Merge pull request #3247 from Yancey1989/convert_all_dataset

convert dataset into recordio format

python/paddle/v2/dataset/cifar.py

@@ -133,7 +133,7 @@ def convert(path):
     """
     Converts dataset to recordio format
     """
-    paddle.v2.dataset.common.convert(path, train100(), 10, "cifar_train100")
-    paddle.v2.dataset.common.convert(path, test100(), 10, "cifar_test100")
-    paddle.v2.dataset.common.convert(path, train10(), 10, "cifar_train10")
-    paddle.v2.dataset.common.convert(path, test10(), 10, "cifar_test10")
+    paddle.v2.dataset.common.convert(path, train100(), 1000, "cifar_train100")
+    paddle.v2.dataset.common.convert(path, test100(), 1000, "cifar_test100")
+    paddle.v2.dataset.common.convert(path, train10(), 1000, "cifar_train10")
+    paddle.v2.dataset.common.convert(path, test10(), 1000, "cifar_test10")

python/paddle/v2/dataset/common.py

@@ -32,17 +32,22 @@ __all__ = [
 
 DATA_HOME = os.path.expanduser('~/.cache/paddle/dataset')
 
+
 # When running unit tests, there could be multiple processes that
 # trying to create DATA_HOME directory simultaneously, so we cannot
 # use a if condition to check for the existence of the directory;
 # instead, we use the filesystem as the synchronization mechanism by
 # catching returned errors.
-try:
-    os.makedirs(DATA_HOME)
-except OSError as exc:
-    if exc.errno != errno.EEXIST:
-        raise
-    pass
+def must_mkdirs(path):
+    try:
+        os.makedirs(DATA_HOME)
+    except OSError as exc:
+        if exc.errno != errno.EEXIST:
+            raise
+        pass
+
+
+must_mkdirs(DATA_HOME)
 
 
 def md5file(fname):
@@ -93,6 +98,19 @@ def fetch_all():
                 "fetch")()
 
 
+def fetch_all_recordio(path):
+    for module_name in filter(lambda x: not x.startswith("__"),
+                              dir(paddle.v2.dataset)):
+        if "convert" in dir(
+                importlib.import_module("paddle.v2.dataset.%s" % module_name)) and \
+                not module_name == "common":
+            ds_path = os.path.join(path, module_name)
+            must_mkdirs(ds_path)
+            getattr(
+                importlib.import_module("paddle.v2.dataset.%s" % module_name),
+                "convert")(ds_path)
+
+
 def split(reader, line_count, suffix="%05d.pickle", dumper=cPickle.dump):
     """
     you can call the function as:

python/paddle/v2/dataset/conll05.py

@@ -233,5 +233,5 @@ def convert(path):
     """
     Converts dataset to recordio format
     """
-    paddle.v2.dataset.common.convert(path, test(), 10, "conl105_train")
-    paddle.v2.dataset.common.convert(path, test(), 10, "conl105_test")
+    paddle.v2.dataset.common.convert(path, test(), 1000, "conl105_train")
+    paddle.v2.dataset.common.convert(path, test(), 1000, "conl105_test")

python/paddle/v2/dataset/imdb.py

@@ -173,5 +173,5 @@ def convert(path):
     Converts dataset to recordio format
     """
     w = word_dict()
-    paddle.v2.dataset.common.convert(path, lambda: train(w), 10, "imdb_train")
-    paddle.v2.dataset.common.convert(path, lambda: test(w), 10, "imdb_test")
+    paddle.v2.dataset.common.convert(path, lambda: train(w), 1000, "imdb_train")
+    paddle.v2.dataset.common.convert(path, lambda: test(w), 1000, "imdb_test")

python/paddle/v2/dataset/imikolov.py

@@ -155,6 +155,7 @@ def convert(path):
     N = 5
     word_dict = build_dict()
     paddle.v2.dataset.common.convert(path,
-                                     train(word_dict, N), 10, "imikolov_train")
+                                     train(word_dict, N), 1000,
+                                     "imikolov_train")
     paddle.v2.dataset.common.convert(path,
-                                     test(word_dict, N), 10, "imikolov_test")
+                                     test(word_dict, N), 1000, "imikolov_test")

python/paddle/v2/dataset/mnist.py

@@ -119,5 +119,5 @@ def convert(path):
     """
     Converts dataset to recordio format
     """
-    paddle.v2.dataset.common.convert(path, train(), 10, "minist_train")
-    paddle.v2.dataset.common.convert(path, test(), 10, "minist_test")
+    paddle.v2.dataset.common.convert(path, train(), 1000, "minist_train")
+    paddle.v2.dataset.common.convert(path, test(), 1000, "minist_test")

python/paddle/v2/dataset/movielens.py

@@ -254,8 +254,8 @@ def convert(path):
     """
     Converts dataset to recordio format
     """
-    paddle.v2.dataset.common.convert(path, train(), 10, "movielens_train")
-    paddle.v2.dataset.common.convert(path, test(), 10, "movielens_test")
+    paddle.v2.dataset.common.convert(path, train(), 1000, "movielens_train")
+    paddle.v2.dataset.common.convert(path, test(), 1000, "movielens_test")
 
 
 if __name__ == '__main__':

python/paddle/v2/dataset/sentiment.py

@@ -137,5 +137,5 @@ def convert(path):
     """
     Converts dataset to recordio format
     """
-    paddle.v2.dataset.common.convert(path, train, 10, "sentiment_train")
-    paddle.v2.dataset.common.convert(path, test, 10, "sentiment_test")
+    paddle.v2.dataset.common.convert(path, train, 1000, "sentiment_train")
+    paddle.v2.dataset.common.convert(path, test, 1000, "sentiment_test")

python/paddle/v2/dataset/uci_housing.py

@@ -119,5 +119,5 @@ def convert(path):
     """
     Converts dataset to recordio format
     """
-    paddle.v2.dataset.common.convert(path, train(), 10, "uci_housing_train")
-    paddle.v2.dataset.common.convert(path, test(), 10, "uci_houseing_test")
+    paddle.v2.dataset.common.convert(path, train(), 1000, "uci_housing_train")
+    paddle.v2.dataset.common.convert(path, test(), 1000, "uci_houseing_test")

python/paddle/v2/dataset/wmt14.py

@@ -169,5 +169,6 @@ def convert(path):
     Converts dataset to recordio format
     """
     dict_size = 30000
-    paddle.v2.dataset.common.convert(path, train(dict_size), 10, "wmt14_train")
-    paddle.v2.dataset.common.convert(path, test(dict_size), 10, "wmt14_test")
+    paddle.v2.dataset.common.convert(path,
+                                     train(dict_size), 1000, "wmt14_train")
+    paddle.v2.dataset.common.convert(path, test(dict_size), 1000, "wmt14_test")