Merge pull request #1028 from reyoung/feature/fix_embedding

Check md5 when download word embedding.

Author: luotao1

demo/model_zoo/embedding/pre_DictAndModel.sh

@@ -14,9 +14,20 @@
 # limitations under the License.
 set -e
 set -x
+BASE_URL='http://paddlepaddle.cdn.bcebos.com/model_zoo/embedding'
 
-# download the dictionary and pretrained model 
-for file in baidu.dict model_32.emb model_64.emb model_128.emb model_256.emb
-do 
-  wget http://paddlepaddle.bj.bcebos.com/model_zoo/embedding/$file
+wget ${BASE_URL}/baidu.dict
+
+DOWNLOAD_ITEMS=(model_32.emb model_64.emb model_128.emb model_256.emb)
+ITEM_MD5=(f88c8325ee6da6187f1080e8fe66c1cd
+          927cf70f27f860aff1a5703ebf7f1584
+	  a52e43655cd25d279777ed509a1ae27b
+	  b92c67fe9ff70fea53596080e351ac80)
+
+for ((i=0; i<${#ITEM_MD5[@]}; i++))
+do
+  FILENAME=${DOWNLOAD_ITEMS[${i}]}
+  REAL_MD5=`wget ${BASE_URL}/${FILENAME} -O - | tee ${FILENAME} | md5sum | cut -d ' ' -f 1`
+  EXPECTED_MD5=${ITEM_MD5[${i}]}
+  [ "${EXPECTED_MD5}" = "${REAL_MD5}" ]
 done