diff --git a/.tools/cache_dataset.py b/.tools/cache_dataset.py
index 0c552f84f635c8beb4538c08af153f76e5c5cff5..eb4951f9633b5358f71c6c18f826f2d28c5d5796 100755
--- a/.tools/cache_dataset.py
+++ b/.tools/cache_dataset.py
@@ -3,14 +3,10 @@ import paddle.v2.dataset as dataset
 import nltk
 
 #cifar
-dataset.common.download(dataset.cifar.CIFAR100_URL, 'cifar',
-                        dataset.cifar.CIFAR100_MD5)
 dataset.common.download(dataset.cifar.CIFAR100_URL, 'cifar',
                         dataset.cifar.CIFAR100_MD5)
 dataset.common.download(dataset.cifar.CIFAR10_URL, 'cifar',
                         dataset.cifar.CIFAR10_MD5)
-dataset.common.download(dataset.cifar.CIFAR10_URL, 'cifar',
-                        dataset.cifar.CIFAR10_MD5)
 
 # Cache conll05
 dataset.common.download(dataset.conll05.WORDDICT_URL, 'conll05st', \
@@ -40,3 +36,9 @@ dataset.common.download(dataset.uci_housing.URL, "uci_housing", \
 # Cache vmt14
 dataset.common.download(dataset.wmt14.URL_TRAIN, "wmt14",\
                         dataset.wmt14.MD5_TRAIN)
+
+#mnist
+dataset.common.download(dataset.mnist.TRAIN_IMAGE_URL, 'mnist', TRAIN_IMAGE_MD5)
+dataset.common.download(dataset.mnist.TRAIN_LABEL_URL, 'mnist', TRAIN_LABEL_MD5)
+dataset.common.download(dataset.mnist.TEST_IMAGE_URL, 'mnist', TEST_IMAGE_MD5)
+dataset.common.download(dataset.mnist.TEST_LABEL_URL, 'mnist', TEST_LABEL_MD5)