已合并
【docs】昇腾产品形态名称更新 #75
luyq11创建于 18 天前
【docs】昇腾产品形态名称更新 #75
已合并
共 5 个文件变更+9-5
| @@ -12,6 +12,8 @@ | |||
| 12 | 12 | ||
| 13 | </div> | 13 | </div> |
| 14 | 14 | ||
| 15 | +简体中文 | [English](README_EN.md) | ||
| 16 | + | ||
| 15 | ## ✨ 最新消息 | 17 | ## ✨ 最新消息 |
| 16 | 18 | ||
| 17 | <span style="font-size:14px;"> | 19 | <span style="font-size:14px;"> |
| @@ -10,6 +10,8 @@ | |||
| 10 | 10 | ||
| 11 | </div> | 11 | </div> |
| 12 | 12 | ||
| 13 | +English | [简体中文](README.md) | ||
| 14 | + | ||
| 13 | ## ✨ Latest News | 15 | ## ✨ Latest News |
| 14 | 16 | ||
| 15 | <span style="font-size:14px;"> | 17 | <span style="font-size:14px;"> |
| @@ -223,7 +223,7 @@ def tiling_func(op_type: str, inputs: list = None, outputs: list = None, lib_pat | |||
| 223 | <td class="cellrowborder" valign="top" width="7.5200000000000005%" headers="mcps1.1.5.1.3 "><p id="p13705103352112"><a name="p13705103352112"></a><a name="p13705103352112"></a>可选参数。</p> | 223 | <td class="cellrowborder" valign="top" width="7.5200000000000005%" headers="mcps1.1.5.1.3 "><p id="p13705103352112"><a name="p13705103352112"></a><a name="p13705103352112"></a>可选参数。</p> |
| 224 | </td> | 224 | </td> |
| 225 | <td class="cellrowborder" valign="top" width="65.27%" headers="mcps1.1.5.1.4 "><p id="zh-cn_topic_0000001821790281_p106751081205"><a name="zh-cn_topic_0000001821790281_p106751081205"></a><a name="zh-cn_topic_0000001821790281_p106751081205"></a>配置为<span id="zh-cn_topic_0000001821790281_ph667548162017"><a name="zh-cn_topic_0000001821790281_ph667548162017"></a><a name="zh-cn_topic_0000001821790281_ph667548162017"></a>昇腾AI处理器</span>的类型。</p> | 225 | <td class="cellrowborder" valign="top" width="65.27%" headers="mcps1.1.5.1.4 "><p id="zh-cn_topic_0000001821790281_p106751081205"><a name="zh-cn_topic_0000001821790281_p106751081205"></a><a name="zh-cn_topic_0000001821790281_p106751081205"></a>配置为<span id="zh-cn_topic_0000001821790281_ph667548162017"><a name="zh-cn_topic_0000001821790281_ph667548162017"></a><a name="zh-cn_topic_0000001821790281_ph667548162017"></a>昇腾AI处理器</span>的类型。</p> |
| 226 | -<div class="note" id="zh-cn_topic_0000001821790281_note481620356579"><a name="zh-cn_topic_0000001821790281_note481620356579"></a><a name="zh-cn_topic_0000001821790281_note481620356579"></a><span class="notetitle"> 说明： </span><div class="notebody"><a name="ul1553919272419"></a><a name="ul1553919272419"></a><ul id="ul1553919272419"><li>非<span id="zh-cn_topic_0000002015877373_ph4604666416"><a name="zh-cn_topic_0000002015877373_ph4604666416"></a><a name="zh-cn_topic_0000002015877373_ph4604666416"></a><span id="zh-cn_topic_0000002015877373_zh-cn_topic_0000001740005657_ph11939124012202"><a name="zh-cn_topic_0000002015877373_zh-cn_topic_0000001740005657_ph11939124012202"></a><a name="zh-cn_topic_0000002015877373_zh-cn_topic_0000001740005657_ph11939124012202"></a><term id="zh-cn_topic_0000002015877373_zh-cn_topic_0000001312391781_term1253731311225"><a name="zh-cn_topic_0000002015877373_zh-cn_topic_0000001312391781_term1253731311225"></a><a name="zh-cn_topic_0000002015877373_zh-cn_topic_0000001312391781_term1253731311225"></a>Atlas A3 训练系列产品</term>/<term id="zh-cn_topic_0000002015877373_zh-cn_topic_0000001312391781_term131434243115"><a name="zh-cn_topic_0000002015877373_zh-cn_topic_0000001312391781_term131434243115"></a><a name="zh-cn_topic_0000002015877373_zh-cn_topic_0000001312391781_term131434243115"></a>Atlas A3 推理系列产品</term></span></span>：在安装<span id="zh-cn_topic_0000002015877373_ph26061361142"><a name="zh-cn_topic_0000002015877373_ph26061361142"></a><a name="zh-cn_topic_0000002015877373_ph26061361142"></a>昇腾AI处理器</span>的服务器执行<strong id="zh-cn_topic_0000002015877373_b116061067419"><a name="zh-cn_topic_0000002015877373_b116061067419"></a><a name="zh-cn_topic_0000002015877373_b116061067419"></a>npu-smi info</strong>命令进行查询，获取<strong id="zh-cn_topic_0000002015877373_b15606662046"><a name="zh-cn_topic_0000002015877373_b15606662046"></a><a name="zh-cn_topic_0000002015877373_b15606662046"></a>Chip Name</strong>信息。实际配置值为AscendChip Name，例如<strong id="zh-cn_topic_0000002015877373_b186061865415"><a name="zh-cn_topic_0000002015877373_b186061865415"></a><a name="zh-cn_topic_0000002015877373_b186061865415"></a>Chip Name</strong>取值为<em id="zh-cn_topic_0000002015877373_i1560656248"><a name="zh-cn_topic_0000002015877373_i1560656248"></a><a name="zh-cn_topic_0000002015877373_i1560656248"></a>xxxyy</em>，实际配置值为Ascend<em id="zh-cn_topic_0000002015877373_i9606666418"><a name="zh-cn_topic_0000002015877373_i9606666418"></a><a name="zh-cn_topic_0000002015877373_i9606666418"></a>xxxyy</em>。当Ascendxxxyy为代码样例的路径时，需要配置为ascend<em id="zh-cn_topic_0000002015877373_zh-cn_topic_0000002163321669_i111162923510"><a name="zh-cn_topic_0000002015877373_zh-cn_topic_0000002163321669_i111162923510"></a><a name="zh-cn_topic_0000002015877373_zh-cn_topic_0000002163321669_i111162923510"></a>xxxyy</em>。</li><li><span id="zh-cn_topic_0000002015877373_ph31312041180"><a name="zh-cn_topic_0000002015877373_ph31312041180"></a><a name="zh-cn_topic_0000002015877373_ph31312041180"></a><term id="zh-cn_topic_0000002015877373_zh-cn_topic_0000001312391781_term1253731311225_1"><a name="zh-cn_topic_0000002015877373_zh-cn_topic_0000001312391781_term1253731311225_1"></a><a name="zh-cn_topic_0000002015877373_zh-cn_topic_0000001312391781_term1253731311225_1"></a>Atlas A3 训练系列产品</term>/<term id="zh-cn_topic_0000002015877373_zh-cn_topic_0000001312391781_term131434243115_1"><a name="zh-cn_topic_0000002015877373_zh-cn_topic_0000001312391781_term131434243115_1"></a><a name="zh-cn_topic_0000002015877373_zh-cn_topic_0000001312391781_term131434243115_1"></a>Atlas A3 推理系列产品</term></span>：在安装<span id="zh-cn_topic_0000002015877373_ph151115122044"><a name="zh-cn_topic_0000002015877373_ph151115122044"></a><a name="zh-cn_topic_0000002015877373_ph151115122044"></a>昇腾AI处理器</span>的服务器执行<strong id="zh-cn_topic_0000002015877373_b6114127415"><a name="zh-cn_topic_0000002015877373_b6114127415"></a><a name="zh-cn_topic_0000002015877373_b6114127415"></a>npu-smi info -t board -i </strong><em id="zh-cn_topic_0000002015877373_i611191210417"><a name="zh-cn_topic_0000002015877373_i611191210417"></a><a name="zh-cn_topic_0000002015877373_i611191210417"></a>id</em><strong id="zh-cn_topic_0000002015877373_b1612131210413"><a name="zh-cn_topic_0000002015877373_b1612131210413"></a><a name="zh-cn_topic_0000002015877373_b1612131210413"></a> -c </strong><em id="zh-cn_topic_0000002015877373_i191261213417"><a name="zh-cn_topic_0000002015877373_i191261213417"></a><a name="zh-cn_topic_0000002015877373_i191261213417"></a>chip_id</em>命令进行查询，获取<strong id="zh-cn_topic_0000002015877373_b312111217419"><a name="zh-cn_topic_0000002015877373_b312111217419"></a><a name="zh-cn_topic_0000002015877373_b312111217419"></a>Chip Name</strong>和<strong id="zh-cn_topic_0000002015877373_b16123129419"><a name="zh-cn_topic_0000002015877373_b16123129419"></a><a name="zh-cn_topic_0000002015877373_b16123129419"></a>NPU Name</strong>信息，实际配置值为Chip Name_NPU Name。例如<strong id="zh-cn_topic_0000002015877373_b4125121149"><a name="zh-cn_topic_0000002015877373_b4125121149"></a><a name="zh-cn_topic_0000002015877373_b4125121149"></a>Chip Name</strong>取值为Ascend<em id="zh-cn_topic_0000002015877373_i18125121646"><a name="zh-cn_topic_0000002015877373_i18125121646"></a><a name="zh-cn_topic_0000002015877373_i18125121646"></a>xxx</em>，<strong id="zh-cn_topic_0000002015877373_b141218121448"><a name="zh-cn_topic_0000002015877373_b141218121448"></a><a name="zh-cn_topic_0000002015877373_b141218121448"></a>NPU Name</strong>取值为1234，实际配置值为Ascend<em id="zh-cn_topic_0000002015877373_i16128121419"><a name="zh-cn_topic_0000002015877373_i16128121419"></a><a name="zh-cn_topic_0000002015877373_i16128121419"></a>xxx</em><em id="zh-cn_topic_0000002015877373_i13127121247"><a name="zh-cn_topic_0000002015877373_i13127121247"></a><a name="zh-cn_topic_0000002015877373_i13127121247"></a>_</em>1234。当Ascend<em id="zh-cn_topic_0000002015877373_i197021346113814"><a name="zh-cn_topic_0000002015877373_i197021346113814"></a><a name="zh-cn_topic_0000002015877373_i197021346113814"></a>xxx</em><em id="zh-cn_topic_0000002015877373_i2702846173818"><a name="zh-cn_topic_0000002015877373_i2702846173818"></a><a name="zh-cn_topic_0000002015877373_i2702846173818"></a>_</em>1234为代码样例的路径时，需要配置为ascend<em id="zh-cn_topic_0000002015877373_i19348065392"><a name="zh-cn_topic_0000002015877373_i19348065392"></a><a name="zh-cn_topic_0000002015877373_i19348065392"></a>xxx</em><em id="zh-cn_topic_0000002015877373_i734817633911"><a name="zh-cn_topic_0000002015877373_i734817633911"></a><a name="zh-cn_topic_0000002015877373_i734817633911"></a>_</em>1234。<p id="zh-cn_topic_0000002015877373_p2020212342"><a name="zh-cn_topic_0000002015877373_p2020212342"></a><a name="zh-cn_topic_0000002015877373_p2020212342"></a>其中：</p> | 226 | +<div class="note" id="zh-cn_topic_0000001821790281_note481620356579"><a name="zh-cn_topic_0000001821790281_note481620356579"></a><a name="zh-cn_topic_0000001821790281_note481620356579"></a><span class="notetitle"> 说明： </span><div class="notebody"><a name="ul1553919272419"></a><a name="ul1553919272419"></a><ul id="ul1553919272419"><li>非<span id="zh-cn_topic_0000002015877373_ph4604666416"><a name="zh-cn_topic_0000002015877373_ph4604666416"></a><a name="zh-cn_topic_0000002015877373_ph4604666416"></a><span id="zh-cn_topic_0000002015877373_zh-cn_topic_0000001740005657_ph11939124012202"><a name="zh-cn_topic_0000002015877373_zh-cn_topic_0000001740005657_ph11939124012202"></a><a name="zh-cn_topic_0000002015877373_zh-cn_topic_0000001740005657_ph11939124012202"></a>昇腾A3系列产品</span></span>：在安装<span id="zh-cn_topic_0000002015877373_ph26061361142"><a name="zh-cn_topic_0000002015877373_ph26061361142"></a><a name="zh-cn_topic_0000002015877373_ph26061361142"></a>昇腾AI处理器</span>的服务器执行<strong id="zh-cn_topic_0000002015877373_b116061067419"><a name="zh-cn_topic_0000002015877373_b116061067419"></a><a name="zh-cn_topic_0000002015877373_b116061067419"></a>npu-smi info</strong>命令进行查询，获取<strong id="zh-cn_topic_0000002015877373_b15606662046"><a name="zh-cn_topic_0000002015877373_b15606662046"></a><a name="zh-cn_topic_0000002015877373_b15606662046"></a>Chip Name</strong>信息。实际配置值为AscendChip Name，例如<strong id="zh-cn_topic_0000002015877373_b186061865415"><a name="zh-cn_topic_0000002015877373_b186061865415"></a><a name="zh-cn_topic_0000002015877373_b186061865415"></a>Chip Name</strong>取值为<em id="zh-cn_topic_0000002015877373_i1560656248"><a name="zh-cn_topic_0000002015877373_i1560656248"></a><a name="zh-cn_topic_0000002015877373_i1560656248"></a>xxxyy</em>，实际配置值为Ascend<em id="zh-cn_topic_0000002015877373_i9606666418"><a name="zh-cn_topic_0000002015877373_i9606666418"></a><a name="zh-cn_topic_0000002015877373_i9606666418"></a>xxxyy</em>。当Ascendxxxyy为代码样例的路径时，需要配置为ascend<em id="zh-cn_topic_0000002015877373_zh-cn_topic_0000002163321669_i111162923510"><a name="zh-cn_topic_0000002015877373_zh-cn_topic_0000002163321669_i111162923510"></a><a name="zh-cn_topic_0000002015877373_zh-cn_topic_0000002163321669_i111162923510"></a>xxxyy</em>。</li><li><span id="zh-cn_topic_0000002015877373_ph31312041180"><a name="zh-cn_topic_0000002015877373_ph31312041180"></a><a name="zh-cn_topic_0000002015877373_ph31312041180"></a>昇腾A3系列产品</span>：在安装<span id="zh-cn_topic_0000002015877373_ph151115122044"><a name="zh-cn_topic_0000002015877373_ph151115122044"></a><a name="zh-cn_topic_0000002015877373_ph151115122044"></a>昇腾AI处理器</span>的服务器执行<strong id="zh-cn_topic_0000002015877373_b6114127415"><a name="zh-cn_topic_0000002015877373_b6114127415"></a><a name="zh-cn_topic_0000002015877373_b6114127415"></a>npu-smi info -t board -i </strong><em id="zh-cn_topic_0000002015877373_i611191210417"><a name="zh-cn_topic_0000002015877373_i611191210417"></a><a name="zh-cn_topic_0000002015877373_i611191210417"></a>id</em><strong id="zh-cn_topic_0000002015877373_b1612131210413"><a name="zh-cn_topic_0000002015877373_b1612131210413"></a><a name="zh-cn_topic_0000002015877373_b1612131210413"></a> -c </strong><em id="zh-cn_topic_0000002015877373_i191261213417"><a name="zh-cn_topic_0000002015877373_i191261213417"></a><a name="zh-cn_topic_0000002015877373_i191261213417"></a>chip_id</em>命令进行查询，获取<strong id="zh-cn_topic_0000002015877373_b312111217419"><a name="zh-cn_topic_0000002015877373_b312111217419"></a><a name="zh-cn_topic_0000002015877373_b312111217419"></a>Chip Name</strong>和<strong id="zh-cn_topic_0000002015877373_b16123129419"><a name="zh-cn_topic_0000002015877373_b16123129419"></a><a name="zh-cn_topic_0000002015877373_b16123129419"></a>NPU Name</strong>信息，实际配置值为Chip Name_NPU Name。例如<strong id="zh-cn_topic_0000002015877373_b4125121149"><a name="zh-cn_topic_0000002015877373_b4125121149"></a><a name="zh-cn_topic_0000002015877373_b4125121149"></a>Chip Name</strong>取值为Ascend<em id="zh-cn_topic_0000002015877373_i18125121646"><a name="zh-cn_topic_0000002015877373_i18125121646"></a><a name="zh-cn_topic_0000002015877373_i18125121646"></a>xxx</em>，<strong id="zh-cn_topic_0000002015877373_b141218121448"><a name="zh-cn_topic_0000002015877373_b141218121448"></a><a name="zh-cn_topic_0000002015877373_b141218121448"></a>NPU Name</strong>取值为1234，实际配置值为Ascend<em id="zh-cn_topic_0000002015877373_i16128121419"><a name="zh-cn_topic_0000002015877373_i16128121419"></a><a name="zh-cn_topic_0000002015877373_i16128121419"></a>xxx</em><em id="zh-cn_topic_0000002015877373_i13127121247"><a name="zh-cn_topic_0000002015877373_i13127121247"></a><a name="zh-cn_topic_0000002015877373_i13127121247"></a>_</em>1234。当Ascend<em id="zh-cn_topic_0000002015877373_i197021346113814"><a name="zh-cn_topic_0000002015877373_i197021346113814"></a><a name="zh-cn_topic_0000002015877373_i197021346113814"></a>xxx</em><em id="zh-cn_topic_0000002015877373_i2702846173818"><a name="zh-cn_topic_0000002015877373_i2702846173818"></a><a name="zh-cn_topic_0000002015877373_i2702846173818"></a>_</em>1234为代码样例的路径时，需要配置为ascend<em id="zh-cn_topic_0000002015877373_i19348065392"><a name="zh-cn_topic_0000002015877373_i19348065392"></a><a name="zh-cn_topic_0000002015877373_i19348065392"></a>xxx</em><em id="zh-cn_topic_0000002015877373_i734817633911"><a name="zh-cn_topic_0000002015877373_i734817633911"></a><a name="zh-cn_topic_0000002015877373_i734817633911"></a>_</em>1234。<p id="zh-cn_topic_0000002015877373_p2020212342"><a name="zh-cn_topic_0000002015877373_p2020212342"></a><a name="zh-cn_topic_0000002015877373_p2020212342"></a>其中：</p> |
| 227 | <a name="zh-cn_topic_0000002015877373_ul9238121944"></a><a name="zh-cn_topic_0000002015877373_ul9238121944"></a><ul id="zh-cn_topic_0000002015877373_ul9238121944"><li>id：设备id，通过<strong id="zh-cn_topic_0000002015877373_b723412048"><a name="zh-cn_topic_0000002015877373_b723412048"></a><a name="zh-cn_topic_0000002015877373_b723412048"></a>npu-smi info -l</strong>命令查出的NPU ID即为设备id。</li><li>chip_id：芯片id，通过<strong id="zh-cn_topic_0000002015877373_b172310125417"><a name="zh-cn_topic_0000002015877373_b172310125417"></a><a name="zh-cn_topic_0000002015877373_b172310125417"></a>npu-smi info -m</strong>命令查出的Chip ID即为芯片id。</li></ul> | 227 | <a name="zh-cn_topic_0000002015877373_ul9238121944"></a><a name="zh-cn_topic_0000002015877373_ul9238121944"></a><ul id="zh-cn_topic_0000002015877373_ul9238121944"><li>id：设备id，通过<strong id="zh-cn_topic_0000002015877373_b723412048"><a name="zh-cn_topic_0000002015877373_b723412048"></a><a name="zh-cn_topic_0000002015877373_b723412048"></a>npu-smi info -l</strong>命令查出的NPU ID即为设备id。</li><li>chip_id：芯片id，通过<strong id="zh-cn_topic_0000002015877373_b172310125417"><a name="zh-cn_topic_0000002015877373_b172310125417"></a><a name="zh-cn_topic_0000002015877373_b172310125417"></a>npu-smi info -m</strong>命令查出的Chip ID即为芯片id。</li></ul> |
| 228 | </li></ul> | 228 | </li></ul> |
| 229 | </div></div> | 229 | </div></div> |
| @@ -24,11 +24,11 @@ | |||
| 24 | python3 build.py | 24 | python3 build.py |
| 25 | ``` | 25 | ``` |
| 26 | 26 | ||
| 27 | -#### 构建产物 | 27 | +### 2.1 构建产物 |
| 28 | 28 | ||
| 29 | 构建成功后，输出将生成在 `artifacts` 目录下。 | 29 | 构建成功后，输出将生成在 `artifacts` 目录下。 |
| 30 | 30 | ||
| 31 | -#### 清理构建 | 31 | +### 2.2 清理构建 |
| 32 | 32 | ||
| 33 | 如需全量重新编译，删除构建目录后重新执行上述步骤： | 33 | 如需全量重新编译，删除构建目录后重新执行上述步骤： |
| 34 | 34 | ||
| @@ -66,7 +66,7 @@ MindStudio Kernel Performance Prediction（算子调用工具，msKL）具有调 | |||
| 66 | 66 | ||
| 67 | > [!NOTE] | 67 | > [!NOTE] |
| 68 | > | 68 | > |
| 69 | - >- 本样例工程以Atlas A2 训练系列产品/Atlas A2 推理系列产品为例。 | 69 | + >- 本样例工程以昇腾A2系列产品为例。 |
| 70 | >- 下载代码样例时，需执行以下命令指定分支版本。 | 70 | >- 下载代码样例时，需执行以下命令指定分支版本。 |
| 71 | > | 71 | > |
| 72 | > ```shell | 72 | > ```shell |
| @@ -194,7 +194,7 @@ MindStudio Kernel Performance Prediction（算子调用工具，msKL）具有调 | |||
| 194 | 194 | ||
| 195 | ### 注意事项 | 195 | ### 注意事项 |
| 196 | 196 | ||
| 197 | -- 自动调优功能仅支持Atlas A2 训练系列产品/Atlas A2 推理系列产品。 | 197 | +- 自动调优功能仅支持昇腾A2系列产品。 |
| 198 | - 单Device仅支持使用单个msKL工具进行自动调优，且不推荐同时运行其他算子程序。 | 198 | - 单Device仅支持使用单个msKL工具进行自动调优，且不推荐同时运行其他算子程序。 |
| 199 | - 需确保先import mskl再import acl，否则需要在运行前设置环境变量。 | 199 | - 需确保先import mskl再import acl，否则需要在运行前设置环境变量。 |
| 200 | 200 | ||