diff --git a/bat/download-kifu.sh b/bat/download-kifu.sh
new file mode 100755
index 0000000..65fd979
--- /dev/null
+++ b/bat/download-kifu.sh
@@ -0,0 +1,17 @@
+# This shell script is supposed to run in a Datalab container to
+# download Shogi kifu from Floodgate.
+set -x
+mkdir -p ../kifu/zip 
+cd ../kifu/zip
+time wget -c --trust-server-names "https://osdn.net/frs/redir.php?m=jaist&f=shogi-server%2F68500%2Fwdoor2016.7z"
+cd ..
+which 7z || apt-get update -y && apt-get install -y p7zip-full --allow-unauthenticated
+[ -d 2016 ] || time 7z x zip/wdoor2016.7z -aos # -y
+
+cd ../python-dlshogi
+pip install python-shogi statistics tqdm
+pip install --no-cache-dir -e .
+time python utils/filter_csa.py ../kifu/2016/ # 20 minutes
+time python utils/make_kifu_list.py ../kifu/2016/ ../kifu/kifulist
+time python pydlshogi/read_kifu.py ../kifu/kifulist_test.txt
+time python pydlshogi/read_kifu.py ../kifu/kifulist_train.txt
diff --git a/pydlshogi/read_kifu.py b/pydlshogi/read_kifu.py
index 1e56b81..7c819f3 100644
--- a/pydlshogi/read_kifu.py
+++ b/pydlshogi/read_kifu.py
@@ -2,13 +2,44 @@
 import shogi.CSA
 import copy
 
+import argparse
+import logging
+import os
+import numpy as np
+import re
+from tqdm import tqdm
+
 from pydlshogi.features import *
 
+logging.basicConfig(format='%(asctime)s\t%(levelname)s\t%(message)s', 
+    datefmt='%Y/%m/%d %H:%M:%S', 
+    level=os.environ.get("LOGLEVEL", "DEBUG"))
+
+
+# 保存済みのダンプファイルを読み込む
+def load_dump(filename):
+    logging.info('Loading from %s' % (filename))
+    with open(filename, 'rb') as f:
+        positions = np.load(f)
+    return positions
+
+
+# ファイルに棋譜データをダンプする
+def save_dump(filename, positions):
+    logging.info('Saving to %s' % (filename))
+    np.array(positions).dump(filename)
+
+
 # read kifu
 def read_kifu(kifu_list_file):
+    logging.info('read kifu start')
+    dump_fname = re.sub(r'\.[^\.]+$', '', kifu_list_file) + '.pickle'
+    logging.info('dump_fname %s' % (dump_fname))
+    if os.path.exists(dump_fname): return load_dump(dump_fname)
+
     positions = []
     with open(kifu_list_file, 'r') as f:
-        for line in f.readlines():
+        for line in tqdm(f.readlines(), ncols=70):
             filepath = line.rstrip('\r\n')
             kifu = shogi.CSA.Parser.parse_file(filepath)[0]
             win_color = shogi.BLACK if kifu['win'] == 'b' else shogi.WHITE
@@ -31,4 +62,15 @@ def read_kifu(kifu_list_file):
 
                 positions.append((piece_bb, occupied, pieces_in_hand, move_label, win))
                 board.push_usi(move)
-    return positions
\ No newline at end of file
+
+    save_dump(dump_fname, positions)
+
+    logging.info('read kifu end')
+    return positions
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('kifulist', type=str, help='kifu list')
+    args = parser.parse_args()
+    read_kifu(args.kifulist)
\ No newline at end of file
diff --git a/train_policy.py b/train_policy.py
index bf4adfd..1563939 100644
--- a/train_policy.py
+++ b/train_policy.py
@@ -11,9 +11,6 @@
 
 import argparse
 import random
-import pickle
-import os
-import re
 
 import logging
 
@@ -48,36 +45,11 @@
     logging.info('Load optimizer state from {}'.format(args.resume))
     serializers.load_npz(args.resume, optimizer)
 
-logging.info('read kifu start')
-# 保存済みのpickleファイルがある場合、pickleファイルを読み込む
-# train date
-train_pickle_filename = re.sub(r'\..*?$', '', args.kifulist_train) + '.pickle'
-if os.path.exists(train_pickle_filename):
-    with open(train_pickle_filename, 'rb') as f:
-        positions_train = pickle.load(f)
-    logging.info('load train pickle')
-else:
-    positions_train = read_kifu(args.kifulist_train)
+# train data
+positions_train = read_kifu(args.kifulist_train)
 
 # test data
-test_pickle_filename = re.sub(r'\..*?$', '', args.kifulist_test) + '.pickle'
-if os.path.exists(test_pickle_filename):
-    with open(test_pickle_filename, 'rb') as f:
-        positions_test = pickle.load(f)
-    logging.info('load test pickle')
-else:
-    positions_test = read_kifu(args.kifulist_test)
-
-# 保存済みのpickleがない場合、pickleファイルを保存する
-if not os.path.exists(train_pickle_filename):
-    with open(train_pickle_filename, 'wb') as f:
-        pickle.dump(positions_train, f, pickle.HIGHEST_PROTOCOL)
-    logging.info('save train pickle')
-if not os.path.exists(test_pickle_filename):
-    with open(test_pickle_filename, 'wb') as f:
-        pickle.dump(positions_test, f, pickle.HIGHEST_PROTOCOL)
-    logging.info('save test pickle')
-logging.info('read kifu end')
+positions_test = read_kifu(args.kifulist_test)
 
 logging.info('train position num = {}'.format(len(positions_train)))
 logging.info('test position num = {}'.format(len(positions_test)))
diff --git a/utils/filter_csa.py b/utils/filter_csa.py
index b313d9a..692badc 100644
--- a/utils/filter_csa.py
+++ b/utils/filter_csa.py
@@ -3,6 +3,8 @@
 import re
 import statistics
 
+from tqdm import tqdm
+
 parser = argparse.ArgumentParser()
 parser.add_argument('dir', type=str)
 args = parser.parse_args()
@@ -16,7 +18,8 @@ def find_all_files(directory):
 
 kifu_count = 0
 rates = []
-for filepath in find_all_files(args.dir):
+n_kifu = len(list(find_all_files(args.dir)))
+for filepath in tqdm(find_all_files(args.dir), total=n_kifu, ncols=70):
     rate = {}
     move_len = 0
     toryo = False