add inmemory dataset

core/trainers/framework/dataset.py

@@ -26,6 +26,7 @@ from paddle.fluid.contrib.utils.hdfs_utils import HDFSClient
 
 __all__ = ["DatasetBase", "DataLoader", "QueueDataset", "InMemoryDataset"]
 
+
 class DatasetBase(object):
     """R
     """
@@ -152,6 +153,7 @@ class QueueDataset(DatasetBase):
                 break
         return dataset
 
+
 class InMemoryDataset(QueueDataset):
     def _get_dataset(self, dataset_name, context):
         with open("context.txt", "w+") as fout:
@@ -197,7 +199,10 @@ class InMemoryDataset(QueueDataset):
             "hadoop.job.ugi": hdfs_ugi
         }
         hdfs_client = HDFSClient(hadoop_home, hdfs_configs)
-        file_list = ["{}/{}".format(hdfs_addr, x) for x in hdfs_client.lsr(train_data_path)]
+        file_list = [
+            "{}/{}".format(hdfs_addr, x)
+            for x in hdfs_client.lsr(train_data_path)
+        ]
         if context["engine"] == EngineMode.LOCAL_CLUSTER:
             file_list = split_files(file_list, context["fleet"].worker_index(),
                                     context["fleet"].worker_num())