From dab90c961e8888a57a3c5af823cd87d93f3fa99b Mon Sep 17 00:00:00 2001 From: is_dc <1140924077@qq.com> Date: Sat, 2 Apr 2022 07:39:01 +0000 Subject: [PATCH 01/16] update TensorFlow/built-in/nlp/albert_xlarge_zh_ID2348_for_TensorFlow/README.md. --- .../README.md | 59 +++++++------------ 1 file changed, 21 insertions(+), 38 deletions(-) diff --git a/TensorFlow/built-in/nlp/albert_xlarge_zh_ID2348_for_TensorFlow/README.md b/TensorFlow/built-in/nlp/albert_xlarge_zh_ID2348_for_TensorFlow/README.md index 2a5ae8606..efef619e1 100644 --- a/TensorFlow/built-in/nlp/albert_xlarge_zh_ID2348_for_TensorFlow/README.md +++ b/TensorFlow/built-in/nlp/albert_xlarge_zh_ID2348_for_TensorFlow/README.md @@ -1,9 +1,9 @@ -- [基本信息](#基本信息.md) -- [概述](#概述.md) -- [训练环境准备](#训练环境准备.md) -- [快速上手](#快速上手.md) -- [迁移学习指导](#迁移学习指导.md) -- [高级参考](#高级参考.md) +- [基本信息](#基本信息.md) +- [概述](#概述.md) +- [训练环境准备](#训练环境准备.md) +- [快速上手](#快速上手.md) +- [迁移学习指导](#迁移学习指导.md) +- [高级参考](#高级参考.md)

基本信息

**发布者（Publisher）：Huawei** @@ -43,15 +43,17 @@ - 适配昇腾 AI 处理器的实现： - https://gitee.com/ascend/modelzoo/tree/master/built-in/TensorFlow/Benchmark/nlp/albert_xlarge_zh_ID2348_for_TensorFlow + https://gitee.com/ascend/ModelZoo-TensorFlow/tree/master/TensorFlow/built-in/nlp/albert_xlarge_zh_ID2348_for_TensorFlow - 通过Git获取对应commit_id的代码方法如下: - git clone {repository_url} # 克隆仓库的代码 - cd {repository_name} # 切换到模型的代码仓目录 - git checkout {branch} # 切换到对应分支 - git reset --hard ｛commit_id｝ # 代码设置到对应的commit_id - cd ｛code_path｝ # 切换到模型代码所在路径，若仓库下只有该模型，则无需切换 + ``` + git clone {repository_url} # 克隆仓库的代码 + cd {repository_name} # 切换到模型的代码仓目录 + git checkout {branch} # 切换到对应分支 + git reset --hard ｛commit_id｝ # 代码设置到对应的commit_id + cd ｛code_path｝ # 切换到模型代码所在路径，若仓库下只有该模型，则无需切换 + ``` ## 默认配置 @@ -99,38 +101,19 @@

训练环境准备

-1. 硬件环境准备请参见各硬件产品文档"[驱动和固件安装升级指南]( https://support.huawei.com/enterprise/zh/category/ai-computing-platform-pid-1557196528909)"。需要在硬件设备上安装与CANN版本配套的固件与驱动。 -2. 宿主机上需要安装Docker并登录[Ascend Hub中心](https://ascendhub.huawei.com/#/detail?name=ascend-tensorflow-arm)获取镜像。 - - 当前模型支持的镜像列表如[表1](#zh-cn_topic_0000001074498056_table1519011227314)所示。 - - **表 1** 镜像列表 - - - - - - - - - - - - -

镜像名称

-

镜像版本

-

配套CANN版本

-
-

21.0.2

-

5.0.2

-
+- 硬件环境和运行环境准备请参见《[CANN软件安装指南](https://support.huawei.com/enterprise/zh/ascend-computing/cann-pid-251168373?category=installation-update)》 +- 运行以下命令安装依赖。 +``` +pip3 install requirements.txt +``` +说明：依赖配置文件requirements.txt文件位于模型的根目录

快速上手

## 数据集准备 -1. 用户需自行准备训练数据集，例如LCQMC数据集，可参考github源。 +1. 用户需自行准备训练数据集，例如LCQMC数据集，可参考github源获取。 2. 用户需提前下载预训练模型，参考github源里提供的albert_xlarge_zh_183k等模型下载方式 ## 模型训练 -- Gitee From a6fa5afb3d8246137c7e15f694b2964a15fb474a Mon Sep 17 00:00:00 2001 From: is_dc <1140924077@qq.com> Date: Sat, 2 Apr 2022 08:02:14 +0000 Subject: [PATCH 02/16] update TensorFlow/built-in/nlp/albert_xlarge_zh_ID2348_for_TensorFlow/README.md. --- .../nlp/albert_xlarge_zh_ID2348_for_TensorFlow/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/TensorFlow/built-in/nlp/albert_xlarge_zh_ID2348_for_TensorFlow/README.md b/TensorFlow/built-in/nlp/albert_xlarge_zh_ID2348_for_TensorFlow/README.md index efef619e1..03c9c6323 100644 --- a/TensorFlow/built-in/nlp/albert_xlarge_zh_ID2348_for_TensorFlow/README.md +++ b/TensorFlow/built-in/nlp/albert_xlarge_zh_ID2348_for_TensorFlow/README.md @@ -45,7 +45,7 @@ https://gitee.com/ascend/ModelZoo-TensorFlow/tree/master/TensorFlow/built-in/nlp/albert_xlarge_zh_ID2348_for_TensorFlow -- 通过Git获取对应commit_id的代码方法如下: +- 通过Git获取对应commit\_id的代码方法如下: ``` git clone {repository_url} # 克隆仓库的代码 -- Gitee From 08721e3143489d5133f16659e9b32808d1dfb825 Mon Sep 17 00:00:00 2001 From: is_dc <1140924077@qq.com> Date: Sat, 2 Apr 2022 08:08:42 +0000 Subject: [PATCH 03/16] update TensorFlow/built-in/nlp/albert_xlarge_zh_ID2348_for_TensorFlow/README.md. --- .../nlp/albert_xlarge_zh_ID2348_for_TensorFlow/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/TensorFlow/built-in/nlp/albert_xlarge_zh_ID2348_for_TensorFlow/README.md b/TensorFlow/built-in/nlp/albert_xlarge_zh_ID2348_for_TensorFlow/README.md index 03c9c6323..7e8a8b906 100644 --- a/TensorFlow/built-in/nlp/albert_xlarge_zh_ID2348_for_TensorFlow/README.md +++ b/TensorFlow/built-in/nlp/albert_xlarge_zh_ID2348_for_TensorFlow/README.md @@ -45,7 +45,7 @@ https://gitee.com/ascend/ModelZoo-TensorFlow/tree/master/TensorFlow/built-in/nlp/albert_xlarge_zh_ID2348_for_TensorFlow -- 通过Git获取对应commit\_id的代码方法如下: +- 通过Git获取对应commit\_id的代码方法如下： ``` git clone {repository_url} # 克隆仓库的代码 -- Gitee From 3c9a5d02a7fb223f66d863e0a19783d8fdfa3f70 Mon Sep 17 00:00:00 2001 From: is_dc <1140924077@qq.com> Date: Sat, 2 Apr 2022 08:12:15 +0000 Subject: [PATCH 04/16] update TensorFlow2/built-in/cv/image_classification/cutmix_ID2502_for_TensorFlow2.X/README.md. --- .../cutmix_ID2502_for_TensorFlow2.X/README.md | 34 ++++--------------- 1 file changed, 7 insertions(+), 27 deletions(-) diff --git a/TensorFlow2/built-in/cv/image_classification/cutmix_ID2502_for_TensorFlow2.X/README.md b/TensorFlow2/built-in/cv/image_classification/cutmix_ID2502_for_TensorFlow2.X/README.md index d8b156685..bae8c4487 100644 --- a/TensorFlow2/built-in/cv/image_classification/cutmix_ID2502_for_TensorFlow2.X/README.md +++ b/TensorFlow2/built-in/cv/image_classification/cutmix_ID2502_for_TensorFlow2.X/README.md @@ -46,7 +46,7 @@ https://gitee.com/ascend/modelzoo/tree/master/built-in/TensorFlow2/Official/cv/image_classification/cutmix_ID2502_for_TensorFlow2.X -- 通过Git获取对应commit_id的代码方法如下: +- 通过Git获取对应commit\_id的代码方法如下： git clone {repository_url} # 克隆仓库的代码 cd {repository_name} # 切换到模型的代码仓目录 @@ -81,34 +81,14 @@ npu_device.open().as_default()) ``` -

训练环境准备

-1. 硬件环境准备请参见各硬件产品文档"[驱动和固件安装升级指南]( https://support.huawei.com/enterprise/zh/category/ai-computing-platform-pid-1557196528909)"。需要在硬件设备上安装与CANN版本配套的固件与驱动。 -2. 宿主机上需要安装Docker并登录[Ascend Hub中心](https://ascendhub.huawei.com/#/detail?name=ascend-tensorflow-arm)获取镜像。 - - 当前模型支持的镜像列表如[表1](#zh-cn_topic_0000001074498056_table1519011227314)所示。 - - **表 1** 镜像列表 - - - - - - - - - - - - -

镜像名称

-

镜像版本

-

配套CANN版本

-
-

21.0.2

-

5.0.2

-
+- 硬件环境和运行环境准备请参见《[CANN软件安装指南](https://support.huawei.com/enterprise/zh/ascend-computing/cann-pid-251168373?category=installation-update)》 +- 运行以下命令安装依赖。 +``` +pip3 install requirements.txt +``` +说明：依赖配置文件requirements.txt文件位于模型的根目录

快速上手

-- Gitee From 71d16050d980cb857f897da57f55249e8db11133 Mon Sep 17 00:00:00 2001 From: is_dc <1140924077@qq.com> Date: Sat, 2 Apr 2022 08:14:52 +0000 Subject: [PATCH 05/16] update TensorFlow2/built-in/cv/image_classification/cutmix_ID2502_for_TensorFlow2.X/README.md. --- .../cutmix_ID2502_for_TensorFlow2.X/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/TensorFlow2/built-in/cv/image_classification/cutmix_ID2502_for_TensorFlow2.X/README.md b/TensorFlow2/built-in/cv/image_classification/cutmix_ID2502_for_TensorFlow2.X/README.md index bae8c4487..7f3c71a27 100644 --- a/TensorFlow2/built-in/cv/image_classification/cutmix_ID2502_for_TensorFlow2.X/README.md +++ b/TensorFlow2/built-in/cv/image_classification/cutmix_ID2502_for_TensorFlow2.X/README.md @@ -44,7 +44,7 @@ - 适配昇腾 AI 处理器的实现： - https://gitee.com/ascend/modelzoo/tree/master/built-in/TensorFlow2/Official/cv/image_classification/cutmix_ID2502_for_TensorFlow2.X + https://gitee.com/ascend/ModelZoo-TensorFlow/tree/master/TensorFlow2/built-in/cv/image_classification/cutmix_ID2502_for_TensorFlow2.X - 通过Git获取对应commit\_id的代码方法如下： -- Gitee From 6c38ac5d5fc37c24f2555167576f1f0c336f76c6 Mon Sep 17 00:00:00 2001 From: is_dc <1140924077@qq.com> Date: Sat, 2 Apr 2022 08:18:40 +0000 Subject: [PATCH 06/16] update TensorFlow/built-in/cv/image_synthesis/DCGAN_ID2196_for_TensorFlow/README.md. --- .../DCGAN_ID2196_for_TensorFlow/README.md | 36 +++++-------------- 1 file changed, 8 insertions(+), 28 deletions(-) diff --git a/TensorFlow/built-in/cv/image_synthesis/DCGAN_ID2196_for_TensorFlow/README.md b/TensorFlow/built-in/cv/image_synthesis/DCGAN_ID2196_for_TensorFlow/README.md index 66bb6a2f5..eefdb8485 100644 --- a/TensorFlow/built-in/cv/image_synthesis/DCGAN_ID2196_for_TensorFlow/README.md +++ b/TensorFlow/built-in/cv/image_synthesis/DCGAN_ID2196_for_TensorFlow/README.md @@ -45,9 +45,9 @@ - 适配昇腾 AI 处理器的实现： - https://gitee.com/ascend/modelzoo/tree/master/built-in/TensorFlow/Research/cv/image_synthesis/DCGAN_ID2196_for_TensorFlow + https://gitee.com/ascend/ModelZoo-TensorFlow/tree/master/TensorFlow/built-in/cv/image_synthesis/DCGAN_ID2196_for_TensorFlow -- 通过Git获取对应commit_id的代码方法如下: +- 通过Git获取对应commit\_id的代码方法如下： git clone {repository_url} # 克隆仓库的代码 cd {repository_name} # 切换到模型的代码仓目录 @@ -90,34 +90,14 @@ custom_op.parameter_map["precision_mode"].s = tf.compat.as_bytes("allow_mix_precision") ``` -

训练环境准备

-1. 硬件环境准备请参见各硬件产品文档"[驱动和固件安装升级指南]( https://support.huawei.com/enterprise/zh/category/ai-computing-platform-pid-1557196528909)"。需要在硬件设备上安装与CANN版本配套的固件与驱动。 -2. 宿主机上需要安装Docker并登录[Ascend Hub中心](https://ascendhub.huawei.com/#/detail?name=ascend-tensorflow-arm)获取镜像。 - - 当前模型支持的镜像列表如[表1](#zh-cn_topic_0000001074498056_table1519011227314)所示。 - - **表 1** 镜像列表 - - - - - - - - - - - - -

镜像名称

-

镜像版本

-

配套CANN版本

-
-

21.0.2

-

5.0.2

-
+- 硬件环境和运行环境准备请参见《[CANN软件安装指南](https://support.huawei.com/enterprise/zh/ascend-computing/cann-pid-251168373?category=installation-update)》 +- 运行以下命令安装依赖。 +``` +pip3 install requirements.txt +``` +说明：依赖配置文件requirements.txt文件位于模型的根目录

快速上手

-- Gitee From 136c659bc22f65c27c69ae92704faaa31de427eb Mon Sep 17 00:00:00 2001 From: is_dc <1140924077@qq.com> Date: Sat, 2 Apr 2022 08:21:11 +0000 Subject: [PATCH 07/16] update TensorFlow2/built-in/cv/image_classification/pointnet_ID2531_for_TensorFlow2.X/README.md. --- .../README.md | 36 +++++-------------- 1 file changed, 8 insertions(+), 28 deletions(-) diff --git a/TensorFlow2/built-in/cv/image_classification/pointnet_ID2531_for_TensorFlow2.X/README.md b/TensorFlow2/built-in/cv/image_classification/pointnet_ID2531_for_TensorFlow2.X/README.md index 680c008a9..5c53068df 100644 --- a/TensorFlow2/built-in/cv/image_classification/pointnet_ID2531_for_TensorFlow2.X/README.md +++ b/TensorFlow2/built-in/cv/image_classification/pointnet_ID2531_for_TensorFlow2.X/README.md @@ -45,9 +45,9 @@ - 适配昇腾 AI 处理器的实现： - https://gitee.com/ascend/modelzoo/tree/master/built-in/TensorFlow2/Official/cv/image_classification/pointnet_ID2531_for_TensorFlow2.X + https://gitee.com/ascend/ModelZoo-TensorFlow/tree/master/TensorFlow2/built-in/cv/image_classification/pointnet_ID2531_for_TensorFlow2.X -- 通过Git获取对应commit_id的代码方法如下: +- 通过Git获取对应commit\_id的代码方法如下： git clone {repository_url} # 克隆仓库的代码 cd {repository_name} # 切换到模型的代码仓目录 @@ -92,34 +92,14 @@ ############维测参数############## ``` -

训练环境准备

-1. 硬件环境准备请参见各硬件产品文档"[驱动和固件安装升级指南]( https://support.huawei.com/enterprise/zh/category/ai-computing-platform-pid-1557196528909)"。需要在硬件设备上安装与CANN版本配套的固件与驱动。 -2. 宿主机上需要安装Docker并登录[Ascend Hub中心](https://ascendhub.huawei.com/#/detail?name=ascend-tensorflow-arm)获取镜像。 - - 当前模型支持的镜像列表如[表1](#zh-cn_topic_0000001074498056_table1519011227314)所示。 - - **表 1** 镜像列表 - - - - - - - - - - - - -

镜像名称

-

镜像版本

-

配套CANN版本

-
-

21.0.2

-

5.0.2

-
+- 硬件环境和运行环境准备请参见《[CANN软件安装指南](https://support.huawei.com/enterprise/zh/ascend-computing/cann-pid-251168373?category=installation-update)》 +- 运行以下命令安装依赖。 +``` +pip3 install requirements.txt +``` +说明：依赖配置文件requirements.txt文件位于模型的根目录

快速上手

-- Gitee From b075636c8389be04d3b0781443c8498983093796 Mon Sep 17 00:00:00 2001 From: is_dc <1140924077@qq.com> Date: Sat, 2 Apr 2022 08:23:56 +0000 Subject: [PATCH 08/16] =?UTF-8?q?=E4=BF=AE=E6=94=B9readme?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../README.md | 36 +++++-------------- 1 file changed, 8 insertions(+), 28 deletions(-) diff --git a/TensorFlow2/built-in/cv/image_segmentation/pointnet_segmentation_ID2532_for_TensorFlow2.X/README.md b/TensorFlow2/built-in/cv/image_segmentation/pointnet_segmentation_ID2532_for_TensorFlow2.X/README.md index f441c4e2a..2bf64ad74 100644 --- a/TensorFlow2/built-in/cv/image_segmentation/pointnet_segmentation_ID2532_for_TensorFlow2.X/README.md +++ b/TensorFlow2/built-in/cv/image_segmentation/pointnet_segmentation_ID2532_for_TensorFlow2.X/README.md @@ -45,9 +45,9 @@ - 适配昇腾 AI 处理器的实现： - https://gitee.com/ascend/modelzoo/tree/master/built-in/TensorFlow2/Official/cv/image_segmentation/pointnet_segmentation_ID2532_for_TensorFlow2.X + https://gitee.com/ascend/ModelZoo-TensorFlow/tree/master/TensorFlow2/built-in/cv/image_segmentation/pointnet_segmentation_ID2532_for_TensorFlow2.X -- 通过Git获取对应commit_id的代码方法如下: +- 通过Git获取对应commit\_id的代码方法如下： git clone {repository_url} # 克隆仓库的代码 cd {repository_name} # 切换到模型的代码仓目录 @@ -96,34 +96,14 @@ ############维测参数############## ``` -

训练环境准备

-1. 硬件环境准备请参见各硬件产品文档"[驱动和固件安装升级指南]( https://support.huawei.com/enterprise/zh/category/ai-computing-platform-pid-1557196528909)"。需要在硬件设备上安装与CANN版本配套的固件与驱动。 -2. 宿主机上需要安装Docker并登录[Ascend Hub中心](https://ascendhub.huawei.com/#/detail?name=ascend-tensorflow-arm)获取镜像。 - - 当前模型支持的镜像列表如[表1](#zh-cn_topic_0000001074498056_table1519011227314)所示。 - - **表 1** 镜像列表 - - - - - - - - - - - - -

镜像名称

-

镜像版本

-

配套CANN版本

-
-

21.0.2

-

5.0.2

-
+- 硬件环境和运行环境准备请参见《[CANN软件安装指南](https://support.huawei.com/enterprise/zh/ascend-computing/cann-pid-251168373?category=installation-update)》 +- 运行以下命令安装依赖。 +``` +pip3 install requirements.txt +``` +说明：依赖配置文件requirements.txt文件位于模型的根目录

快速上手

-- Gitee From cdacf3fef24b123503e76dbcd304b63b215ff1f5 Mon Sep 17 00:00:00 2001 From: is_dc <1140924077@qq.com> Date: Sat, 2 Apr 2022 08:26:04 +0000 Subject: [PATCH 09/16] =?UTF-8?q?=E4=BF=AE=E6=94=B9README.md.?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../pointnet_segmentation_ID2532_for_TensorFlow2.X/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/TensorFlow2/built-in/cv/image_segmentation/pointnet_segmentation_ID2532_for_TensorFlow2.X/README.md b/TensorFlow2/built-in/cv/image_segmentation/pointnet_segmentation_ID2532_for_TensorFlow2.X/README.md index 2bf64ad74..a288630cd 100644 --- a/TensorFlow2/built-in/cv/image_segmentation/pointnet_segmentation_ID2532_for_TensorFlow2.X/README.md +++ b/TensorFlow2/built-in/cv/image_segmentation/pointnet_segmentation_ID2532_for_TensorFlow2.X/README.md @@ -8,7 +8,7 @@ **发布者（Publisher）：Huawei** -**应用领域（Application Domain）： Image Segmentation** +**应用领域（Application Domain）： Instance Segmentation** **版本（Version）：1.1** -- Gitee From b792ea150ccde7b79988774625c8af10b0dbfa21 Mon Sep 17 00:00:00 2001 From: is_dc <1140924077@qq.com> Date: Sat, 2 Apr 2022 08:28:54 +0000 Subject: [PATCH 10/16] update TensorFlow/built-in/nlp/Roberta_ID2349_for_TensorFlow/README.md. --- .../Roberta_ID2349_for_TensorFlow/README.md | 48 ++++++------------- 1 file changed, 14 insertions(+), 34 deletions(-) diff --git a/TensorFlow/built-in/nlp/Roberta_ID2349_for_TensorFlow/README.md b/TensorFlow/built-in/nlp/Roberta_ID2349_for_TensorFlow/README.md index ae2c21ad1..290030209 100644 --- a/TensorFlow/built-in/nlp/Roberta_ID2349_for_TensorFlow/README.md +++ b/TensorFlow/built-in/nlp/Roberta_ID2349_for_TensorFlow/README.md @@ -1,9 +1,9 @@ -- [基本信息](#基本信息.md) -- [概述](#概述.md) -- [训练环境准备](#训练环境准备.md) -- [快速上手](#快速上手.md) -- [迁移学习指导](#迁移学习指导.md) -- [高级参考](#高级参考.md) +- [基本信息](#基本信息.md) +- [概述](#概述.md) +- [训练环境准备](#训练环境准备.md) +- [快速上手](#快速上手.md) +- [迁移学习指导](#迁移学习指导.md) +- [高级参考](#高级参考.md)

基本信息

**发布者（Publisher）：Huawei** @@ -44,9 +44,9 @@ - 适配昇腾 AI 处理器的实现： - https://gitee.com/ascend/modelzoo/tree/master/built-in/TensorFlow/Benchmark/nlp/Roberta_ID2349_for_TensorFlow + https://gitee.com/ascend/ModelZoo-TensorFlow/tree/master/TensorFlow/built-in/nlp/Roberta_ID2349_for_TensorFlow -- 通过Git获取对应commit_id的代码方法如下: +- 通过Git获取对应commit\_id的代码方法如下： git clone {repository_url} # 克隆仓库的代码 cd {repository_name} # 切换到模型的代码仓目录 @@ -88,34 +88,14 @@ custom_op.parameter_map["precision_mode"].s = tf.compat.as_bytes("allow_mix_precision") ``` -

训练环境准备

-1. 硬件环境准备请参见各硬件产品文档"[驱动和固件安装升级指南]( https://support.huawei.com/enterprise/zh/category/ai-computing-platform-pid-1557196528909)"。需要在硬件设备上安装与CANN版本配套的固件与驱动。 -2. 宿主机上需要安装Docker并登录[Ascend Hub中心](https://ascendhub.huawei.com/#/detail?name=ascend-tensorflow-arm)获取镜像。 - - 当前模型支持的镜像列表如[表1](#zh-cn_topic_0000001074498056_table1519011227314)所示。 - - **表 1** 镜像列表 - - - - - - - - - - - - -

镜像名称

-

镜像版本

-

配套CANN版本

-
-

21.0.2

-

5.0.2

-
+- 硬件环境和运行环境准备请参见《[CANN软件安装指南](https://support.huawei.com/enterprise/zh/ascend-computing/cann-pid-251168373?category=installation-update)》 +- 运行以下命令安装依赖。 +``` +pip3 install requirements.txt +``` +说明：依赖配置文件requirements.txt文件位于模型的根目录

快速上手

-- Gitee From c09f900827658fec3e7dceff6462e0221f3aa316 Mon Sep 17 00:00:00 2001 From: is_dc <1140924077@qq.com> Date: Sat, 2 Apr 2022 08:31:04 +0000 Subject: [PATCH 11/16] update TensorFlow/built-in/nlp/Roberta_ID2366_for_TensorFlow/README.md. --- .../Roberta_ID2366_for_TensorFlow/README.md | 36 +++++-------------- 1 file changed, 8 insertions(+), 28 deletions(-) diff --git a/TensorFlow/built-in/nlp/Roberta_ID2366_for_TensorFlow/README.md b/TensorFlow/built-in/nlp/Roberta_ID2366_for_TensorFlow/README.md index 154d9c58f..90930e827 100644 --- a/TensorFlow/built-in/nlp/Roberta_ID2366_for_TensorFlow/README.md +++ b/TensorFlow/built-in/nlp/Roberta_ID2366_for_TensorFlow/README.md @@ -44,9 +44,9 @@ - 适配昇腾 AI 处理器的实现： - https://gitee.com/ascend/modelzoo/tree/master/built-in/TensorFlow/Benchmark/nlp/Roberta_ID2366_for_TensorFlow + https://gitee.com/ascend/ModelZoo-TensorFlow/tree/master/TensorFlow/built-in/nlp/Roberta_ID2366_for_TensorFlow -- 通过Git获取对应commit_id的代码方法如下: +- 通过Git获取对应commit\_id的代码方法如下： git clone {repository_url} # 克隆仓库的代码 cd {repository_name} # 切换到模型的代码仓目录 @@ -90,34 +90,14 @@ session_config.graph_options.rewrite_options.memory_optimization = RewriterConfig.OFF ``` -

训练环境准备

-1. 硬件环境准备请参见各硬件产品文档"[驱动和固件安装升级指南]( https://support.huawei.com/enterprise/zh/category/ai-computing-platform-pid-1557196528909)"。需要在硬件设备上安装与CANN版本配套的固件与驱动。 -2. 宿主机上需要安装Docker并登录[Ascend Hub中心](https://ascendhub.huawei.com/#/detail?name=ascend-tensorflow-arm)获取镜像。 - - 当前模型支持的镜像列表如[表1](#zh-cn_topic_0000001074498056_table1519011227314)所示。 - - **表 1** 镜像列表 - - - - - - - - - - - - -

镜像名称

-

镜像版本

-

配套CANN版本

-
-

21.0.2

-

5.0.2

-
+- 硬件环境和运行环境准备请参见《[CANN软件安装指南](https://support.huawei.com/enterprise/zh/ascend-computing/cann-pid-251168373?category=installation-update)》 +- 运行以下命令安装依赖。 +``` +pip3 install requirements.txt +``` +说明：依赖配置文件requirements.txt文件位于模型的根目录

快速上手

-- Gitee From e84a4173d84d4499a3f49e012c8e77098237bf89 Mon Sep 17 00:00:00 2001 From: is_dc <1140924077@qq.com> Date: Sat, 2 Apr 2022 08:32:17 +0000 Subject: [PATCH 12/16] update TensorFlow/built-in/nlp/Roberta_ID2366_for_TensorFlow/README.md. --- .../nlp/Roberta_ID2366_for_TensorFlow/README.md | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/TensorFlow/built-in/nlp/Roberta_ID2366_for_TensorFlow/README.md b/TensorFlow/built-in/nlp/Roberta_ID2366_for_TensorFlow/README.md index 90930e827..1dfa3c189 100644 --- a/TensorFlow/built-in/nlp/Roberta_ID2366_for_TensorFlow/README.md +++ b/TensorFlow/built-in/nlp/Roberta_ID2366_for_TensorFlow/README.md @@ -1,9 +1,9 @@ -- [基本信息](#基本信息.md) -- [概述](#概述.md) -- [训练环境准备](#训练环境准备.md) -- [快速上手](#快速上手.md) -- [迁移学习指导](#迁移学习指导.md) -- [高级参考](#高级参考.md) +- [基本信息](#基本信息.md) +- [概述](#概述.md) +- [训练环境准备](#训练环境准备.md) +- [快速上手](#快速上手.md) +- [迁移学习指导](#迁移学习指导.md) +- [高级参考](#高级参考.md)

基本信息

**发布者（Publisher）：Huawei** -- Gitee From f8ab00d91ebf7774d72471a0c8f8b1a3b4b9e3ee Mon Sep 17 00:00:00 2001 From: is_dc <1140924077@qq.com> Date: Sat, 2 Apr 2022 08:35:59 +0000 Subject: [PATCH 13/16] update TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md. --- .../README.md | 54 +++++++++---------- 1 file changed, 26 insertions(+), 28 deletions(-) diff --git a/TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md b/TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md index 18a869bd3..38c6ae1e1 100644 --- a/TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md +++ b/TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md @@ -36,7 +36,7 @@ SSD-Resnet50V1-FPN将边界框的输出空间离散为一组默认框，每个 - 适配昇腾 AI 处理器的实现： - https://gitee.com/ascend/modelzoo/tree/master/built-in/TensorFlow/Research/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow + https://gitee.com/ascend/ModelZoo-TensorFlow/tree/master/TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow - 通过Git获取对应commit\_id的代码方法如下： @@ -77,34 +77,32 @@ SSD-Resnet50V1-FPN将边界框的输出空间离散为一组默认框，每个 | 分布式训练 | 是 | | 混合精度 | 是 | +## 混合精度训练 + +昇腾910 AI处理器提供自动混合精度功能，可以针对全网中float32数据类型的算子，按照内置的优化策略，自动将部分float32的算子降低精度到float16，从而在精度损失很小的情况下提升系统性能并减少内存使用。 + +## 开启混合精度 +相关代码示例。 + +``` + config_proto = tf.ConfigProto(allow_soft_placement=True) + custom_op = config_proto.graph_options.rewrite_options.custom_optimizers.add() + custom_op.name = 'NpuOptimizer' + custom_op.parameter_map["use_off_line"].b = True + custom_op.parameter_map["precision_mode"].s = tf.compat.as_bytes("allow_mix_precision") + config_proto.graph_options.rewrite_options.remapping = RewriterConfig.OFF + session_config = npu_config_proto(config_proto=config_proto) +``` + +

训练环境准备

+ +- 硬件环境和运行环境准备请参见《[CANN软件安装指南](https://support.huawei.com/enterprise/zh/ascend-computing/cann-pid-251168373?category=installation-update)》 +- 运行以下命令安装依赖。 +``` +pip3 install requirements.txt +``` +说明：依赖配置文件requirements.txt文件位于模型的根目录 -## 训练环境准备 - -1. 硬件环境准备请参见各硬件产品文档"[驱动和固件安装升级指南]( https://support.huawei.com/enterprise/zh/category/ai-computing-platform-pid-1557196528909)"。需要在硬件设备上安装与CANN版本配套的固件与驱动。 -2. 宿主机上需要安装Docker并登录[Ascend Hub中心](https://ascendhub.huawei.com/#/detail?name=ascend-tensorflow-arm)获取镜像。 - - 当前模型支持的镜像列表如[表1](#zh-cn_topic_0000001074498056_table1519011227314)所示。 - - **表 1** 镜像列表 - - - - - - - - - - - - -

镜像名称

-

镜像版本

-

配套CANN版本

-
-

21.0.2

-

5.0.2

-
## 安装依赖 -- Gitee From 9e62675708afe4354616f546f953742a7f58970a Mon Sep 17 00:00:00 2001 From: is_dc <1140924077@qq.com> Date: Sat, 2 Apr 2022 08:42:26 +0000 Subject: [PATCH 14/16] update TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md. --- .../README.md | 34 +++++++++++-------- 1 file changed, 19 insertions(+), 15 deletions(-) diff --git a/TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md b/TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md index 38c6ae1e1..3ba870175 100644 --- a/TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md +++ b/TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md @@ -1,14 +1,12 @@ -# SSD-Resnet50V1-FPN_for_TensorFlow +- [基本信息](#基本信息.md) +- [概述](#概述.md) +- [训练环境准备](#训练环境准备.md) +- [快速上手](#快速上手.md) +- [迁移学习指导](#迁移学习指导.md) +- [高级参考](#高级参考.md) -## 目录 -* [基本信息](#基本信息) -* [概述](#概述) -* [训练环境准备](#训练环境准备) -* [快速上手](#快速上手) -* [高级参考](#高级参考) - -## 基本信息 +

基本信息

**发布者（Publisher）：Huawei** **应用领域（Application Domain）：Object Detection @@ -22,7 +20,7 @@ **应用级别（Categories）：Official **描述（Description）：基于tensorflow实现，以Resnet50为backbone的SSD目标检测网络。 -## 概述 +

概述

SSD-Resnet50V1-FPN将边界框的输出空间离散为一组默认框，每个特征地图位置的纵横比和比例不同。在预测时，网络为每个默认框中每个对象类别生成分数，并对该框进行调整，以更好地匹配对象形状。此外，该网络结合了来自不同分辨率的多个特征图的预测，从而自然地处理不同尺寸的物体 @@ -112,7 +110,7 @@ pip3 install requirements.txt 3、mpl_toolkits -## 快速上手 +

快速上手

### 数据集准备 @@ -124,10 +122,11 @@ download_all.sh nvidia_ssd ### 模型训练 - 单击“立即下载”，并选择合适的下载方式下载源码包。 +- 开始训练 + + 1. 启动训练之前，首先要配置程序运行相关环境变量。 -- 启动训练之前，首先要配置程序运行相关环境变量。 - - 环境变量配置信息参见： + 环境变量配置信息参见： [Ascend 910训练平台环境变量设置](https://gitee.com/ascend/modelzoo/wikis/Ascend%20910%E8%AE%AD%E7%BB%83%E5%B9%B3%E5%8F%B0%E7%8E%AF%E5%A2%83%E5%8F%98%E9%87%8F%E8%AE%BE%E7%BD%AE?sort_id=3148819) @@ -162,7 +161,7 @@ download_all.sh nvidia_ssd ## 高级参考 -### 核心脚本和示例代码 +## 脚本和示例代码 ``` ├── README.md //代码说明文档 @@ -236,3 +235,8 @@ download_all.sh nvidia_ssd --model_dir 训练模型存放路径 --checkpoint_dir 待评测的模型路径（仅eval模式使用） ``` + +## 训练过程 +``` + NA +``` \ No newline at end of file -- Gitee From ec76a6d250e0817c0de5b67d9ac467b50ef4f3fa Mon Sep 17 00:00:00 2001 From: is_dc <1140924077@qq.com> Date: Sat, 2 Apr 2022 08:42:51 +0000 Subject: [PATCH 15/16] update TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md. --- .../SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md b/TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md index 3ba870175..c2d31f3ed 100644 --- a/TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md +++ b/TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md @@ -9,15 +9,25 @@

基本信息

**发布者（Publisher）：Huawei** + **应用领域（Application Domain）：Object Detection + **版本（Version）：1.1 + **修改时间（Modified） ：2021.07.19 + **大小（Size）：6M + **框架（Framework）：TensorFlow 1.15.0 + **模型格式（Model Format）：ckpt + **精度（Precision）：Mixed + **处理器（Processor）：昇腾910 + **应用级别（Categories）：Official + **描述（Description）：基于tensorflow实现，以Resnet50为backbone的SSD目标检测网络。

概述

-- Gitee From c8acc2d5b12bb6e43e93fb760401772460a8dc5d Mon Sep 17 00:00:00 2001 From: is_dc <1140924077@qq.com> Date: Sat, 2 Apr 2022 08:44:21 +0000 Subject: [PATCH 16/16] update TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md. --- .../SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md b/TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md index c2d31f3ed..e3c21915e 100644 --- a/TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md +++ b/TensorFlow/built-in/cv/detection/SSD-Resnet50V1-FPN_ID1463_for_TensorFlow/README.md @@ -169,7 +169,7 @@ download_all.sh nvidia_ssd bash examples/SSD320_evaluate.sh /checkpoints/ ``` -## 高级参考 +

高级参考

## 脚本和示例代码 -- Gitee