diff --git a/docs/zh/RELEASENOTE/RELEASENOTE.md b/docs/zh/RELEASENOTE/RELEASENOTE.md
index 7b396e98350823f608ed9f46c32b2b0d12524882..70e83fcf4533bbfdcfb71bbb7577b7667c53f9b3 100644
--- a/docs/zh/RELEASENOTE/RELEASENOTE.md
+++ b/docs/zh/RELEASENOTE/RELEASENOTE.md
@@ -1,22 +1,32 @@
 # FrameworkPTAdapter 2.0.3 版本说明书
--   [用户须知](#用户须知md)
--   [新增特性](#新增特性md)
--   [特性修改](#特性修改md)
--   [已修复问题](#已修复问题md)
--   [已知问题](#已知问题md)
--   [兼容性](#兼容性md)
-<h2 id='用户须知md'>用户须知</h2>
+-   [FrameworkPTAdapter 2.0.3](#FrameworkPTAdapter-2-0-3md)
+    -   [用户须知](#用户须知md)
+    -   [新增特性](#新增特性md)
+    -   [特性修改](#特性修改md)
+    -   [已修复问题](#已修复问题md)
+    -   [已知问题](#已知问题md)
+    -   [兼容性](#兼容性md)
+-   [FrameworkPTAdapter 2.0.2](#FrameworkPTAdapter-2-0-2md)
+    -   [用户须知](#用户须知-0md)
+    -   [新增特性](#新增特性-1md)
+    -   [特性修改](#特性修改-2md)
+    -   [已修复问题](#已修复问题-3md)
+    -   [已知问题](#已知问题-4md)
+    -   [兼容性](#兼容性-5md)
+<h2 id="FrameworkPTAdapter-2-0-3md">FrameworkPTAdapter 2.0.3</h2>
+
+
+<h3 id="用户须知md">用户须知</h3>
 
 本框架基于Facebook主导的开源PyTorch1.5.0进行修改，延续原生的PyTorch特性，使用NPU进行动态图训练；以算子粒度进行模型适配，代码重用性好，支持现有的网络只修改设备类型或数据类型，即可迁移到NPU上使用。
 
 从此版本开始，PyTorch1.8.1版本提供支持，此版本延续PyTorch1.5.0特性，功能保持一致（profiling工具除外）。除此之外，对后端算子适配提供较好开发体验。当期1.8.1版本仅支持Resent50网络模型。
 
-<h2 id='新增特性md'>新增特性</h2>
+<h3 id="新增特性md">新增特性</h3>
 
 **表 1** PyTorch支持的版本特性列表
 
 <a name="t76c34275cbb74753970f7c5a9eb594fa"></a>
-
 <table><thead align="left"><tr id="r0c10e7163bf54fe8816ab5ca2d77ccc4"><th class="cellrowborder" valign="top" width="10.459999999999999%" id="mcps1.2.4.1.1"><p id="a7888762cf8294977b7d114b1c898d1bd"><a name="a7888762cf8294977b7d114b1c898d1bd"></a><a name="a7888762cf8294977b7d114b1c898d1bd"></a>一级特性</p>
 </th>
 <th class="cellrowborder" valign="top" width="26.27%" id="mcps1.2.4.1.2"><p id="a4581ffde4a5f455faadfba144243a9d4"><a name="a4581ffde4a5f455faadfba144243a9d4"></a><a name="a4581ffde4a5f455faadfba144243a9d4"></a>二级特性</p>
@@ -187,15 +197,15 @@
 </tbody>
 </table>
 
-<h2 id="特性修改md">特性修改</h2>
+<h3 id="特性修改md">特性修改</h3>
 
 不涉及
 
-<h2 id="已修复问题md">已修复问题</h2>
+<h3 id="已修复问题md">已修复问题</h3>
 
 不涉及
 
-<h2 id="已知问题md">已知问题</h2>
+<h3 id="已知问题md">已知问题</h3>
 
 <a name="table1969972073016"></a>
 <table><thead align="left"><tr id="row3699162017307"><th class="cellrowborder" valign="top" width="18.22%" id="mcps1.1.3.1.1"><p id="p16992020153010"><a name="p16992020153010"></a><a name="p16992020153010"></a>已知问题</p>
@@ -234,9 +244,144 @@
 </tbody>
 </table>
 
-<h2 id="兼容性md">兼容性</h2>
+<h3 id="兼容性md">兼容性</h3>
 
 A800-9010：CentOS 7.6/Ubuntu 18.04/BC-Linux 7.6/Debian 9.9/Debian 10/OpenEuler 20.03 LTS
 
 A800-9000：CentOS 7.6/Euler 2.8/Kylin v10/BC-Linux 7.6/OpenEuler 20.03 LTS/UOS 20 1020e
 
+<h2 id="FrameworkPTAdapter-2-0-2md">FrameworkPTAdapter 2.0.2</h2>
+
+
+<h3 id="用户须知-0md">用户须知</h3>
+
+本框架基于Facebook主导的开源PyTorch1.5.0进行修改，延续原生的PyTorch特性，使用NPU进行动态图训练；以算子粒度进行模型适配，代码重用性好，支持现有的网络只修改设备类型或数据类型，即可迁移到NPU上使用。
+
+<h3 id="新增特性-1md">新增特性</h3>
+
+**表 1** PyTorch支持的版本特性列表
+
+<a name="t76c34275cbb74753970f7c5a9eb594fa"></a>
+<table><thead align="left"><tr id="r0c10e7163bf54fe8816ab5ca2d77ccc4"><th class="cellrowborder" valign="top" width="10.489999999999998%" id="mcps1.2.4.1.1"><p id="a7888762cf8294977b7d114b1c898d1bd"><a name="a7888762cf8294977b7d114b1c898d1bd"></a><a name="a7888762cf8294977b7d114b1c898d1bd"></a>一级特性</p>
+</th>
+<th class="cellrowborder" valign="top" width="26.3%" id="mcps1.2.4.1.2"><p id="a4581ffde4a5f455faadfba144243a9d4"><a name="a4581ffde4a5f455faadfba144243a9d4"></a><a name="a4581ffde4a5f455faadfba144243a9d4"></a>二级特性</p>
+</th>
+<th class="cellrowborder" valign="top" width="63.21%" id="mcps1.2.4.1.3"><p id="a2a1562364b09433a83133fa10b3cf2b3"><a name="a2a1562364b09433a83133fa10b3cf2b3"></a><a name="a2a1562364b09433a83133fa10b3cf2b3"></a>说明</p>
+</th>
+</tr>
+</thead>
+<tbody><tr id="row171322953713"><td class="cellrowborder" rowspan="7" valign="top" width="10.489999999999998%" headers="mcps1.2.4.1.1 "><p id="p10237517181"><a name="p10237517181"></a><a name="p10237517181"></a>适配训练模型</p>
+</td>
+<td class="cellrowborder" valign="top" width="26.3%" headers="mcps1.2.4.1.2 "><p id="p6134294377"><a name="p6134294377"></a><a name="p6134294377"></a>YOLOV4</p>
+</td>
+<td class="cellrowborder" valign="top" width="63.21%" headers="mcps1.2.4.1.3 "><p id="p314533811397"><a name="p314533811397"></a><a name="p314533811397"></a>-</p>
+</td>
+</tr>
+<tr id="row15990182233714"><td class="cellrowborder" valign="top" headers="mcps1.2.4.1.1 "><p id="p399019224374"><a name="p399019224374"></a><a name="p399019224374"></a>YOLOV3</p>
+</td>
+<td class="cellrowborder" valign="top" headers="mcps1.2.4.1.2 "><p id="p1714553816396"><a name="p1714553816396"></a><a name="p1714553816396"></a>-</p>
+</td>
+</tr>
+<tr id="row5301255373"><td class="cellrowborder" valign="top" headers="mcps1.2.4.1.1 "><p id="p1530325123719"><a name="p1530325123719"></a><a name="p1530325123719"></a>DB</p>
+</td>
+<td class="cellrowborder" valign="top" headers="mcps1.2.4.1.2 "><p id="p5146193819390"><a name="p5146193819390"></a><a name="p5146193819390"></a>-</p>
+</td>
+</tr>
+<tr id="row274203413712"><td class="cellrowborder" valign="top" headers="mcps1.2.4.1.1 "><p id="p197423343379"><a name="p197423343379"></a><a name="p197423343379"></a>RFCN</p>
+</td>
+<td class="cellrowborder" valign="top" headers="mcps1.2.4.1.2 "><p id="p15828941153919"><a name="p15828941153919"></a><a name="p15828941153919"></a>-</p>
+</td>
+</tr>
+<tr id="row78671918163714"><td class="cellrowborder" valign="top" headers="mcps1.2.4.1.1 "><p id="p188671318183713"><a name="p188671318183713"></a><a name="p188671318183713"></a>CRNN</p>
+</td>
+<td class="cellrowborder" valign="top" headers="mcps1.2.4.1.2 "><p id="p10828241123917"><a name="p10828241123917"></a><a name="p10828241123917"></a>-</p>
+</td>
+</tr>
+<tr id="row16912221171"><td class="cellrowborder" valign="top" headers="mcps1.2.4.1.1 "><p id="p6913132676"><a name="p6913132676"></a><a name="p6913132676"></a>Densenset161</p>
+</td>
+<td class="cellrowborder" valign="top" headers="mcps1.2.4.1.2 "><p id="p209131221474"><a name="p209131221474"></a><a name="p209131221474"></a>-</p>
+</td>
+</tr>
+<tr id="row1016073314719"><td class="cellrowborder" valign="top" headers="mcps1.2.4.1.1 "><p id="p3160153320717"><a name="p3160153320717"></a><a name="p3160153320717"></a>Densenset191</p>
+</td>
+<td class="cellrowborder" valign="top" headers="mcps1.2.4.1.2 "><p id="p1516012331173"><a name="p1516012331173"></a><a name="p1516012331173"></a>-</p>
+</td>
+</tr>
+<tr id="row9627165910386"><td class="cellrowborder" rowspan="4" valign="top" width="10.489999999999998%" headers="mcps1.2.4.1.1 "><p id="p1561535993811"><a name="p1561535993811"></a><a name="p1561535993811"></a>适配NPU的PyTorch特性</p>
+</td>
+<td class="cellrowborder" valign="top" width="26.3%" headers="mcps1.2.4.1.2 "><p id="p136151659163819"><a name="p136151659163819"></a><a name="p136151659163819"></a>框架基础功能</p>
+</td>
+<td class="cellrowborder" valign="top" width="63.21%" headers="mcps1.2.4.1.3 "><p id="p1661535983813"><a name="p1661535983813"></a><a name="p1661535983813"></a>新增适配算子开发（详见算子清单）。</p>
+</td>
+</tr>
+<tr id="row7627155917380"><td class="cellrowborder" valign="top" headers="mcps1.2.4.1.1 "><p id="p1561525916383"><a name="p1561525916383"></a><a name="p1561525916383"></a>精度对比工具</p>
+</td>
+<td class="cellrowborder" valign="top" headers="mcps1.2.4.1.2 "><p id="p19616115953816"><a name="p19616115953816"></a><a name="p19616115953816"></a>新增精度对比工具，支持训练精度定界。</p>
+</td>
+</tr>
+<tr id="row46269593383"><td class="cellrowborder" valign="top" headers="mcps1.2.4.1.1 "><p id="p0616559163819"><a name="p0616559163819"></a><a name="p0616559163819"></a>昇腾710芯片</p>
+</td>
+<td class="cellrowborder" valign="top" headers="mcps1.2.4.1.2 "><p id="p186161459193813"><a name="p186161459193813"></a><a name="p186161459193813"></a>新增支持昇腾710芯片在线推理。</p>
+</td>
+</tr>
+<tr id="row76261059153817"><td class="cellrowborder" valign="top" headers="mcps1.2.4.1.1 "><p id="p11616175953812"><a name="p11616175953812"></a><a name="p11616175953812"></a>OS兼容性</p>
+</td>
+<td class="cellrowborder" valign="top" headers="mcps1.2.4.1.2 "><p id="p8935115971714"><a name="p8935115971714"></a><a name="p8935115971714"></a>新增支持ubuntu 18.04.5、OpenEuler 20.03 LTS系统</p>
+</td>
+</tr>
+</tbody>
+</table>
+
+<h3 id="特性修改-2md">特性修改</h3>
+
+不涉及
+
+<h3 id="已修复问题-3md">已修复问题</h3>
+
+不涉及
+
+<h3 id="已知问题-4md">已知问题</h3>
+
+<a name="table1969972073016"></a>
+<table><thead align="left"><tr id="row3699162017307"><th class="cellrowborder" valign="top" width="18.22%" id="mcps1.1.3.1.1"><p id="p16992020153010"><a name="p16992020153010"></a><a name="p16992020153010"></a>已知问题</p>
+</th>
+<th class="cellrowborder" valign="top" width="81.78%" id="mcps1.1.3.1.2"><p id="p269919203308"><a name="p269919203308"></a><a name="p269919203308"></a>问题描述</p>
+</th>
+</tr>
+</thead>
+<tbody><tr id="row9699142003011"><td class="cellrowborder" valign="top" width="18.22%" headers="mcps1.1.3.1.1 "><p id="p1769932017300"><a name="p1769932017300"></a><a name="p1769932017300"></a>数据类型支持</p>
+</td>
+<td class="cellrowborder" valign="top" width="81.78%" headers="mcps1.1.3.1.2 "><p id="p13699152010301"><a name="p13699152010301"></a><a name="p13699152010301"></a>NPU不支持float16类型的inf/nan数据输入输出。</p>
+</td>
+</tr>
+<tr id="row146991520153016"><td class="cellrowborder" valign="top" width="18.22%" headers="mcps1.1.3.1.1 "><p id="p156997200308"><a name="p156997200308"></a><a name="p156997200308"></a>数据Format</p>
+</td>
+<td class="cellrowborder" valign="top" width="81.78%" headers="mcps1.1.3.1.2 "><p id="p10699182020308"><a name="p10699182020308"></a><a name="p10699182020308"></a>出现4D以上的format时不能降维。</p>
+</td>
+</tr>
+<tr id="row11121205610549"><td class="cellrowborder" rowspan="3" valign="top" width="18.22%" headers="mcps1.1.3.1.1 "><p id="p1647216219558"><a name="p1647216219558"></a><a name="p1647216219558"></a>集合通信约束</p>
+<p id="p0465121912402"><a name="p0465121912402"></a><a name="p0465121912402"></a></p>
+</td>
+<td class="cellrowborder" valign="top" width="81.78%" headers="mcps1.1.3.1.2 "><p id="p3116115695415"><a name="p3116115695415"></a><a name="p3116115695415"></a>要求一次训练任务中不同device上执行的图相同。</p>
+</td>
+</tr>
+<tr id="row51211656105411"><td class="cellrowborder" valign="top" headers="mcps1.1.3.1.1 "><p id="p1311616560541"><a name="p1311616560541"></a><a name="p1311616560541"></a>当前只支持1/2/4/8P粒度的分配。</p>
+</td>
+</tr>
+<tr id="row8647195765419"><td class="cellrowborder" valign="top" headers="mcps1.1.3.1.1 "><p id="p2064225716544"><a name="p2064225716544"></a><a name="p2064225716544"></a>只支持int8，int32，float16和float32数据类型。</p>
+</td>
+</tr>
+<tr id="row4646195719548"><td class="cellrowborder" valign="top" width="18.22%" headers="mcps1.1.3.1.1 "><p id="p4642195718541"><a name="p4642195718541"></a><a name="p4642195718541"></a>Apex功能支持</p>
+</td>
+<td class="cellrowborder" valign="top" width="81.78%" headers="mcps1.1.3.1.2 "><p id="p864205725412"><a name="p864205725412"></a><a name="p864205725412"></a>Apex当前版本的实现方式主要为python实现，不支持APEX中的自定义优化CUDA Kernel。</p>
+</td>
+</tr>
+</tbody>
+</table>
+
+<h3 id="兼容性-5md">兼容性</h3>
+
+A800-9010：CentOS 7.6/Ubuntu 18.04/BC-Linux 7.6/Debian 9.9/Debian 10/OpenEuler 20.03 LTS
+
+A800-9000：CentOS 7.6/Euler 2.8/Kylin v10/BC-Linux 7.6/OpenEuler 20.03 LTS
+