!103 modify_Readme

Merge pull request !103 from 欧士博/dev

official/nlp/bert/README.md

-# Contents
+# Contents
 
 [查看中文](./README_CN.md)
 
@@ -588,7 +588,15 @@ epoch: 0.0, current epoch percent: 0.002, step: 200, outputs are (Tensor(shape=[
 
 #### evaluation on cola dataset when running on Ascend
 
-Before running the command below, please check the load pretrain checkpoint path has been set. Please set the checkpoint path to be the absolute full path, e.g:"/username/pretrain/checkpoint_100_300.ckpt".
+Before running the command below, please check the load pretrain checkpoint path has been set. Please set the checkpoint path to be the absolute full path, e.g:
+
+--load_pretrain_checkpoint_path="/data/scripts/checkpoint_bert-20000_1.ckpt" \
+
+--train_data_file_path="/data/tnews/train.tf_record" \
+
+--eval_data_file_path="/data/tnews/dev.tf_record" \
+
+--schema_file_path="/data/tnews/dataset.json"
 
 ```bash
 bash scripts/run_classifier.sh
@@ -604,6 +612,18 @@ acc_num XXX, total_num XXX, accuracy 0.588986
 
 #### evaluation on cluener dataset when running on Ascend
 
+Before running the command below, please check the load pretrain checkpoint path has been set. Please set the checkpoint path to be the absolute full path, e.g:
+
+--label_file_path="/data/cluener/label_file" \
+
+--load_pretrain_checkpoint_path="/data/scripts/checkpoint_bert-20000_1.ckpt" \
+
+--train_data_file_path="/data/cluener/train.tf_record" \
+
+--eval_data_file_path="/data/cluener/dev.tf_record" \
+
+--schema_file_path="/data/cluener/dataset.json"
+
 ```bash
 bash scripts/run_ner.sh
 ```
@@ -642,6 +662,16 @@ F1 0.931243
 
 #### evaluation on squad v1.1 dataset when running on Ascend
 
+Before running the command below, please check the load pretrain checkpoint path has been set. Please set the checkpoint path to be the absolute full path, e.g:
+
+--vocab_file_path="/data/squad/vocab_bert_large_en.txt" \
+
+--load_pretrain_checkpoint_path="/data/scripts/bert_converted.ckpt" \
+
+--train_data_file_path="/data/squad/train.tf_record" \
+
+--eval_json_path="/data/squad/dev-v1.1.json" \
+
 ```bash
 bash scripts/squad.sh
 ```

official/nlp/bert/README_CN.md

-
-# 目录
+# 目录
 
 [View English](./README.md)
 
@@ -86,13 +85,17 @@ BERT的主干结构为Transformer。对于BERT_base，Transformer包含12个编
 - 在Ascend上运行
 
 ```bash
+
 # 单机运行预训练示例
+
 bash scripts/run_standalone_pretrain_ascend.sh 0 1 /path/cn-wiki-128
 
 # 分布式运行预训练示例
+
 bash scripts/run_distributed_pretrain_ascend.sh /path/cn-wiki-128 /path/hccl.json
 
 # 运行微调和评估示例
+
 - 如需运行微调任务，请先准备预训练生成的权重文件（ckpt）。
 - 在`finetune_eval_config.py`中设置BERT网络配置和优化器超参。
 
@@ -115,13 +118,17 @@ bash scripts/run_distributed_pretrain_ascend.sh /path/cn-wiki-128 /path/hccl.jso
 - 在GPU上运行
 
 ```bash
+
 # 单机运行预训练示例
+
 bash run_standalone_pretrain_for_gpu.sh 0 1 /path/cn-wiki-128
 
 # 分布式运行预训练示例
+
 bash scripts/run_distributed_pretrain_for_gpu.sh 8 40 /path/cn-wiki-128
 
 # 运行微调和评估示例
+
 - 如需运行微调任务，请先准备预训练生成的权重文件（ckpt）。
 - 在`finetune_eval_config.py`中设置BERT网络配置和优化器超参。
 
@@ -503,7 +510,9 @@ bash scripts/run_standalone_pretrain_ascend.sh 0 1 /path/cn-wiki-128
 以上命令后台运行，您可以在pretraining_log.txt中查看训练日志。训练结束后，您可以在默认脚本路径下脚本文件夹中找到检查点文件，得到如下损失值：
 
 ```text
+
 # grep "epoch" pretraining_log.txt
+
 epoch: 0.0, current epoch percent: 0.000, step: 1, outputs are (Tensor(shape=[1], dtype=Float32, [ 1.0856101e+01]), Tensor(shape=[], dtype=Bool, False), Tensor(shape=[], dtype=Float32, 65536))
 epoch: 0.0, current epoch percent: 0.000, step: 2, outputs are (Tensor(shape=[1], dtype=Float32, [ 1.0821701e+01]), Tensor(shape=[], dtype=Bool, False), Tensor(shape=[], dtype=Float32, 65536))
 ...
@@ -533,7 +542,9 @@ bash scripts/run_distributed_pretrain_ascend.sh /path/cn-wiki-128 /path/hccl.jso
 以上命令后台运行，您可以在pretraining_log.txt中查看训练日志。训练结束后，您可以在默认LOG*文件夹下找到检查点文件，得到如下损失值：
 
 ```text
+
 # grep "epoch" LOG*/pretraining_log.txt
+
 epoch: 0.0, current epoch percent: 0.001, step: 100, outputs are (Tensor(shape=[1], dtype=Float32, [ 1.08209e+01]), Tensor(shape=[], dtype=Bool, False), Tensor(shape=[], dtype=Float32, 65536))
 epoch: 0.0, current epoch percent: 0.002, step: 200, outputs are (Tensor(shape=[1], dtype=Float32, [ 1.07566e+01]), Tensor(shape=[], dtype=Bool, False), Tensor(shape=[], dtype=Float32, 65536))
 ...
@@ -550,7 +561,15 @@ epoch: 0.0, current epoch percent: 0.002, step: 200, outputs are (Tensor(shape=[
 
 #### Ascend处理器上运行后评估cola数据集
 
-运行以下命令前，确保已设置加载与训练检查点路径。请将检查点路径设置为绝对全路径，例如，/username/pretrain/checkpoint_100_300.ckpt。
+运行以下命令前，确保已设置加载与训练检查点路径。请将检查点路径设置为绝对全路径，例如，
+
+--load_pretrain_checkpoint_path="/data/scripts/checkpoint_bert-20000_1.ckpt" \
+
+--train_data_file_path="/data/tnews/train.tf_record" \
+
+--eval_data_file_path="/data/tnews/dev.tf_record" \
+
+--schema_file_path="/data/tnews/dataset.json"
 
 ```bash
 bash scripts/run_classifier.sh
@@ -566,6 +585,18 @@ acc_num XXX, total_num XXX, accuracy 0.588986
 
 #### Ascend处理器上运行后评估cluener数据集
 
+运行以下命令前，确保已设置加载与训练检查点路径。请将检查点路径设置为绝对全路径，例如，
+
+--label_file_path="/data/finetune/cluener/label_file" \
+
+--load_pretrain_checkpoint_path="/data/scripts/checkpoint_bert-20000_1.ckpt" \
+
+--train_data_file_path="/data/cluener/train.tf_record" \
+
+--eval_data_file_path="/data/cluener/dev.tf_record" \
+
+--schema_file_path="/data/cluener/dataset.json"
+
 ```bash
 bash scripts/run_ner.sh
 ```
@@ -603,6 +634,16 @@ F1 0.931243
 
 #### Ascend处理器上运行后评估squad v1.1数据集
 
+运行以下命令前，确保已设置加载与训练检查点路径。请将检查点路径设置为绝对全路径，例如，
+
+--vocab_file_path="/data/squad/vocab_bert_large_en.txt" \
+
+--load_pretrain_checkpoint_path="/data/scripts/bert_converted.ckpt" \
+
+--train_data_file_path="/data/squad/train.tf_record" \
+
+--eval_json_path="/data/squad/dev-v1.1.json" \
+
 ```bash
 bash scripts/squad.sh
 ```
@@ -627,6 +668,7 @@ python export.py --config_path [../../*.yaml] --export_ckpt_file [CKPT_PATH] --e
 - 在ModelArts上导出
 
 ```python
+
 # (1) 上传你的代码到 s3 桶上
 # (2) 在ModelArts上创建训练任务
 # (3) 选择代码目录 /{path}/bert
@@ -650,6 +692,7 @@ python export.py --config_path [../../*.yaml] --export_ckpt_file [CKPT_PATH] --e
 # (9) 在网页上的’资源池选择‘项目下， 选择单卡规格的资源
 # (10) 创建训练作业
 # 你将在{Output file path}下看到 'bert_ner.mindir'文件
+
 ```
 
 参数`export_ckpt_file` 是必需的，`file_format` 必须在 ["AIR", "MINDIR"]中进行选择。
@@ -661,7 +704,9 @@ python export.py --config_path [../../*.yaml] --export_ckpt_file [CKPT_PATH] --e
 在执行推理之前，需要通过export.py导出mindir文件。输入数据文件为bin格式。
 
 ```shell
+
 # Ascend310 推理
+
 bash run_infer_310.sh [MINDIR_PATH] [LABEL_PATH] [DATA_FILE_PATH] [DATASET_FORMAT] [SCHEMA_PATH] [USE_CRF] [NEED_PREPROCESS] [DEVICE_ID]
 ```
 

official/nlp/emotect/README_CN.md

-
-# 目录
+# 目录
 
 - [目录](#目录)
 - [概述](#概述)
@@ -151,9 +150,13 @@ bash script/download_data.sh
 
 ```bash
 bash scripts/convert_dataset.sh
+
 # `convert_dataset.sh` depend on ERNIE vocabulary,
+
 # you should download ERNIE model first by:
+
 # sh script/download_model.sh
+
 ```
 
 #### Ascend处理器或GPU上运行
@@ -167,16 +170,22 @@ bash script/download_model.sh
 预训练模型ERNIE下载后，将其转换为MindSpore可加载权重
 
 ```shell
+
 #--input_dir ./pretrain_models/ernie
+
 bash script/paddle_to_midnspore.sh
+
 # only support x86 platform since Paddle don't support ARM
+
 ```
 
 将ERNIE迁移至Mindspore后，执行训练脚本:
 
 ```bash
 bash scripts/run_classifier_finetune_{platform}.sh
+
 # platform: gpu or ascend
+
 ```
 
 模型保存在 ```./save_models```。
@@ -187,11 +196,19 @@ bash scripts/run_classifier_finetune_{platform}.sh
 
 #### Ascend处理器或GPU上运行后评估
 
-根据训练结果，可选择最优的step进行评估，修改```scripts/run_classifier_eval.sh``` 脚本中```load_finetune_checkpoint_path``` 参数，然后执行
+根据训练结果，可选择最优的step进行评估，修改```scripts/run_classifier_eval.sh``` 脚本中参数，例如，
+
+--load_pretrain_checkpoint_path="/data/pretrain_models/ernie.ckpt" \
+
+--train_data_file_path="/emotect_dataset/data/train.mindrecord" \
+
+--eval_data_file_path="/emotect_dataset/data/dev.mindrecord" \
 
 ```shell
 bash scripts/run_classifier_eval_{platform}.sh
+
 # platform: gpu or ascend
+
 ```
 
 ## 310推理
@@ -200,7 +217,9 @@ bash scripts/run_classifier_eval_{platform}.sh
 
 ```shell
 bash scripts/export.sh
+
 # export finetune ckpt to mindir
+
 ```
 
 参数`ckpt_file`，`file_format`已在`export.sh`中设置。
@@ -210,7 +229,9 @@ bash scripts/export.sh
 以下展示了使用minir模型执行推理的示例。
 
 ```shell
+
 # Ascend310推理
+
 bash scripts/run_infer_310.sh [MINDIR_PATH] [DATA_FILE_PATH] [NEED_PREPROCESS] [DEVICE_ID]
 ```
 

official/nlp/gru/README.md

-![](https://www.mindspore.cn/static/img/logo_black.6a5c850d.png)
+![](https://www.mindspore.cn/static/img/logo_black.6a5c850d.png)
 
 <!-- TOC -->
 
@@ -278,6 +278,8 @@ Parameters for both training and evaluation can be set in config.py. All the dat
     cd ./scripts
     bash run_standalone_train_{platform}.sh [DATASET_PATH]
     # platform: ascend or gpu
+    #example：
+    bash run_standalone_train_ascend.sh /Muti30k/mindrecord/multi30k_train_mindrecord_32_0
     ```
 
 - Running scripts for distributed training of GRU. Task training on multiple device and run the following command in bash to be executed in `scripts/`:
@@ -297,6 +299,8 @@ Parameters for both training and evaluation can be set in config.py. All the dat
     cd ./scripts
     bash run_eval_{platform}.sh [CKPT_FILE] [DATASET_PATH]
     # platform: ascend or gpu
+    #example:
+    bash run_eval_ascend.sh /data/ckpt_0/0-20_1807.ckpt /data/mindrecord/multi30k_test_mindrecord_32
     ```
 
 - After evalulation, we will get eval/target.txt and eval/output.txt.Then we can use scripts/parse_output.sh to get the translation.

official/nlp/gru/requirements.txt

+nltk
+numpy
\ No newline at end of file

official/nlp/pangu_alpha/README.md

-# Contents
+# Contents
 
 - [Contents](#contents)
 - [PanGu-Alpha Description](#pangu-alpha-description)
@@ -137,6 +137,8 @@ Currently the scripts provide three default configures : `2.6B` `13B` and `200B`
 # run distributed training example
 
 bash scripts/run_distribute_train.sh DATASET RANK_TABLE RANK_SIZE TYPE MODE STAGE_NUM MICRO_SIZE PER_BATCH RANK_START
+#example:
+bash scripts/run_distribute_train.sh /data/pangu_30_step_ba64/ /root/hccl_8p.json 8 fp32 2.6B 1 1 8 0 8
 
 ```
 

official/nlp/tinybert/README.md

@@ -494,6 +494,8 @@ The best acc is 0.891176
 
 ```shell
 python export.py --ckpt_file [CKPT_PATH] --file_name [FILE_NAME] --file_format [FILE_FORMAT]
+#example:
+python export.py --ckpt_file ./2021-09-03_time_16_00_12/tiny_bert_936_100.ckpt --file_name SST-2 --file_format MINDIR --config_path ./td_config/td_config_sst2.yaml
 ```
 
 The ckpt_file parameter is required,

official/nlp/tinybert/README_CN.md

@@ -493,6 +493,8 @@ The best acc is 0.891176
 
     ```shell
     python export.py --ckpt_file [CKPT_PATH] --file_name [FILE_NAME] --file_format [FILE_FORMAT]
+    #example:
+    python export.py --ckpt_file ./2021-09-03_time_16_00_12/tiny_bert_936_100.ckpt --file_name SST-2 --file_format MINDIR --config_path ./td_config/td_config_sst2.yaml
     ```
 
 - 在ModelArts上导出

official/nlp/transformer/README.md

@@ -61,7 +61,7 @@ After dataset preparation, you can start training and evaluation as follows:
 
 ```bash
 # run training example
-bash scripts/run_standalone_train_ascend.sh 0 52 /path/ende-l128-mindrecord
+bash scripts/run_standalone_train_ascend.sh Ascend 0 52 /path/ende-l128-mindrecord
 
 # run distributed training example
 bash scripts/run_distribute_train_ascend.sh 8 52 /path/ende-l128-mindrecord rank_table.json ./default_config.yaml

official/nlp/transformer/README_CN.md

@@ -63,7 +63,7 @@ Transformer具体包括六个编码模块和六个解码模块。每个编码模
 
 ```bash
 # 运行训练示例
-bash scripts/run_standalone_train_ascend.sh 0 52 /path/ende-l128-mindrecord
+bash scripts/run_standalone_train_ascend.sh Ascend 0 52 /path/ende-l128-mindrecord
 
 # 运行分布式训练示例
 bash scripts/run_distribute_train_ascend.sh 8 52 /path/ende-l128-mindrecord rank_table.json ./default_config.yaml

official/nlp/transformer/requirements.txt

+numpy
+easydict
+pyyaml

official/recommend/deep_and_cross/README_CN.md

-# 目录
+# 目录
 
 <!-- TOC -->
 
@@ -78,14 +78,14 @@ GPU单卡训练命令如下：
 #单卡训练示例
 python train.py --device_target="GPU" > output.train.log 2>&1 &
 #或
-sh scripts/run_train_gpu.sh
+bash scripts/run_train_gpu.sh
 ```
 
 GPU 8卡训练命令如下：
 
 ```bash
 #8卡训练示例
-sh scripts/run_train_multi_gpu.sh
+bash scripts/run_train_multi_gpu.sh
 ```
 
 5. 开始验证。
@@ -95,7 +95,7 @@ sh scripts/run_train_multi_gpu.sh
 ```bash
 python eval.py --ckpt_path=CHECKPOINT_PATH
 #或
-sh scripts/run_eval.sh CHECKPOINT_PATH
+bash scripts/run_eval.sh CHECKPOINT_PATH
 ```
 
 # 脚本说明
@@ -164,7 +164,7 @@ sh scripts/run_eval.sh CHECKPOINT_PATH
 - GPU处理器环境运行
 
   ```bash
-  sh scripts/run_train_multi_gpu.sh
+  bash scripts/run_train_multi_gpu.sh
   ```
 
   上述shell脚本将在后台运行分布训练。您可以通过output.multi_gpu.train.log文件查看结果。
@@ -186,7 +186,7 @@ sh scripts/run_eval.sh CHECKPOINT_PATH
   或者，
 
   ```bash
-  sh scripts/run_eval.sh [CHECKPOINT_PATH]
+  bash scripts/run_eval.sh [CHECKPOINT_PATH]
   ```
 
   上述python命令将在后台运行，您可以通过output.eval.log文件查看结果。

official/recommend/deep_and_cross/requirements.txt

+numpy
+scikit-learn
\ No newline at end of file

official/recommend/wide_and_deep/README.md

@@ -92,7 +92,7 @@ python train_and_eval.py --data_path=./data/mindrecord --dataset_type=mindrecord
 To evaluate the model, command as follows:
 
 ```bash
-python eval.py  --data_path=./data/mindrecord --dataset_type=mindrecord
+python eval.py  --data_path=./data/mindrecord --dataset_type=mindrecord --ckpt_path=./ckpt/widedeep_train-15_2582.ckpt
 ```
 
 - Running on ModelArts (If you want to run in modelarts, please check the official documentation of [modelarts](https://support.huaweicloud.com/modelarts/), and you can start training as follows)

official/recommend/wide_and_deep/README_CN.md

-# 目录
+# 目录
 
 - [目录](#目录)
 - [Wide&Deep概述](#widedeep概述)
@@ -89,13 +89,13 @@ python src/preprocess_data.py  --data_path=./data/ --dense_dim=13 --slot_dim=26
 数据集准备就绪后，即可在Ascend上单机训练和评估模型。
 
 ```bash
-python train_and_eval.py --data_path=./data/mindrecord --data_type=mindrecord
+python train_and_eval.py --data_path=./data/mindrecord --dataset_type=mindrecord
 ```
 
 按如下操作评估模型：
 
 ```bash
-python eval.py  --data_path=./data/mindrecord --data_type=mindrecord
+python eval.py  --data_path=./data/mindrecord --dataset_type=mindrecord --ckpt_path=./ckpt/widedeep_train-15_2582.ckpt
 ```
 
 - 在ModelArts上运行（如果想在modelarts中运行，请查看【modelarts】官方文档（https://support.huaweicloud.com/modelarts/），如下开始训练即可）

official/recommend/wide_and_deep_multitable/README.md

@@ -66,7 +66,7 @@ python train_and_eval.py --data_path=./data/mindrecord --data_type=mindrecord
 To evaluate the model, command as follows:
 
 ```bash
-python eval.py  --data_path=./data/mindrecord --data_type=mindrecord
+python eval.py  --data_path=./data/mindrecord --data_type=mindrecord --ckpt_path=widedeep_train-8_455.ckpt
 ```
 
 ## [Script Description](#contents)

official/recommend/wide_and_deep_multitable/README_CN.md

@@ -67,7 +67,7 @@ python train_and_eval.py --data_path=./data/mindrecord --data_type=mindrecord
 运行如下命令评估模型：
 
 ```bash
-python eval.py  --data_path=./data/mindrecord --data_type=mindrecord
+python eval.py  --data_path=./data/mindrecord --data_type=mindrecord --ckpt_path=widedeep_train-8_455.ckpt
 ```
 
 ## 脚本说明