearby/", mode="overwrite")
 
     rdd.unpersist()