diff --git a/develop/doc/_sources/design/ops/sequence_decoder.md.txt b/develop/doc/_sources/design/ops/sequence_decoder.md.txt
index 9db5fb8e9a9f89b004bf71ddc064cd976c0d0bee..c4a9bbeeefca0e05c335dd60233691e8bac33015 100644
--- a/develop/doc/_sources/design/ops/sequence_decoder.md.txt
+++ b/develop/doc/_sources/design/ops/sequence_decoder.md.txt
@@ -22,7 +22,7 @@ The current `LoDTensor` is designed to store levels of variable-length sequences
 The integers in each level represent the begin and end (not inclusive) offset of a sequence **in the underlying tensor**,
 let's call this format the **absolute-offset LoD** for clarity.
 
-The relative-offset LoD can retrieve any sequence very quickly but fails to represent empty sequences, for example, a two-level LoD is as follows
+The absolute-offset LoD can retrieve any sequence very quickly but fails to represent empty sequences, for example, a two-level LoD is as follows
 ```python
 [[0, 3, 9]
  [0, 2, 3, 3, 3, 9]]
@@ -119,7 +119,7 @@ def generate():
         encoder_ctx_expanded = pd.lod_expand(encoder_ctx, target_word)
         decoder_input = pd.fc(
             act=pd.activation.Linear(),
-            input=[target_word, encoder_ctx],
+            input=[target_word, encoder_ctx_expanded],
             size=3 * decoder_dim)
         gru_out, cur_mem = pd.gru_step(
             decoder_input, mem=decoder_mem, size=decoder_dim)
diff --git a/develop/doc/design/ops/sequence_decoder.html b/develop/doc/design/ops/sequence_decoder.html
index c3be3d4aea5bdd4bb7044f88b04ac8667da8a413..1cf28a854ea13f2e429203eda204352a9c87a42f 100644
--- a/develop/doc/design/ops/sequence_decoder.html
+++ b/develop/doc/design/ops/sequence_decoder.html
@@ -228,7 +228,7 @@ the selected candidate&#8217;s IDs in each time step can be stored in a <code cl
 <p>The current <code class="docutils literal"><span class="pre">LoDTensor</span></code> is designed to store levels of variable-length sequences. It stores several arrays of integers where each represents a level.</p>
 <p>The integers in each level represent the begin and end (not inclusive) offset of a sequence <strong>in the underlying tensor</strong>,
 let&#8217;s call this format the <strong>absolute-offset LoD</strong> for clarity.</p>
-<p>The relative-offset LoD can retrieve any sequence very quickly but fails to represent empty sequences, for example, a two-level LoD is as follows</p>
+<p>The absolute-offset LoD can retrieve any sequence very quickly but fails to represent empty sequences, for example, a two-level LoD is as follows</p>
 <div class="highlight-python"><div class="highlight"><pre><span></span><span class="p">[[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">9</span><span class="p">]</span>
  <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">9</span><span class="p">]]</span>
 </pre></div>
@@ -315,7 +315,7 @@ It is easy to find out the second sequence in the first-level LoD has two empty
         <span class="n">encoder_ctx_expanded</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">lod_expand</span><span class="p">(</span><span class="n">encoder_ctx</span><span class="p">,</span> <span class="n">target_word</span><span class="p">)</span>
         <span class="n">decoder_input</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">fc</span><span class="p">(</span>
             <span class="n">act</span><span class="o">=</span><span class="n">pd</span><span class="o">.</span><span class="n">activation</span><span class="o">.</span><span class="n">Linear</span><span class="p">(),</span>
-            <span class="nb">input</span><span class="o">=</span><span class="p">[</span><span class="n">target_word</span><span class="p">,</span> <span class="n">encoder_ctx</span><span class="p">],</span>
+            <span class="nb">input</span><span class="o">=</span><span class="p">[</span><span class="n">target_word</span><span class="p">,</span> <span class="n">encoder_ctx_expanded</span><span class="p">],</span>
             <span class="n">size</span><span class="o">=</span><span class="mi">3</span> <span class="o">*</span> <span class="n">decoder_dim</span><span class="p">)</span>
         <span class="n">gru_out</span><span class="p">,</span> <span class="n">cur_mem</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">gru_step</span><span class="p">(</span>
             <span class="n">decoder_input</span><span class="p">,</span> <span class="n">mem</span><span class="o">=</span><span class="n">decoder_mem</span><span class="p">,</span> <span class="n">size</span><span class="o">=</span><span class="n">decoder_dim</span><span class="p">)</span>
diff --git a/develop/doc_cn/_sources/design/ops/sequence_decoder.md.txt b/develop/doc_cn/_sources/design/ops/sequence_decoder.md.txt
index 9db5fb8e9a9f89b004bf71ddc064cd976c0d0bee..c4a9bbeeefca0e05c335dd60233691e8bac33015 100644
--- a/develop/doc_cn/_sources/design/ops/sequence_decoder.md.txt
+++ b/develop/doc_cn/_sources/design/ops/sequence_decoder.md.txt
@@ -22,7 +22,7 @@ The current `LoDTensor` is designed to store levels of variable-length sequences
 The integers in each level represent the begin and end (not inclusive) offset of a sequence **in the underlying tensor**,
 let's call this format the **absolute-offset LoD** for clarity.
 
-The relative-offset LoD can retrieve any sequence very quickly but fails to represent empty sequences, for example, a two-level LoD is as follows
+The absolute-offset LoD can retrieve any sequence very quickly but fails to represent empty sequences, for example, a two-level LoD is as follows
 ```python
 [[0, 3, 9]
  [0, 2, 3, 3, 3, 9]]
@@ -119,7 +119,7 @@ def generate():
         encoder_ctx_expanded = pd.lod_expand(encoder_ctx, target_word)
         decoder_input = pd.fc(
             act=pd.activation.Linear(),
-            input=[target_word, encoder_ctx],
+            input=[target_word, encoder_ctx_expanded],
             size=3 * decoder_dim)
         gru_out, cur_mem = pd.gru_step(
             decoder_input, mem=decoder_mem, size=decoder_dim)
diff --git a/develop/doc_cn/design/ops/sequence_decoder.html b/develop/doc_cn/design/ops/sequence_decoder.html
index 6a3419fb5ccf27951a70aae7bcd9402ae7c69c34..e516c928fc0623acc2b90b1205b67782096773db 100644
--- a/develop/doc_cn/design/ops/sequence_decoder.html
+++ b/develop/doc_cn/design/ops/sequence_decoder.html
@@ -247,7 +247,7 @@ the selected candidate&#8217;s IDs in each time step can be stored in a <code cl
 <p>The current <code class="docutils literal"><span class="pre">LoDTensor</span></code> is designed to store levels of variable-length sequences. It stores several arrays of integers where each represents a level.</p>
 <p>The integers in each level represent the begin and end (not inclusive) offset of a sequence <strong>in the underlying tensor</strong>,
 let&#8217;s call this format the <strong>absolute-offset LoD</strong> for clarity.</p>
-<p>The relative-offset LoD can retrieve any sequence very quickly but fails to represent empty sequences, for example, a two-level LoD is as follows</p>
+<p>The absolute-offset LoD can retrieve any sequence very quickly but fails to represent empty sequences, for example, a two-level LoD is as follows</p>
 <div class="highlight-python"><div class="highlight"><pre><span></span><span class="p">[[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">9</span><span class="p">]</span>
  <span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">3</span><span class="p">,</span> <span class="mi">9</span><span class="p">]]</span>
 </pre></div>
@@ -334,7 +334,7 @@ It is easy to find out the second sequence in the first-level LoD has two empty
         <span class="n">encoder_ctx_expanded</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">lod_expand</span><span class="p">(</span><span class="n">encoder_ctx</span><span class="p">,</span> <span class="n">target_word</span><span class="p">)</span>
         <span class="n">decoder_input</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">fc</span><span class="p">(</span>
             <span class="n">act</span><span class="o">=</span><span class="n">pd</span><span class="o">.</span><span class="n">activation</span><span class="o">.</span><span class="n">Linear</span><span class="p">(),</span>
-            <span class="nb">input</span><span class="o">=</span><span class="p">[</span><span class="n">target_word</span><span class="p">,</span> <span class="n">encoder_ctx</span><span class="p">],</span>
+            <span class="nb">input</span><span class="o">=</span><span class="p">[</span><span class="n">target_word</span><span class="p">,</span> <span class="n">encoder_ctx_expanded</span><span class="p">],</span>
             <span class="n">size</span><span class="o">=</span><span class="mi">3</span> <span class="o">*</span> <span class="n">decoder_dim</span><span class="p">)</span>
         <span class="n">gru_out</span><span class="p">,</span> <span class="n">cur_mem</span> <span class="o">=</span> <span class="n">pd</span><span class="o">.</span><span class="n">gru_step</span><span class="p">(</span>
             <span class="n">decoder_input</span><span class="p">,</span> <span class="n">mem</span><span class="o">=</span><span class="n">decoder_mem</span><span class="p">,</span> <span class="n">size</span><span class="o">=</span><span class="n">decoder_dim</span><span class="p">)</span>