diff --git a/docs/images/qrcode/dinggroup1.png b/docs/images/qrcode/dinggroup1.png
index b758ea4d2..cf3fd59e8 100644
Binary files a/docs/images/qrcode/dinggroup1.png and b/docs/images/qrcode/dinggroup1.png differ
diff --git a/docs/images/qrcode/dinggroup2.png b/docs/images/qrcode/dinggroup2.png
index f8cfb15eb..bc5344367 100644
Binary files a/docs/images/qrcode/dinggroup2.png and b/docs/images/qrcode/dinggroup2.png differ
diff --git a/docs/source/feature/feature.rst b/docs/source/feature/feature.rst
index 8cc25af48..97b0cdbd1 100644
--- a/docs/source/feature/feature.rst
+++ b/docs/source/feature/feature.rst
@@ -295,6 +295,8 @@ Sequence类特征格式一般为“XX\|XX\|XX”，如用户行为序列特征
 -  embedding_dim: embedding的dimension
 -  hash_bucket_size: 同离散值特征
 -  sub_feature_type: 用于描述序列特征里子特征的类型，目前支持 IdFeature 和 RawFeature 两种形式，默认为 IdFeature
+-  pad_sequence_length: 序列补齐或截断的长度
+-  max_seq_len: 最大序列长度，超过该长度的序列会被截断；当配置了`pad_sequence_length`时，`max_seq_len`会被忽略
 -  NOTE：SequenceFeature一般用在DIN算法或者BST算法里面。
 
 在模型中可支持对序列特征使用Target Attention（DIN)，方法如下：
diff --git a/easy_rec/python/builders/loss_builder.py b/easy_rec/python/builders/loss_builder.py
index b9698ba5b..55f0836e8 100644
--- a/easy_rec/python/builders/loss_builder.py
+++ b/easy_rec/python/builders/loss_builder.py
@@ -240,8 +240,7 @@ def build_kd_loss(kds, prediction_dict, label_dict, feature_dict):
   loss_dict = {}
   for kd in kds:
     assert kd.pred_name in prediction_dict, 'invalid predict_name: %s available ones: %s' % (
-        kd.pred_name,
-        ','.join(prediction_dict.keys()))
+        kd.pred_name, ','.join(prediction_dict.keys()))
 
     loss_name = kd.loss_name
     if not loss_name:
diff --git a/easy_rec/python/feature_column/feature_column.py b/easy_rec/python/feature_column/feature_column.py
index 8701b55fc..f60a2a949 100644
--- a/easy_rec/python/feature_column/feature_column.py
+++ b/easy_rec/python/feature_column/feature_column.py
@@ -165,8 +165,11 @@ def _cmp_embed_config(a, b):
         config = self._share_embed_infos[embed_name]
         max_seq_len = config.max_seq_len if config.HasField(
             'max_seq_len') else -1
+        pad_sequence_length = config.pad_sequence_length if config.HasField(
+            'pad_sequence_length') else -1
         for fc in share_embed_fcs:
           fc.max_seq_length = max_seq_len
+          fc.pad_sequence_length = pad_sequence_length
         self._deep_share_embed_columns[embed_name] = share_embed_fcs
 
       # for handling wide share embedding columns
@@ -182,8 +185,11 @@ def _cmp_embed_config(a, b):
         config = self._share_embed_infos[embed_name]
         max_seq_len = config.max_seq_len if config.HasField(
             'max_seq_len') else -1
+        pad_sequence_length = config.pad_sequence_length if config.HasField(
+            'pad_sequence_length') else -1
         for fc in share_embed_fcs:
           fc.max_seq_length = max_seq_len
+          fc.pad_sequence_length = pad_sequence_length
         self._wide_share_embed_columns[embed_name] = share_embed_fcs
 
     for fc_name in self._deep_columns:
@@ -647,6 +653,8 @@ def _add_deep_embedding_column(self, fc, config):
           ev_params=ev_params)
       fc.max_seq_length = config.max_seq_len if config.HasField(
           'max_seq_len') else -1
+      fc.pad_sequence_length = config.pad_sequence_length if config.HasField(
+          'pad_sequence_length') else -1
 
     if config.feature_type != config.SequenceFeature:
       self._deep_columns[feature_name] = fc
diff --git a/easy_rec/python/layers/input_layer.py b/easy_rec/python/layers/input_layer.py
index 27bc9bdf4..f21cc4a20 100644
--- a/easy_rec/python/layers/input_layer.py
+++ b/easy_rec/python/layers/input_layer.py
@@ -180,7 +180,11 @@ def get_sequence_feature(self, features, group_name):
       with variable_scope.variable_scope('input_layer/' +
                                          fc.categorical_column.name):
         tmp_embedding, tmp_seq_len = fc._get_sequence_dense_tensor(builder)
-        if fc.max_seq_length > 0:
+        # If pad_sequence_length is set, pad or truncate to fixed length
+        if fc.pad_sequence_length > 0:
+          tmp_embedding, tmp_seq_len = shape_utils.pad_or_truncate_sequence(
+              tmp_embedding, tmp_seq_len, fc.pad_sequence_length)
+        elif fc.max_seq_length > 0:
           tmp_embedding, tmp_seq_len = shape_utils.truncate_sequence(
               tmp_embedding, tmp_seq_len, fc.max_seq_length)
         seq_features.append((tmp_embedding, tmp_seq_len))
diff --git a/easy_rec/python/protos/feature_config.proto b/easy_rec/python/protos/feature_config.proto
index 8c0c0c214..69a2f273c 100644
--- a/easy_rec/python/protos/feature_config.proto
+++ b/easy_rec/python/protos/feature_config.proto
@@ -80,6 +80,8 @@ message FeatureConfig {
     optional string seq_multi_sep = 101;
     // truncate sequence data to max_seq_len
     optional uint32 max_seq_len = 102;
+    // pad or truncate sequence to fixed length (enables padding for shorter sequences)
+    optional uint32 pad_sequence_length = 103;
 
     optional string vocab_file = 11;
     repeated string vocab_list = 12;
diff --git a/easy_rec/version.py b/easy_rec/version.py
index 80ce168e3..ce2f7d72a 100644
--- a/easy_rec/version.py
+++ b/easy_rec/version.py
@@ -1,4 +1,4 @@
 # -*- encoding:utf-8 -*-
 # Copyright (c) Alibaba, Inc. and its affiliates.
 
-__version__ = '0.8.7'
+__version__ = '0.8.8'