diff --git a/PaddleCV/video/metrics/youtube8m/eval_util.py b/PaddleCV/video/metrics/youtube8m/eval_util.py
index 67bc2e54d14df350ab659ebaedeec9dd9dd176bc..f7742236f1176073eae84fdc7c3a3a1a2e294fe0 100644
--- a/PaddleCV/video/metrics/youtube8m/eval_util.py
+++ b/PaddleCV/video/metrics/youtube8m/eval_util.py
@@ -37,7 +37,7 @@ def calculate_hit_at_one(predictions, actuals):
     float: The average hit at one across the entire batch.
   """
     top_prediction = numpy.argmax(predictions, 1)
-    hits = actuals[:, top_prediction]
+    hits = actuals[numpy.arange(actuals.shape[0]), top_prediction]
     return numpy.average(hits)
 
 
diff --git a/PaddleCV/video/models/bmn/bmn.py b/PaddleCV/video/models/bmn/bmn.py
index 39638dd5b3590922e9ee93f72132f0de2c7e7ee4..e81b6abf2616a28d7b7c86e6f44d5b9eb1157639 100644
--- a/PaddleCV/video/models/bmn/bmn.py
+++ b/PaddleCV/video/models/bmn/bmn.py
@@ -27,7 +27,15 @@ __all__ = ["BMN"]
 
 
 class BMN(ModelBase):
-    """BMN model"""
+    """BMN model from
+    `"BMN: Boundary-Matching Network for Temporal Action Proposal Generation" <https://arxiv.org/abs/1907.09702>`_
+    Args:
+        tscale (int): sequence length, default 100.
+        dscale (int): max duration length, default 100.
+        prop_boundary_ratio (float): ratio of expanded temporal region in proposal boundary, default 0.5. 
+        num_sample (int): number of samples betweent starting boundary and ending boundary of each propoasl, default 32.
+        num_sample_perbin (int):  number of selected points in each sample, default 3.
+    """
 
     def __init__(self, name, cfg, mode='train'):
         super(BMN, self).__init__(name, cfg, mode=mode)
diff --git a/PaddleCV/video/reader/ctcn_reader.py b/PaddleCV/video/reader/ctcn_reader.py
index 35ceca6e67bb754a3a480fd272a1a605043a70d6..7488cf988ac46aeb558b6b6b23322a41160f89d9 100644
--- a/PaddleCV/video/reader/ctcn_reader.py
+++ b/PaddleCV/video/reader/ctcn_reader.py
@@ -425,9 +425,9 @@ class CTCNReader(DataReader):
                     tmp_list = fl[i * file_num:]
                 reader_lists[i] = tmp_list
 
-            queue = multiprocessing.Queue(queue_size)
+            manager = multiprocessing.Manager()
+            queue = manager.Queue(queue_size)
             p_list = [None] * len(reader_lists)
-            # for reader_list in reader_lists:
             for i in range(len(reader_lists)):
                 reader_list = reader_lists[i]
                 p_list[i] = multiprocessing.Process(
diff --git a/dygraph/slowfast/README.md b/dygraph/slowfast/README.md
index 3a448d3cb104a2a7456f6a67980a9937d9941061..b7ab9d4bb9ce6604ee1d2eb97f1556825505291f 100644
--- a/dygraph/slowfast/README.md
+++ b/dygraph/slowfast/README.md
@@ -68,31 +68,8 @@ SlowFast Overview
 
 ## 数据准备
 
-SlowFast模型的训练数据采用Kinetics400数据集，直接输入mp4文件进行训练，数据准备方式如下:
+SlowFast模型的训练数据采用Kinetics400数据集，数据下载及准备请参考[数据说明](../PaddleCV/video/data/dataset/README.md)
 
-### mp4视频下载
-
-在Code\_Root目录下创建文件夹
-
-```
-    cd $Code_Root/ && mkdir data
-    cd data && mkdir data_k400 && cd data_k400
-    mkdir train_mp4 && mkdir val_mp4
-```
-
-ActivityNet官方提供了Kinetics的下载工具，具体参考其[官方repo](https://github.com/activitynet/ActivityNet/tree/master/Crawler/Kinetics)即可下载Kinetics400的mp4视频集合。将kinetics400的训练与验证集合分别下载到data/data\_k400/train\_mp4与data/data\_k400/val\_mp4。
-
-### 生成训练和验证集list
-
-验证(Validation)和评估(Evaluation)使用同一份数据集，预测(Predict)可以从验证数据集中挑选少量样本进行预测
-
-```
-    cd $Code_Root/data/
-    ls $Code_Root/data/data_k400/train_mp4/* > train.csv
-    ls $Code_Root/data/data_k400/val_mp4/* > val.csv
-    ls $Code_Root/data/data_k400/val_mp4/* > test.csv
-    ls $Code_Root/data/data_k400/val_mp4/* > infer.csv
-```
 
 ## 模型训练
 
diff --git a/dygraph/slowfast/train.py b/dygraph/slowfast/train.py
index 48dabcf4b4b1d4daad963785f2fc41373e54ae29..71391997519345ddb4ca364ae72e1c2a0af94f0e 100644
--- a/dygraph/slowfast/train.py
+++ b/dygraph/slowfast/train.py
@@ -268,7 +268,7 @@ def train(args):
 
     random.seed(0)
     np.random.seed(0)
-    paddle.framework.manual_seed(0)
+    paddle.framework.seed(0)
     with fluid.dygraph.guard(place):
         # 1. init net
         if args.use_data_parallel: