diff --git a/official/nlp/transformer/README.md b/official/nlp/transformer/README.md
index 22c6cdef300321cf57dfda00d894b76cab271e86..aba3b341c7066ce0854857f499625349c87dd622 100644
--- a/official/nlp/transformer/README.md
+++ b/official/nlp/transformer/README.md
@@ -59,15 +59,36 @@ Note that you can run the scripts based on the dataset mentioned in original pap
 
 After dataset preparation, you can start training and evaluation as follows:
 
+In Ascend environment
+
+```bash
+# run training example
+bash scripts/run_standalone_train.sh Ascend [DEVICE_ID] [EPOCH_SIZE] [GRADIENT_ACCUMULATE_STEP] [DATA_PATH]
+# EPOCH_SIZE recommend 52, GRADIENT_ACCUMULATE_STEP recommend 8 or 1
+
+# run distributed training example
+bash scripts/run_distribute_train_ascend.sh [DEVICE_NUM] [EPOCH_SIZE] [DATA_PATH] [RANK_TABLE_FILE] [CONFIG_PATH]
+# EPOCH_SIZE recommend 52
+
+# run evaluation example
+bash scripts/run_eval.sh Ascend [DEVICE_ID] [MINDRECORD_DATA] [CKPT_PATH] [CONFIG_PATH]
+# CONFIG_PATH　should be consistent with training
+```
+
+In GPU environment
+
 ```bash
 # run training example
-bash scripts/run_standalone_train_ascend.sh Ascend 0 52 /path/ende-l128-mindrecord
+bash scripts/run_standalone_train.sh GPU [DEVICE_ID] [EPOCH_SIZE] [GRADIENT_ACCUMULATE_STEP] [DATA_PATH]
+# EPOCH_SIZE recommend 52, GRADIENT_ACCUMULATE_STEP recommend 8 or 1
 
 # run distributed training example
-bash scripts/run_distribute_train_ascend.sh 8 52 /path/ende-l128-mindrecord rank_table.json ./default_config.yaml
+bash scripts/run_distribute_train_gpu.sh [DEVICE_NUM] [EPOCH_SIZE] [DATA_PATH] [CONFIG_PATH]
+# EPOCH_SIZE recommend 52
 
 # run evaluation example
-python eval.py > eval.log 2>&1 &
+bash scripts/run_eval.sh GPU [DEVICE_ID] [MINDRECORD_DATA] [CKPT_PATH] [CONFIG_PATH]
+# CONFIG_PATH　should be consistent with training
 ```
 
 - Running on [ModelArts](https://support.huaweicloud.com/modelarts/)
@@ -322,25 +343,28 @@ Parameters for learning rate:
 - Run `run_standalone_train.sh` for non-distributed training of Transformer model.
 
     ``` bash
-    bash scripts/run_standalone_train.sh DEVICE_TARGET DEVICE_ID EPOCH_SIZE GRADIENT_ACCUMULATE_STEP DATA_PATH
+    bash scripts/run_standalone_train.sh [DEVICE_TARGET] [DEVICE_ID] [EPOCH_SIZE] [GRADIENT_ACCUMULATE_STEP] [DATA_PATH]
     ```
 
 - Run `run_distribute_train_ascend.sh` for distributed training of Transformer model.
 
     ``` bash
-    bash scripts/run_distribute_train_ascend.sh DEVICE_NUM EPOCH_SIZE DATA_PATH RANK_TABLE_FILE CONFIG_PATH
+    # Ascend environment
+    bash scripts/run_distribute_train_ascend.sh [DEVICE_NUM] [EPOCH_SIZE] [DATA_PATH] [RANK_TABLE_FILE] [CONFIG_PATH]
+    # GPU environment
+    bash scripts/run_distribute_train_gpu.sh [DEVICE_NUM] [EPOCH_SIZE] [DATA_PATH] [CONFIG_PATH]
     ```
 
 **Attention**: data sink mode can not be used in transformer since the input data have different sequence lengths.
 
 ## [Evaluation Process](#contents)
 
-- Set options in `default_config.yaml`. Make sure the 'data_file', 'model_file' and 'output_file' are set to your own path.
+- Set options in [CONFIG_PATH], that should be consistent with training. Make sure the 'device_target', 'data_file', 'model_file' and 'output_file' are set to your own path.
 
 - Run `eval.py` for evaluation of Transformer model.
 
     ```bash
-    python eval.py
+    python eval.py --config_path=[CONFIG_PATH]
     ```
 
 - Run `process_output.sh` to process the output token ids to get the real translation results.
@@ -390,33 +414,33 @@ Inference result is saved in current path, 'output_file' will generate in path s
 
 #### Training Performance
 
-| Parameters                 | Ascend                                                         |
-| -------------------------- | -------------------------------------------------------------- |
-| Resource                   | Ascend 910; OS Euler2.8                                                 |
-| uploaded Date              | 07/05/2021 (month/day/year)                                    |
-| MindSpore Version          | 1.3.0                                                          |
-| Dataset                    | WMT Englis-German                                              |
-| Training Parameters        | epoch=52, batch_size=96                                        |
-| Optimizer                  | Adam                                                           |
-| Loss Function              | Softmax Cross Entropy                                          |
-| BLEU Score                 | 28.7                                                           |
-| Speed                      | 400ms/step (8pcs)                                              |
-| Loss                       | 2.8                                                            |
-| Params (M)                 | 213.7                                                          |
-| Checkpoint for inference   | 2.4G (.ckpt file)                                              |
+| Parameters                 | Ascend                                     | GPU                             |
+| -------------------------- | -------------------------------------------| --------------------------------|
+| Resource                   | Ascend 910; OS Euler2.8                    | GPU(Tesla V100 SXM2)            |
+| uploaded Date              | 07/05/2021 (month/day/year)                | 12/21/2021 (month/day/year)     |
+| MindSpore Version          | 1.3.0                                      | 1.5.0                           |
+| Dataset                    | WMT Englis-German                          | WMT Englis-German               |
+| Training Parameters        | epoch=52, batch_size=96                    | epoch=52, batch_size=96         |
+| Optimizer                  | Adam                                       | Adam                            |
+| Loss Function              | Softmax Cross Entropy                      | Softmax Cross Entropy           |
+| BLEU Score                 | 28.7                                       | 29.1                            |
+| Speed                      | 400ms/step (8pcs)                          | 337 ms/step (8pcs)              |
+| Loss                       | 2.8                                        | 2.9                             |
+| Params (M)                 | 213.7                                      | 213.7                           |
+| Checkpoint for inference   | 2.4G (.ckpt file)                          | 2.4G (.ckpt file)               |
 | Scripts                    | [Transformer scripts](https://gitee.com/mindspore/models/tree/master/official/nlp/transformer) |
 
 #### Evaluation Performance
 
-| Parameters          | Ascend                      |
-| ------------------- | --------------------------- |
-| Resource            | Ascend 910; OS Euler2.8                |
-| Uploaded Date       | 07/05/2021 (month/day/year) |
-| MindSpore Version   | 1.3.0                       |
-| Dataset             | WMT newstest2014            |
-| batch_size          | 1                           |
-| outputs             | BLEU score                  |
-| Accuracy            | BLEU=28.7                   |
+| Parameters          | Ascend                      | GPU                         |
+| ------------------- | --------------------------- | ----------------------------|
+| Resource            | Ascend 910; OS Euler2.8     | GPU(Tesla V100 SXM2)        |
+| Uploaded Date       | 07/05/2021 (month/day/year) | 12/21/2021 (month/day/year) |
+| MindSpore Version   | 1.3.0                       | 1.5.0                       |
+| Dataset             | WMT newstest2014            | WMT newstest2014            |
+| batch_size          | 1                           | 1                           |
+| outputs             | BLEU score                  | BLEU score                  |
+| Accuracy            | BLEU=28.7                   | BLEU=29.1                   |
 
 ## [Description of Random Situation](#contents)
 
diff --git a/official/nlp/transformer/README_CN.md b/official/nlp/transformer/README_CN.md
index 24b75eae4a414392cde75cf8f7767ad088246d0a..e49963b45c6b709374d33d13dea0587f8fdb4300 100644
--- a/official/nlp/transformer/README_CN.md
+++ b/official/nlp/transformer/README_CN.md
@@ -49,8 +49,8 @@ Transformer具体包括六个编码模块和六个解码模块。每个编码模
 
 ## 环境要求
 
-- 硬件（Ascend处理器）
-    - 使用Ascend处理器准备硬件环境。
+- 硬件（Ascend处理器/CPU处理器）
+    - 使用Ascend/GPu处理器准备硬件环境。
 - 框架
     - [MindSpore](https://gitee.com/mindspore/mindspore)
 - 如需查看详情，请参见如下资源：
@@ -61,15 +61,36 @@ Transformer具体包括六个编码模块和六个解码模块。每个编码模
 
 数据集准备完成后，请按照如下步骤开始训练和评估：
 
+Ascend环境下:
+
+```bash
+# 运行训练示例
+bash scripts/run_standalone_train.sh Ascend [DEVICE_ID] [EPOCH_SIZE] [GRADIENT_ACCUMULATE_STEP] [DATA_PATH]
+# EPOCH_SIZE 推荐52, GRADIENT_ACCUMULATE_STEP 推荐8或者1
+
+# 运行分布式训练示例
+bash scripts/run_distribute_train_ascend.sh [DEVICE_NUM] [EPOCH_SIZE] [DATA_PATH] [RANK_TABLE_FILE] [CONFIG_PATH]
+# EPOCH_SIZE 推荐52
+
+# 运行评估示例
+bash scripts/run_eval.sh Ascend [DEVICE_ID] [MINDRECORD_DATA] [CKPT_PATH] [CONFIG_PATH]
+# CONFIG_PATH要和训练时保持一致
+```
+
+GPU环境下:
+
 ```bash
 # 运行训练示例
-bash scripts/run_standalone_train_ascend.sh Ascend 0 52 /path/ende-l128-mindrecord
+bash scripts/run_standalone_train.sh GPU [DEVICE_ID] [EPOCH_SIZE] [GRADIENT_ACCUMULATE_STEP] [DATA_PATH]
+# EPOCH_SIZE 推荐52, GRADIENT_ACCUMULATE_STEP 推荐8或者1
 
 # 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
+bash scripts/run_distribute_train_gpu.sh [DEVICE_NUM] [EPOCH_SIZE] [DATA_PATH] [CONFIG_PATH]
+# EPOCH_SIZE 推荐52
 
 # 运行评估示例
-python eval.py > eval.log 2>&1 &
+bash scripts/run_eval.sh GPU [DEVICE_ID] [MINDRECORD_DATA] [CKPT_PATH] [CONFIG_PATH]
+# CONFIG_PATH要和训练时保持一致
 ```
 
 - 在 ModelArts 进行训练 (如果你想在modelarts上运行，可以参考以下文档 [modelarts](https://support.huaweicloud.com/modelarts/))
@@ -325,25 +346,28 @@ Parameters for learning rate:
 - 运行`run_standalone_train.sh`，进行Transformer模型的非分布式训练。
 
     ``` bash
-    bash scripts/run_standalone_train.sh DEVICE_TARGET DEVICE_ID EPOCH_SIZE GRADIENT_ACCUMULATE_STEP DATA_PATH
+    bash scripts/run_standalone_train.sh [DEVICE_TARGET] [DEVICE_ID] [EPOCH_SIZE] [GRADIENT_ACCUMULATE_STEP] [DATA_PATH]
     ```
 
 - 运行`run_distribute_train_ascend.sh`，进行Transformer模型的非分布式训练。
 
     ``` bash
-    bash scripts/run_distribute_train_ascend.sh DEVICE_NUM EPOCH_SIZE DATA_PATH RANK_TABLE_FILE CONFIG_PATH
+    # Ascend environment
+    bash scripts/run_distribute_train_ascend.sh [DEVICE_NUM] [EPOCH_SIZE] [DATA_PATH] [RANK_TABLE_FILE] [CONFIG_PATH]
+    # GPU environment
+    bash scripts/run_distribute_train_gpu.sh [DEVICE_NUM] [EPOCH_SIZE] [DATA_PATH] [CONFIG_PATH]
     ```
 
 **注意**：由于网络输入中有不同句长的数据，所以数据下沉模式不可使用。
 
 ### 评估过程
 
-- 在`default_config.yaml`中设置选项。确保已设置了‘data_file'、'model_file’和'output_file'文件路径。
+- 在[CONFIG_PATH]中设置选项，此时的[CONFIG_PATH]要和训练时保持一致。确保已设置了'device_target', 'data_file'、'model_file'和'output_file'文件路径。
 
 - 运行`eval.py`，评估Transformer模型。
 
     ```bash
-    python eval.py
+    python eval.py --config_path=[CONFIG_PATH]
     ```
 
 - 运行`process_output.sh`，处理输出标记ids，获得真实翻译结果。
@@ -393,33 +417,33 @@ bash run_infer_310.sh [MINDIR_PATH] [NEED_PREPROCESS] [DEVICE_ID]
 
 #### 训练性能
 
-| 参数                | Ascend                                                    |
-| -------------------------- | -------------------------------------------------------------- |
-| 资源                  | Ascend 910；系统 Euler2.8                                                   |
-| 上传日期              | 2021-07-05                                    |
-| MindSpore版本          | 1.3.0                                                     |
-| 数据集                    | WMT英-德翻译数据集                                              |
-| 训练参数        | epoch=52, batch_size=96                                        |
-| 优化器                 | Adam                                                           |
-| 损失函数              | Softmax Cross Entropy                                          |
-| BLEU分数                 | 28.7                                                           |
-| 速度                      | 400毫秒/步(8卡)                                              |
-| 损失                       | 2.8                                                            |
-| 参数 (M)                 | 213.7                                                          |
-| 推理检查点   | 2.4G （.ckpt文件）                                              |
-| 脚本                    | <https://gitee.com/mindspore/models/tree/master/official/nlp/transformer> |
+| 参数                        | Ascend                           | GPU                             |
+| -------------------------- | -------------------------------- | --------------------------------|
+| 资源                        | Ascend 910；系统 Euler2.8         | GPU(Tesla V100 SXM2)            |
+| 上传日期                    | 2021-07-05                        | 2021-12-21                      |
+| MindSpore版本               | 1.3.0                            | 1.5.0                           |
+| 数据集                      | WMT英-德翻译数据集                  | WMT英-德翻译数据集                |
+| 训练参数                     | epoch=52, batch_size=96          | epoch=52, batch_size=96         |
+| 优化器                      | Adam                              | Adam                            |
+| 损失函数                     | Softmax Cross Entropy            | Softmax Cross Entropy           |
+| BLEU分数                    | 28.7                              | 29.1                           |
+| 速度                        | 400毫秒/步(8卡)                    | 337 ms/step(8卡)                |
+| 损失                        | 2.8                               | 2.9                            |
+| 参数 (M)                    | 213.7                             | 213.7                          |
+| 推理检查点                   | 2.4G （.ckpt文件）                 | 2.4G                            |
+| 脚本                        | <https://gitee.com/mindspore/models/tree/master/official/nlp/transformer> |
 
 #### 评估性能
 
-| 参数          | Ascend                   |
-| ------------------- | --------------------------- |
-|资源| Ascend 910；系统 Euler2.8  |
-| 上传日期       | 2021-07-05 |
-| MindSpore版本   | 1.3.0                  |
-| 数据集             | WMT newstest2014            |
-| batch_size          | 1                           |
-| 输出             | BLEU score                  |
-| 准确率            | BLEU=28.7                   |
+| 参数          | Ascend                   |                   GPU                 |
+| ------------------- | --------------------------- | ----------------------------|
+|资源| Ascend 910；系统 Euler2.8  |                GPU(Tesla V100 SXM2)         |
+| 上传日期       | 2021-07-05 |              2021-12-21                        |
+| MindSpore版本   | 1.3.0                  |        1.5.0                      |
+| 数据集             | WMT newstest2014            | WMT newstest2014            |
+| batch_size          | 1                           | 1                           |
+| 输出             | BLEU score                  | BLEU score                  |
+| 准确率            | BLEU=28.7                   | BLEU=29.1                   |
 
 ## 随机情况说明
 
diff --git a/official/nlp/transformer/scripts/run_distribute_train_ascend.sh b/official/nlp/transformer/scripts/run_distribute_train_ascend.sh
index 41be7cdef414660863d77b4eb8f34aefee89bf53..3c6f5b0135e6f62091a69b9bcd5b69ded21c3c35 100644
--- a/official/nlp/transformer/scripts/run_distribute_train_ascend.sh
+++ b/official/nlp/transformer/scripts/run_distribute_train_ascend.sh
@@ -16,7 +16,7 @@
 if [ $# != 5 ] ; then
 echo "=============================================================================================================="
 echo "Please run the script as: "
-echo "sh run_distribute_train_ascend.sh DEVICE_NUM EPOCH_SIZE DATA_PATH RANK_TABLE_FILE CONFIG_PATH"
+echo "bash scripts/run_distribute_train_ascend.sh DEVICE_NUM EPOCH_SIZE DATA_PATH RANK_TABLE_FILE CONFIG_PATH"
 echo "for example: sh run_distribute_train_ascend.sh 8 52 /path/ende-l128-mindrecord00 /path/hccl.json ./default_config_large.yaml"
 echo "It is better to use absolute path."
 echo "=============================================================================================================="
diff --git a/official/nlp/transformer/scripts/run_distribute_train_ascend_multi_machines.sh b/official/nlp/transformer/scripts/run_distribute_train_ascend_multi_machines.sh
index 14fd9d12d6bea02a4e0ee5eef445577ea3cc2bbe..6577f0d9894fb2b22171cb68614fcfa6aee2e546 100644
--- a/official/nlp/transformer/scripts/run_distribute_train_ascend_multi_machines.sh
+++ b/official/nlp/transformer/scripts/run_distribute_train_ascend_multi_machines.sh
@@ -16,7 +16,7 @@
 if [ $# != 6 ] ; then
 echo "=============================================================================================================="
 echo "Please run the script as: "
-echo "sh run_distribute_train_ascend_multi_machines.sh DEVICE_NUM SERVER_ID EPOCH_SIZE DATA_PATH RANK_TABLE_FILE CONFIG_PATH"
+echo "bash scripts/run_distribute_train_ascend_multi_machines.sh DEVICE_NUM SERVER_ID EPOCH_SIZE DATA_PATH RANK_TABLE_FILE CONFIG_PATH"
 echo "for example: sh run_distribute_train_ascend_multi_machines.sh 32 0 52 /path/ende-l128-mindrecord00 /path/hccl.json ./default_config_large.yaml"
 echo "It is better to use absolute path."
 echo "=============================================================================================================="
diff --git a/official/nlp/transformer/scripts/run_distribute_train_gpu.sh b/official/nlp/transformer/scripts/run_distribute_train_gpu.sh
index a7a884a78d14cf2e5babd215338f2ede6d4eb2e7..e878616ae72595efabaca5db6fd9eb545cd1b306 100644
--- a/official/nlp/transformer/scripts/run_distribute_train_gpu.sh
+++ b/official/nlp/transformer/scripts/run_distribute_train_gpu.sh
@@ -16,7 +16,7 @@
 if [ $# != 4 ] ; then
 echo "=============================================================================================================="
 echo "Please run the script as: "
-echo "sh run_distribute_train_gpu.sh DEVICE_NUM EPOCH_SIZE DATA_PATH CONFIG_PATH"
+echo "bash scripts/run_distribute_train_gpu.sh DEVICE_NUM EPOCH_SIZE DATA_PATH CONFIG_PATH"
 echo "for example: sh run_distribute_train_gpu.sh 8 55 /path/ende-l128-mindrecord00 ./default_config_large_gpu.yaml"
 echo "It is better to use absolute path."
 echo "=============================================================================================================="
diff --git a/official/nlp/transformer/scripts/run_eval.sh b/official/nlp/transformer/scripts/run_eval.sh
index 8a2d2a1fe2a823bfd38927bab20966af380396c5..9628e782092804f5cb6d1be03dfb0c1c234a1533 100644
--- a/official/nlp/transformer/scripts/run_eval.sh
+++ b/official/nlp/transformer/scripts/run_eval.sh
@@ -16,7 +16,7 @@
 if [ $# != 5 ] ; then
 echo "=============================================================================================================="
 echo "Please run the script as: "
-echo "sh run_eval.sh DEVICE_TARGET DEVICE_ID MINDRECORD_DATA CKPT_PATH CONFIG_PATH"
+echo "sh scripts/run_eval.sh DEVICE_TARGET DEVICE_ID MINDRECORD_DATA CKPT_PATH CONFIG_PATH"
 echo "for example: sh run_eval.sh Ascend 0 /your/path/evaluation.mindrecord /your/path/checkpoint_file ./default_config_large_gpu.yaml"
 echo "Note: set the checkpoint and dataset path in default_config.yaml"
 echo "=============================================================================================================="
diff --git a/official/nlp/transformer/scripts/run_standalone_train.sh b/official/nlp/transformer/scripts/run_standalone_train.sh
index 50a7779ee8fb6487ecca0a0035120d634c1ece7a..3250dff510b0bcd2afb7b6e8a6a79b4515bcae9c 100644
--- a/official/nlp/transformer/scripts/run_standalone_train.sh
+++ b/official/nlp/transformer/scripts/run_standalone_train.sh
@@ -16,7 +16,7 @@
 if [ $# != 5 ] ; then
 echo "=============================================================================================================="
 echo "Please run the script as: "
-echo "sh run_standalone_train.sh DEVICE_TARGET DEVICE_ID EPOCH_SIZE GRADIENT_ACCUMULATE_STEP DATA_PATH"
+echo "sh scripts/run_standalone_train.sh DEVICE_TARGET DEVICE_ID EPOCH_SIZE GRADIENT_ACCUMULATE_STEP DATA_PATH"
 echo "for example: sh run_standalone_train.sh Ascend 0 52 8 /path/ende-l128-mindrecord00"
 echo "It is better to use absolute path."
 echo "=============================================================================================================="