Nigel
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

!17593 update readme 

From: @huchunmei
Reviewed-by: @wuxuejian,@c_34
Signed-off-by: @wuxuejian
This commit is contained in:
mindspore-ci-bot 2021-06-03 15:01:04 +08:00 committed by Gitee
parent 730e1d3b65 40d9b537cb
commit 44dd7d994b
14 changed files with 895 additions and 93 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

76
model_zoo/official/cv/alexnet/README.md
View File

@ -114,10 +114,10 @@ Major parameters in train.py and config.py as follows:
#### Training
- running on Ascend
- Running on Ascend
```bash
python train.py --data_path cifar-10-batches-bin --ckpt_path ckpt > log 2>&1 &
python train.py --config_path default_config.yaml --data_path cifar-10-batches-bin --ckpt_path ckpt > log 2>&1 &
# or enter script dir, and run the script
sh run_standalone_train_ascend.sh cifar-10-batches-bin ckpt
```
@ -139,7 +139,7 @@ Major parameters in train.py and config.py as follows:
- running on GPU
```bash
python train.py --device_target "GPU" --data_path cifar-10-batches-bin --ckpt_path ckpt > log 2>&1 &
python train.py --config_path default_config.yaml --device_target "GPU" --data_path cifar-10-batches-bin --ckpt_path ckpt > log 2>&1 &
# or enter script dir, and run the script
sh run_standalone_train_for_gpu.sh cifar-10-batches-bin ckpt
```
@ -164,7 +164,7 @@ Before running the command below, please check the checkpoint path used for eval
- running on Ascend
```bash
python eval.py --data_path cifar-10-verify-bin --ckpt_path ckpt/checkpoint_alexnet-1_1562.ckpt > eval_log.txt 2>&1 &
python eval.py --config_path default_config.yaml --data_path cifar-10-verify-bin --ckpt_path ckpt/checkpoint_alexnet-1_1562.ckpt > eval_log.txt 2>&1 &
# or enter script dir, and run the script
sh run_standalone_eval_ascend.sh cifar-10-verify-bin ckpt/checkpoint_alexnet-1_1562.ckpt
```
@ -179,7 +179,7 @@ Before running the command below, please check the checkpoint path used for eval
- running on GPU
```bash
python eval.py --device_target "GPU" --data_path cifar-10-verify-bin --ckpt_path ckpt/checkpoint_alexnet-30_1562.ckpt > eval_log 2>&1 &
python eval.py --config_path default_config.yaml --device_target "GPU" --data_path cifar-10-verify-bin --ckpt_path ckpt/checkpoint_alexnet-30_1562.ckpt > eval_log 2>&1 &
# or enter script dir, and run the script
sh run_standalone_eval_for_gpu.sh cifar-10-verify-bin ckpt/checkpoint_alexnet-30_1562.ckpt
```
@ -196,7 +196,7 @@ Before running the command below, please check the checkpoint path used for eval
### [Export MindIR](#contents)
```shell
python export.py --ckpt_file [CKPT_PATH] --file_name [FILE_NAME] --file_format [FILE_FORMAT]
python export.py --config_path [CONFIG_PATH] --ckpt_file [CKPT_PATH] --file_name [FILE_NAME] --file_format [FILE_FORMAT]
```
The ckpt_file parameter is required,
@ -225,6 +225,70 @@ Inference result is saved in current path, you can find result like this in acc.
'acc': 0.88772
```
- Running on [ModelArts](https://support.huaweicloud.com/modelarts/)
```bash
# Train 8p with Ascend
# (1) Perform a or b.
# a. Set "enable_modelarts=True" on default_config.yaml file.
# Set "distribute=True" on default_config.yaml file.
# Set "data_path='/cache/data'" on default_config.yaml file.
# Set "ckpt_path='/cache/train'" on default_config.yaml file.
# (optional)Set "checkpoint_url='s3://dir_to_your_pretrained/'" on default_config.yaml file.
# Set other parameters on default_config.yaml file you need.
# b. Add "enable_modelarts=True" on the website UI interface.
# Add "distribute=True" on the website UI interface.
# Add "data_path=/cache/data" on the website UI interface.
# Add "ckpt_path=/cache/train" on the website UI interface.
# (optional)Add "checkpoint_url='s3://dir_to_your_pretrained/'" on the website UI interface.
# Add other parameters on the website UI interface.
# (2) Prepare model code
# (3) Upload or copy your pretrained model to S3 bucket if you want to finetune.
# (4) Upload the original cifar10 dataset to S3 bucket.
# (5) Set the code directory to "/path/alexnet" on the website UI interface.
# (6) Set the startup file to "train.py" on the website UI interface.
# (7) Set the "Dataset path" and "Output file path" and "Job log path" to your path on the website UI interface.
# (8) Create your job.
#
# Train 1p with Ascend
# (1) Perform a or b.
# a. Set "enable_modelarts=True" on default_config.yaml file.
# Set "data_path='/cache/data'" on default_config.yaml file.
# Set "ckpt_path='/cache/train'" on default_config.yaml file.
# (optional)Set "checkpoint_url='s3://dir_to_your_pretrained/'" on default_config.yaml file.
# Set other parameters on default_config.yaml file you need.
# b. Add "enable_modelarts=True" on the website UI interface.
# Add "data_path=/cache/data" on the website UI interface.
# Add "ckpt_path=/cache/train" on the website UI interface.
# (optional)Add "checkpoint_url='s3://dir_to_your_pretrained/'" on the website UI interface.
# Add other parameters on the website UI interface.
# (2) Prepare model code
# (3) Upload or copy your pretrained model to S3 bucket if you want to finetune.
# (4) Upload the original cifar10 dataset to S3 bucket.
# (5) Set the code directory to "/path/alexnet" on the website UI interface.
# (6) Set the startup file to "train.py" on the website UI interface.
# (7) Set the "Dataset path" and "Output file path" and "Job log path" to your path on the website UI interface.
# (8) Create your job.
#
# Eval 1p with Ascend
# (1) Perform a or b.
# a. Set "enable_modelarts=True" on default_config.yaml file.
# Set "data_path='/cache/data'" on default_config.yaml file.
# Set "ckpt_file='/cache/train/checkpoint_alexnet-30_1562.ckpt'" on default_config.yaml file.
# Set other parameters on default_config.yaml file you need.
# b. Add "enable_modelarts=True" on the website UI interface.
# Add "data_path=/cache/data" on the website UI interface.
# Add "ckpt_file=/cache/train/checkpoint_alexnet-30_1562.ckpt" on the website UI interface.
# Add other parameters on the website UI interface.
# (2) Prepare model code
# (3) Upload or copy your trained model to S3 bucket.
# (4) Upload the original cifar10 dataset to S3 bucket.
# (5) Set the code directory to "/path/alexnet" on the website UI interface.
# (6) Set the startup file to "eval.py" on the website UI interface.
# (7) Set the "Dataset path" and "Output file path" and "Job log path" to your path on the website UI interface.
# (8) Create your job.
```
## [Model Description](#contents)
### [Performance](#contents)

74
model_zoo/official/cv/alexnet/README_CN.md
View File

@ -71,6 +71,70 @@ sh run_standalone_train_ascend.sh [DATA_PATH] [CKPT_SAVE_PATH]
sh run_standalone_eval_ascend.sh [DATA_PATH] [CKPT_NAME]
```
- 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/))
```bash
# 在 ModelArts 上使用8卡训练
# (1) 执行a或者b
# a. 在 default_config.yaml 文件中设置 "enable_modelarts=True"
# 在 default_config.yaml 文件中设置 "distribute=True"
# 在 default_config.yaml 文件中设置 "data_path='/cache/data'"
# 在 default_config.yaml 文件中设置 "ckpt_path='/cache/train'"
# (可选)在 default_config.yaml 文件中设置 "checkpoint_url='s3://dir_to_your_pretrained/'"
# 在 default_config.yaml 文件中设置 其他参数
# b. 在网页上设置 "enable_modelarts=True"
# 在网页上设置 "distribute=True"
# 在网页上设置 "data_path='/cache/data'"
# 在网页上设置 "ckpt_path='/cache/train'"
# (可选)在网页上设置 "checkpoint_url='s3://dir_to_your_pretrained/'"
# 在网页上设置 其他参数
# (2) 准备模型代码
# (3) 如果选择微调您的模型，请上传你的预训练模型到 S3 桶上
# (4) 上传原始 cifar10 数据集到 S3 桶上
# (5) 在网页上设置你的代码路径为 "/path/alexnet"
# (6) 在网页上设置启动文件为 "train.py"
# (7) 在网页上设置"训练数据集"、"训练输出文件路径"、"作业日志路径"等
# (8) 创建训练作业
#
# 在 ModelArts 上使用单卡训练
# (1) 执行a或者b
# a. 在 default_config.yaml 文件中设置 "enable_modelarts=True"
# 在 default_config.yaml 文件中设置 "data_path='/cache/data'"
# 在 default_config.yaml 文件中设置 "ckpt_path='/cache/train'"
# (可选)在 default_config.yaml 文件中设置 "checkpoint_url='s3://dir_to_your_pretrained/'"
# 在 default_config.yaml 文件中设置 其他参数
# b. 在网页上设置 "enable_modelarts=True"
# 在网页上设置 "data_path='/cache/data'"
# 在网页上设置 "ckpt_path='/cache/train'"
# (可选)在网页上设置 "checkpoint_url='s3://dir_to_your_pretrained/'"
# 在网页上设置 其他参数
# (2) 准备模型代码
# (3) 如果选择微调您的模型，上传你的预训练模型到 S3 桶上
# (4) 上传原始 cifar10 数据集到 S3 桶上
# (5) 在网页上设置你的代码路径为 "/path/alexnet"
# (6) 在网页上设置启动文件为 "train.py"
# (7) 在网页上设置"训练数据集"、"训练输出文件路径"、"作业日志路径"等
# (8) 创建训练作业
#
# 在 ModelArts 上使用单卡验证
# (1) 执行a或者b
# a. 在 default_config.yaml 文件中设置 "enable_modelarts=True"
# 在 default_config.yaml 文件中设置 "data_path='/cache/data'"
# 在 default_config.yaml 文件中设置 "ckpt_file='/cache/train/checkpoint_alexnet-30_1562.ckpt'"
# 在 default_config.yaml 文件中设置 其他参数
# b. 在网页上设置 "enable_modelarts=True"
# 在网页上设置 "data_path='/cache/data'"
# 在网页上设置 "ckpt_file='/cache/train/checkpoint_alexnet-30_1562.ckpt'"
# 在网页上设置 其他参数
# (2) 准备模型代码
# (3) 上传你训练好的模型到 S3 桶上
# (4) 上传原始 cifar10 数据集到 S3 桶上
# (5) 在网页上设置你的代码路径为 "/path/alexnet"
# (6) 在网页上设置启动文件为 "train.py"
# (7) 在网页上设置"训练数据集"、"训练输出文件路径"、"作业日志路径"等
# (8) 创建训练作业
```
## 脚本说明
### 脚本及样例代码
@ -121,7 +185,7 @@ train.py和config.py中主要参数如下：
- Ascend处理器环境运行
```bash
python train.py --data_path cifar-10-batches-bin --ckpt_path ckpt > log 2>&1 &
python train.py --config_path default_config.yaml --data_path cifar-10-batches-bin --ckpt_path ckpt > log 2>&1 &
# 或进入脚本目录，执行脚本
sh run_standalone_train_ascend.sh cifar-10-batches-bin ckpt
```
@ -143,7 +207,7 @@ train.py和config.py中主要参数如下：
- GPU环境运行
```bash
python train.py --device_target "GPU" --data_path cifar-10-batches-bin --ckpt_path ckpt > log 2>&1 &
python train.py --config_path default_config.yaml --device_target "GPU" --data_path cifar-10-batches-bin --ckpt_path ckpt > log 2>&1 &
# 或进入脚本目录，执行脚本
sh run_standalone_train_for_gpu.sh cifar-10-batches-bin ckpt
```
@ -168,7 +232,7 @@ train.py和config.py中主要参数如下：
- Ascend处理器环境运行
```bash
python eval.py --data_path cifar-10-verify-bin --ckpt_path ckpt/checkpoint_alexnet-1_1562.ckpt > eval_log.txt 2>&1 &
python eval.py --config_path default_config.yaml --data_path cifar-10-verify-bin --ckpt_path ckpt/checkpoint_alexnet-1_1562.ckpt > eval_log.txt 2>&1 &
#或进入脚本目录，执行脚本
sh run_standalone_eval_ascend.sh cifar-10-verify-bin ckpt/checkpoint_alexnet-1_1562.ckpt
```
@ -183,7 +247,7 @@ train.py和config.py中主要参数如下：
- GPU环境运行
```bash
python eval.py --device_target "GPU" --data_path cifar-10-verify-bin --ckpt_path ckpt/checkpoint_alexnet-30_1562.ckpt > eval_log 2>&1 &
python eval.py --config_path default_config.yaml --device_target "GPU" --data_path cifar-10-verify-bin --ckpt_path ckpt/checkpoint_alexnet-30_1562.ckpt > eval_log 2>&1 &
#或进入脚本目录，执行脚本
sh run_standalone_eval_for_gpu.sh cifar-10-verify-bin ckpt/checkpoint_alexnet-30_1562.ckpt
```
@ -200,7 +264,7 @@ train.py和config.py中主要参数如下：
### [导出MindIR](#contents)
```shell
python export.py --ckpt_file [CKPT_PATH] --file_name [FILE_NAME] --file_format [FILE_FORMAT]
python export.py --config_path [CONFIG_PATH] --ckpt_file [CKPT_PATH] --file_name [FILE_NAME] --file_format [FILE_FORMAT]
```
参数ckpt_file为必填项，

4
model_zoo/official/cv/alexnet/default_config.yaml
View File

@ -11,8 +11,8 @@ checkpoint_file: './checkpoint/checkpoint_alexnet-30_1562.ckpt'
device_target: Ascend
enable_profiling: False
ckpt_path: "/cache/data"
ckpt_file: "/cache/data/checkpoint_alexnet-30_1562.ckpt"
ckpt_path: "/cache/train"
ckpt_file: "/cache/train/checkpoint_alexnet-30_1562.ckpt"
# ==============================================================================
# Training options
epoch_size: 30

4
model_zoo/official/cv/alexnet/eval.py
View File

@ -38,7 +38,6 @@ def modelarts_process():
@moxing_wrapper(pre_process=modelarts_process)
def eval_alexnet():
print("============== Starting Testing ==============")
device_num = get_device_num()
if device_num > 1:
# context.set_context(mode=context.GRAPH_MODE, device_target=config.device_target)
@ -53,8 +52,7 @@ def eval_alexnet():
network = AlexNet(config.num_classes, phase='test')
loss = nn.SoftmaxCrossEntropyWithLogits(sparse=True, reduction="mean")
opt = nn.Momentum(network.trainable_params(), config.learning_rate, config.momentum)
ds_eval = create_dataset_cifar10(config, config.data_path, config.batch_size, status="test", \
target=config.device_target)
ds_eval = create_dataset_cifar10(config, config.data_path, config.batch_size, target=config.device_target)
param_dict = load_checkpoint(config.ckpt_path)
print("load checkpoint from [{}].".format(config.ckpt_path))
load_param_into_net(network, param_dict)

104
model_zoo/official/cv/inceptionv3/README.md
View File

@ -66,6 +66,98 @@ For FP16 operators, if the input data type is FP32, the backend of MindSpore wil
- [MindSpore Tutorials](https://www.mindspore.cn/tutorial/training/en/master/index.html)
- [MindSpore Python API](https://www.mindspore.cn/doc/api_python/en/master/index.html)
- Running on [ModelArts](https://support.huaweicloud.com/modelarts/)
```bash
# Train 8p with Ascend
# (1) Perform a or b.
# a. Set "enable_modelarts=True" on default_config.yaml file.
# Set "distribute=True" on default_config.yaml file.
# Set "need_modelarts_dataset_unzip=True" on default_config.yaml file.
# Set "modelarts_dataset_unzip_name='ImageNet_Original'" on default_config.yaml file.
# Set "lr_init=0.00004" on default_config.yaml file.
# Set "dataset_path='/cache/data'" on default_config.yaml file.
# Set "epoch_size=250" on default_config.yaml file.
# (optional)Set "checkpoint_url='s3://dir_to_your_pretrained/'" on default_config.yaml file.
# Set other parameters on default_config.yaml file you need.
# b. Add "enable_modelarts=True" on the website UI interface.
# Add "need_modelarts_dataset_unzip=True" on the website UI interface.
# Add "modelarts_dataset_unzip_name='ImageNet_Original'" on the website UI interface.
# Add "distribute=True" on the website UI interface.
# Add "lr_init=0.00004" on the website UI interface.
# Add "dataset_path=/cache/data" on the website UI interface.
# Add "epoch_size=250" on the website UI interface.
# (optional)Add "checkpoint_url='s3://dir_to_your_pretrained/'" on the website UI interface.
# Add other parameters on the website UI interface.
# (2) Prepare model code
# (3) Upload or copy your pretrained model to S3 bucket if you want to finetune.
# (4) Perform a or b. (suggested option a)
# a. First, zip MindRecord dataset to one zip file.
# Second, upload your zip dataset to S3 bucket.(you could also upload the origin mindrecord dataset, but it can be so slow.)
# b. Upload the original coco dataset to S3 bucket.
# (Data set conversion occurs during training process and costs a lot of time. it happens every time you train.)
# (5) Set the code directory to "/path/inceptionv3" on the website UI interface.
# (6) Set the startup file to "train.py" on the website UI interface.
# (7) Set the "Dataset path" and "Output file path" and "Job log path" to your path on the website UI interface.
# (8) Create your job.
#
# Train 1p with Ascend
# (1) Perform a or b.
# a. Set "enable_modelarts=True" on default_config.yaml file.
# Set "need_modelarts_dataset_unzip=True" on default_config.yaml file.
# Set "modelarts_dataset_unzip_name='ImageNet_Original'" on default_config.yaml file.
# Set "dataset_path='/cache/data'" on default_config.yaml file.
# Set "epoch_size=250" on default_config.yaml file.
# (optional)Set "checkpoint_url='s3://dir_to_your_pretrained/'" on default_config.yaml file.
# Set other parameters on default_config.yaml file you need.
# b. Add "enable_modelarts=True" on the website UI interface.
# Add "need_modelarts_dataset_unzip=True" on the website UI interface.
# Add "modelarts_dataset_unzip_name='ImageNet_Original'" on the website UI interface.
# Add "dataset_path='/cache/data'" on the website UI interface.
# Add "epoch_size=250" on the website UI interface.
# (optional)Add "checkpoint_url='s3://dir_to_your_pretrained/'" on the website UI interface.
# Add other parameters on the website UI interface.
# (2) Prepare model code
# (3) Upload or copy your pretrained model to S3 bucket if you want to finetune.
# (4) Perform a or b. (suggested option a)
# a. zip MindRecord dataset to one zip file.
# Second, upload your zip dataset to S3 bucket.(you could also upload the origin mindrecord dataset, but it can be so slow.)
# b. Upload the original coco dataset to S3 bucket.
# (Data set conversion occurs during training process and costs a lot of time. it happens every time you train.)
# (5) Set the code directory to "/path/inceptionv3" on the website UI interface.
# (6) Set the startup file to "train.py" on the website UI interface.
# (7) Set the "Dataset path" and "Output file path" and "Job log path" to your path on the website UI interface.
# (8) Create your job.
#
# Eval 1p with Ascend
# (1) Perform a or b.
# a. Set "enable_modelarts=True" on default_config.yaml file.
# Set "need_modelarts_dataset_unzip=True" on default_config.yaml file.
# Set "modelarts_dataset_unzip_name='ImageNet_Original'" on default_config.yaml file.
# Set "checkpoint_url='s3://dir_to_your_trained_model/'" on base_config.yaml file.
# Set "checkpoint='./inceptionv3/inceptionv3-rank3_1-247_1251.ckpt'" on default_config.yaml file.
# Set "dataset_path='/cache/data'" on default_config.yaml file.
# Set other parameters on default_config.yaml file you need.
# b. Add "enable_modelarts=True" on the website UI interface.
# Add "need_modelarts_dataset_unzip=True" on the website UI interface.
# Add "modelarts_dataset_unzip_name='ImageNet_Original'" on the website UI interface.
# Add "checkpoint_url='s3://dir_to_your_trained_model/'" on the website UI interface.
# Add "checkpoint='./inceptionv3/inceptionv3-rank3_1-247_1251.ckpt'" on the website UI interface.
# Add "dataset_path='/cache/data'" on the website UI interface.
# Add other parameters on the website UI interface.
# (2) Prepare model code
# (3) Upload or copy your trained model to S3 bucket.
# (4) Perform a or b. (suggested option a)
# a. First, zip MindRecord dataset to one zip file.
# Second, upload your zip dataset to S3 bucket.(you could also upload the origin mindrecord dataset, but it can be so slow.)
# b. Upload the original coco dataset to S3 bucket.
# (Data set conversion occurs during training process and costs a lot of time. it happens every time you train.)
# (5) Set the code directory to "/path/inceptionv3" on the website UI interface.
# (6) Set the startup file to "eval.py" on the website UI interface.
# (7) Set the "Dataset path" and "Output file path" and "Job log path" to your path on the website UI interface.
# (8) Create your job.
```
# [Script description](#contents)
## [Script and sample code](#contents)
@ -167,8 +259,8 @@ sh scripts/run_standalone_train_cpu.sh DATA_PATH
```python
# training example
python:
Ascend: python train.py --dataset_path DATA_PATH --platform Ascend
CPU: python train.py --dataset_path DATA_PATH --platform CPU
Ascend: python train.py --config_path CONFIG_FILE --dataset_path DATA_PATH --platform Ascend
CPU: python train.py --config_path CONFIG_FILE --dataset_path DATA_PATH --platform CPU
shell:
Ascend:
@ -229,8 +321,8 @@ You can start training using python or shell scripts. The usage of shell scripts
```python
# eval example
python:
Ascend: python eval.py --dataset_path DATA_PATH --checkpoint PATH_CHECKPOINT --platform Ascend
CPU: python eval.py --dataset_path DATA_PATH --checkpoint PATH_CHECKPOINT --platform CPU
Ascend: python eval.py --config_path CONFIG_FILE --dataset_path DATA_PATH --checkpoint PATH_CHECKPOINT --platform Ascend
CPU: python eval.py --config_path CONFIG_FILE --dataset_path DATA_PATH --checkpoint PATH_CHECKPOINT --platform CPU
shell:
Ascend: sh scripts/run_eval.sh DEVICE_ID DATA_PATH PATH_CHECKPOINT
@ -250,7 +342,7 @@ metric: {'Loss': 1.778, 'Top1-Acc':0.788, 'Top5-Acc':0.942}
## Model Export
```shell
python export.py --ckpt_file [CKPT_PATH] --device_target [DEVICE_TARGET] --file_format[EXPORT_FORMAT]
python export.py --config_path CONFIG_FILE --ckpt_file [CKPT_PATH] --device_target [DEVICE_TARGET] --file_format[EXPORT_FORMAT]
```
`EXPORT_FORMAT` should be in ["AIR", "MINDIR"]
@ -290,7 +382,7 @@ accuracy:78.742
| MindSpore Version | 0.6.0-beta |
| Dataset | 1200k images |
| Batch_size | 128 |
| Training Parameters | src/config.py |
| Training Parameters | src/model_utils/default_config.yaml |
| Optimizer | RMSProp |
| Loss Function | SoftmaxCrossEntropy |
| Outputs | probability |

156
model_zoo/official/cv/inceptionv3/README_CN.md
View File

@ -77,6 +77,98 @@ InceptionV3的总体网络架构如下：
- [MindSpore教程](https://www.mindspore.cn/tutorial/training/zh-CN/master/index.html)
- [MindSpore Python API](https://www.mindspore.cn/doc/api_python/zh-CN/master/index.html)
- 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/))
```bash
# 在 ModelArts 上使用8卡训练
# (1) 执行a或者b
# a. 在 default_config.yaml 文件中设置 "enable_modelarts=True"
# 在 default_config.yaml 文件中设置 "distribute=True"
# 在 default_config.yaml 文件中设置 "need_modelarts_dataset_unzip=True"
# 在 default_config.yaml 文件中设置 "modelarts_dataset_unzip_name='imagenet_original'"
# 在 default_config.yaml 文件中设置 "lr_init=0.00004"
# 在 default_config.yaml 文件中设置 "dataset_path='/cache/data'"
# 在 default_config.yaml 文件中设置 "epoch_size=250"
# (可选)在 default_config.yaml 文件中设置 "checkpoint_url='s3://dir_to_your_pretrained/'"
# 在 default_config.yaml 文件中设置 其他参数
# b. 在网页上设置 "enable_modelarts=True"
# 在网页上设置 "need_modelarts_dataset_unzip=True"
# 在网页上设置 "modelarts_dataset_unzip_name='imagenet_original'"
# 在网页上设置 "distribute=True"
# 在网页上设置 "lr_init=0.00004"
# 在网页上设置 "dataset_path=/cache/data"
# 在网页上设置 "epoch_size=250"
# (可选)在网页上设置 "checkpoint_url='s3://dir_to_your_pretrained/'"
# 在网页上设置 其他参数
# (2) 准备模型代码
# (3) 如果选择微调您的模型，请上传你的预训练模型到 S3 桶上
# (4) 执行a或者b (推荐选择 a)
# a. 第一, 将该数据集压缩为一个 ".zip" 文件。
# 第二, 上传你的压缩数据集到 S3 桶上 (你也可以上传未压缩的数据集，但那可能会很慢。)
# b. 上传原始 coco 数据集到 S3 桶上。
# (数据集转换发生在训练过程中，需要花费较多的时间。每次训练的时候都会重新进行转换。)
# (5) 在网页上设置你的代码路径为 "/path/inceptionv3"
# (6) 在网页上设置启动文件为 "train.py"
# (7) 在网页上设置"训练数据集"、"训练输出文件路径"、"作业日志路径"等
# (8) 创建训练作业
#
# 在 ModelArts 上使用单卡训练
# (1) 执行a或者b
# a. 在 default_config.yaml 文件中设置 "enable_modelarts=True"
# 在 default_config.yaml 文件中设置 "need_modelarts_dataset_unzip=True"
# 在 default_config.yaml 文件中设置 "modelarts_dataset_unzip_name='imagenet_original'"
# 在 default_config.yaml 文件中设置 "dataset_path='/cache/data'"
# 在 default_config.yaml 文件中设置 "epoch_size=250"
# (可选)在 default_config.yaml 文件中设置 "checkpoint_url='s3://dir_to_your_pretrained/'"
# 在 default_config.yaml 文件中设置 其他参数
# b. 在网页上设置 "enable_modelarts=True"
# 在网页上设置 "need_modelarts_dataset_unzip=True"
# 在网页上设置 "modelarts_dataset_unzip_name='imagenet_original'"
# 在网页上设置 "dataset_path='/cache/data'"
# 在网页上设置 "epoch_size=250"
# (可选)在网页上设置 "checkpoint_url='s3://dir_to_your_pretrained/'"
# 在网页上设置 其他参数
# (2) 准备模型代码
# (3) 如果选择微调您的模型，上传你的预训练模型到 S3 桶上
# (4) 执行a或者b (推荐选择 a)
# a. 第一, 将该数据集压缩为一个 ".zip" 文件。
# 第二, 上传你的压缩数据集到 S3 桶上 (你也可以上传未压缩的数据集，但那可能会很慢。)
# b. 上传原始 coco 数据集到 S3 桶上。
# (数据集转换发生在训练过程中，需要花费较多的时间。每次训练的时候都会重新进行转换。)
# (5) 在网页上设置你的代码路径为 "/path/inceptionv3"
# (6) 在网页上设置启动文件为 "train.py"
# (7) 在网页上设置"训练数据集"、"训练输出文件路径"、"作业日志路径"等
# (8) 创建训练作业
#
# 在 ModelArts 上使用单卡验证
# (1) 执行a或者b
# a. 在 default_config.yaml 文件中设置 "enable_modelarts=True"
# 在 default_config.yaml 文件中设置 "need_modelarts_dataset_unzip=True"
# 在 default_config.yaml 文件中设置 "modelarts_dataset_unzip_name='imagenet_original'"
# 在 default_config.yaml 文件中设置 "checkpoint_url='s3://dir_to_your_trained_model/'"
# 在 default_config.yaml 文件中设置 "checkpoint='./inceptionv3/inceptionv3-rank3_1-247_1251.ckpt'"
# 在 default_config.yaml 文件中设置 "dataset_path='/cache/data'"
# 在 default_config.yaml 文件中设置 其他参数
# b. 在网页上设置 "enable_modelarts=True"
# 在网页上设置 "need_modelarts_dataset_unzip=True"
# 在网页上设置 "modelarts_dataset_unzip_name='imagenet_original'"
# 在网页上设置 "checkpoint_url='s3://dir_to_your_trained_model/'"
# 在网页上设置 "checkpoint='./inceptionv3/inceptionv3-rank3_1-247_1251.ckpt'"
# 在网页上设置 "dataset_path='/cache/data'"
# 在网页上设置 其他参数
# (2) 准备模型代码
# (3) 上传你训练好的模型到 S3 桶上
# (4) 执行a或者b (推荐选择 a)
# a. 第一, 将该数据集压缩为一个 ".zip" 文件。
# 第二, 上传你的压缩数据集到 S3 桶上 (你也可以上传未压缩的数据集，但那可能会很慢。)
# b. 上传原始 coco 数据集到 S3 桶上。
# (数据集转换发生在训练过程中，需要花费较多的时间。每次训练的时候都会重新进行转换。)
# (5) 在网页上设置你的代码路径为 "/path/inceptionv3"
# (6) 在网页上设置启动文件为 "train.py"
# (7) 在网页上设置"训练数据集"、"训练输出文件路径"、"作业日志路径"等
# (8) 创建训练作业
```
# 脚本说明
## 脚本和样例代码
@ -172,8 +264,8 @@ train.py和config.py中主要参数如下：
``` launch
# 训练示例
python:
Ascend: python train.py --dataset_path /dataset/train --platform Ascend
CPU: python train.py --dataset_path DATA_PATH --platform CPU
Ascend: python train.py --config_path default_config.yaml --dataset_path /dataset/train --platform Ascend
CPU: python train.py --config_path CONFIG_FILE --dataset_path DATA_PATH --platform CPU
shell:
Ascend:
@ -234,8 +326,8 @@ epoch time: 6358482.104 ms, per step time: 16303.800 ms
``` launch
# 评估示例
python:
Ascend: python eval.py --dataset_path DATA_DIR --checkpoint PATH_CHECKPOINT --platform Ascend
CPU: python eval.py --dataset_path DATA_PATH --checkpoint PATH_CHECKPOINT --platform CPU
Ascend: python eval.py --config_path CONFIG_FILE --dataset_path DATA_DIR --checkpoint PATH_CHECKPOINT --platform Ascend
CPU: python eval.py --config_path CONFIG_FILE --dataset_path DATA_PATH --checkpoint PATH_CHECKPOINT --platform CPU
shell:
Ascend: sh scripts/run_eval.sh DEVICE_ID DATA_DIR PATH_CHECKPOINT
@ -255,7 +347,7 @@ metric:{'Loss':1.778, 'Top1-Acc':0.788, 'Top5-Acc':0.942}
## 模型导出
```shell
python export.py --ckpt_file [CKPT_PATH] --device_target [DEVICE_TARGET] --file_format[EXPORT_FORMAT]
python export.py --config_path [CONFIG_FILE] --ckpt_file [CKPT_PATH] --device_target [DEVICE_TARGET] --file_format[EXPORT_FORMAT]
```
`EXPORT_FORMAT` 可选 ["AIR", "MINDIR"]
@ -287,39 +379,39 @@ accuracy:78.742
### 训练性能
| 参数 | Ascend |
| -------------------------- | ---------------------------------------------- |
| 模型版本 | InceptionV3 |
| 资源 | Ascend 910；CPU 2.60GHz，192核；内存 755G；系统 Euler2.8 |
| 上传日期 | 2020-08-21 |
| MindSpore版本 | 0.6.0-beta |
| 数据集 | 120万张图像 |
| Batch_size | 128 |
| 训练参数 | src/config.py |
| 优化器 | RMSProp |
| 损失函数 | Softmax交叉熵 |
| 输出 | 概率 |
| 损失 | 1.98 |
| 总时长（8卡） | 11小时 |
| 参数(M) | 103M |
| 微调检查点 | 313M |
| 训练速度 | 单卡：1050img/s;8卡：8000 img/s |
| 脚本 | [inceptionv3脚本](https://gitee.com/mindspore/mindspore/tree/master/model_zoo/official/cv/inceptionv3) |
| 参数 | Ascend |
| -------------------------- | ------------------------------------------------------- |
| 模型版本 | InceptionV3 |
| 资源 | Ascend 910；CPU 2.60GHz，192核；内存 755G；系统 Euler2.8|
| 上传日期 | 2020-08-21 |
| MindSpore版本 | 0.6.0-beta |
| 数据集 | 120万张图像 |
| Batch_size | 128 |
| 训练参数 | src/model_utils/default_config.yaml |
| 优化器 | RMSProp |
| 损失函数 | Softmax交叉熵 |
| 输出 | 概率 |
| 损失 | 1.98 |
| 总时长（8卡） | 11小时 |
| 参数(M) | 103M |
| 微调检查点 | 313M |
| 训练速度 | 单卡：1050img/s;8卡：8000 img/s |
| 脚本 | [inceptionv3脚本](https://gitee.com/mindspore/mindspore/tree/master/model_zoo/official/cv/inceptionv3) |
#### 推理性能
| 参数 | Ascend |
| 参数 | Ascend |
| ------------------- | --------------------------- |
| 模型版本 | InceptionV3 |
| 资源 | Ascend 910；CPU 2.60GHz，192核；内存 755G；系统 Euler2.8 |
| 上传日期 | 2020-08-22 |
| 模型版本 | InceptionV3 |
| 资源 | Ascend 910；CPU 2.60GHz，192核；内存 755G；系统 Euler2.8|
| 上传日期 | 2020-08-22 |
| MindSpore 版本 | 0.6.0-beta |
| 数据集 | 5万张图像 |
| Batch_size | 128 |
| 数据集 | 5万张图像 |
| Batch_size | 128 |
| 输出 | 概率 |
| 准确率 | ACC1[78.8%] ACC5[94.2%] |
| 总时长 | 2分钟 |
| 推理模型 | 92M (.onnx文件) |
| 准确率 | ACC1[78.8%] ACC5[94.2%] |
| 总时长 | 2分钟 |
| 推理模型 | 92M (.onnx文件) |
# 随机情况说明

1
model_zoo/official/cv/inceptionv3/train.py
View File

@ -190,6 +190,5 @@ def train_inceptionv3():
model.train(config.epoch_size, dataset, callbacks=callbacks, dataset_sink_mode=config.ds_sink_mode)
print("train success")
if __name__ == '__main__':
train_inceptionv3()

96
model_zoo/official/cv/inceptionv4/README.md
View File

@ -59,6 +59,98 @@ For FP16 operators, if the input data type is FP32, the backend of MindSpore wil
- [MindSpore Tutorials](https://www.mindspore.cn/tutorial/training/en/master/index.html)
- [MindSpore Python API](https://www.mindspore.cn/doc/api_python/en/master/index.html)
- Running on [ModelArts](https://support.huaweicloud.com/modelarts/)
```bash
# Train 8p with Ascend
# (1) Perform a or b.
# a. Set "enable_modelarts=True" on default_config.yaml file.
# Set "distribute=True" on default_config.yaml file.
# Set "need_modelarts_dataset_unzip=True" on default_config.yaml file.
# Set "modelarts_dataset_unzip_name='ImageNet_Original'" on default_config.yaml file.
# Set "lr_init=0.00004" on default_config.yaml file.
# Set "dataset_path='/cache/data'" on default_config.yaml file.
# Set "epoch_size=250" on default_config.yaml file.
# (optional)Set "checkpoint_url='s3://dir_to_your_pretrained/'" on default_config.yaml file.
# Set other parameters on default_config.yaml file you need.
# b. Add "enable_modelarts=True" on the website UI interface.
# Add "need_modelarts_dataset_unzip=True" on the website UI interface.
# Add "modelarts_dataset_unzip_name='ImageNet_Original'" on the website UI interface.
# Add "distribute=True" on the website UI interface.
# Add "lr_init=0.00004" on the website UI interface.
# Add "dataset_path=/cache/data" on the website UI interface.
# Add "epoch_size=250" on the website UI interface.
# (optional)Add "checkpoint_url='s3://dir_to_your_pretrained/'" on the website UI interface.
# Add other parameters on the website UI interface.
# (2) Prepare model code
# (3) Upload or copy your pretrained model to S3 bucket if you want to finetune.
# (4) Perform a or b. (suggested option a)
# a. First, zip MindRecord dataset to one zip file.
# Second, upload your zip dataset to S3 bucket.(you could also upload the origin mindrecord dataset, but it can be so slow.)
# b. Upload the original coco dataset to S3 bucket.
# (Data set conversion occurs during training process and costs a lot of time. it happens every time you train.)
# (5) Set the code directory to "/path/inceptionv4" on the website UI interface.
# (6) Set the startup file to "train.py" on the website UI interface.
# (7) Set the "Dataset path" and "Output file path" and "Job log path" to your path on the website UI interface.
# (8) Create your job.
#
# Train 1p with Ascend
# (1) Perform a or b.
# a. Set "enable_modelarts=True" on default_config.yaml file.
# Set "need_modelarts_dataset_unzip=True" on default_config.yaml file.
# Set "modelarts_dataset_unzip_name='ImageNet_Original'" on default_config.yaml file.
# Set "dataset_path='/cache/data'" on default_config.yaml file.
# Set "epoch_size=250" on default_config.yaml file.
# (optional)Set "checkpoint_url='s3://dir_to_your_pretrained/'" on default_config.yaml file.
# Set other parameters on default_config.yaml file you need.
# b. Add "enable_modelarts=True" on the website UI interface.
# Add "need_modelarts_dataset_unzip=True" on the website UI interface.
# Add "modelarts_dataset_unzip_name='ImageNet_Original'" on the website UI interface.
# Add "dataset_path='/cache/data'" on the website UI interface.
# Add "epoch_size=250" on the website UI interface.
# (optional)Add "checkpoint_url='s3://dir_to_your_pretrained/'" on the website UI interface.
# Add other parameters on the website UI interface.
# (2) Prepare model code
# (3) Upload or copy your pretrained model to S3 bucket if you want to finetune.
# (4) Perform a or b. (suggested option a)
# a. zip MindRecord dataset to one zip file.
# Second, upload your zip dataset to S3 bucket.(you could also upload the origin mindrecord dataset, but it can be so slow.)
# b. Upload the original coco dataset to S3 bucket.
# (Data set conversion occurs during training process and costs a lot of time. it happens every time you train.)
# (5) Set the code directory to "/path/inceptionv4" on the website UI interface.
# (6) Set the startup file to "train.py" on the website UI interface.
# (7) Set the "Dataset path" and "Output file path" and "Job log path" to your path on the website UI interface.
# (8) Create your job.
#
# Eval 1p with Ascend
# (1) Perform a or b.
# a. Set "enable_modelarts=True" on default_config.yaml file.
# Set "need_modelarts_dataset_unzip=True" on default_config.yaml file.
# Set "modelarts_dataset_unzip_name='ImageNet_Original'" on default_config.yaml file.
# Set "checkpoint_url='s3://dir_to_your_trained_model/'" on base_config.yaml file.
# Set "checkpoint_path='./inceptionv4/inceptionv4-train-250_1251.ckpt'" on default_config.yaml file.
# Set "dataset_path='/cache/data'" on default_config.yaml file.
# Set other parameters on default_config.yaml file you need.
# b. Add "enable_modelarts=True" on the website UI interface.
# Add "need_modelarts_dataset_unzip=True" on the website UI interface.
# Add "modelarts_dataset_unzip_name='ImageNet_Original'" on the website UI interface.
# Add "checkpoint_url='s3://dir_to_your_trained_model/'" on the website UI interface.
# Add "checkpoint_path='./inceptionv4/inceptionv4-train-250_1251.ckpt'" on the website UI interface.
# Add "dataset_path='/cache/data'" on the website UI interface.
# (2) Prepare model code
# Add other parameters on the website UI interface.
# (3) Upload or copy your trained model to S3 bucket.
# (4) Perform a or b. (suggested option a)
# a. First, zip MindRecord dataset to one zip file.
# Second, upload your zip dataset to S3 bucket.(you could also upload the origin mindrecord dataset, but it can be so slow.)
# b. Upload the original coco dataset to S3 bucket.
# (Data set conversion occurs during training process and costs a lot of time. it happens every time you train.)
# (5) Set the code directory to "/path/inceptionv4" on the website UI interface.
# (6) Set the startup file to "eval.py" on the website UI interface.
# (7) Set the "Dataset path" and "Output file path" and "Job log path" to your path on the website UI interface.
# (8) Create your job.
```
# [Script description](#contents)
## [Script and sample code](#contents)
@ -248,7 +340,7 @@ metric: {'Loss': 0.8144, 'Top1-Acc': 0.8009, 'Top5-Acc': 0.9457}
## Model Export
```shell
python export.py --ckpt_file [CKPT_PATH] --device_target [DEVICE_TARGET] --file_format[EXPORT_FORMAT]
python export.py --config_path [CONFIG_FILE] --ckpt_file [CKPT_PATH] --device_target [DEVICE_TARGET] --file_format[EXPORT_FORMAT]
```
`EXPORT_FORMAT` should be in ["AIR", "MINDIR"]
@ -288,7 +380,7 @@ accuracy:80.044
| MindSpore Version | 1.0.0 | 1.0.0 |
| Dataset | 1200k images | 1200K images |
| Batch_size | 128 | 128 |
| Training Parameters | src/config.py (Ascend) | src/config.py (GPU) |
| Training Parameters | src/model_utils/default_config.yaml (Ascend) | src/model_utils/default_config.yaml (GPU)|
| Optimizer | RMSProp | RMSProp |
| Loss Function | SoftmaxCrossEntropyWithLogits | SoftmaxCrossEntropyWithLogits |
| Outputs | probability | probability |

74
model_zoo/official/cv/lenet/README.md
View File

@ -78,6 +78,72 @@ sh run_standalone_train_ascend.sh [DATA_PATH] [CKPT_SAVE_PATH]
sh run_standalone_eval_ascend.sh [DATA_PATH] [CKPT_NAME]
```
- Running on [ModelArts](https://support.huaweicloud.com/modelarts/)
```bash
# Train 8p with Ascend
# (1) Perform a or b.
# a. Set "enable_modelarts=True" on default_config.yaml file.
# Set "distribute=True" on default_config.yaml file.
# Set "data_path='/cache/data'" on default_config.yaml file.
# Set "ckpt_path='/cache/data'" on default_config.yaml file.
# (optional)Set "checkpoint_url='s3://dir_to_your_pretrained/'" on default_config.yaml file.
# Set other parameters on default_config.yaml file you need.
# b. Add "enable_modelarts=True" on the website UI interface.
# Add "need_modelarts_dataset_unzip=True" on the website UI interface.
# Add "data_path='/cache/data'" on the website UI interface.
# Add "ckpt_path='/cache/data'" on the website UI interface.
# (optional)Add "checkpoint_url='s3://dir_to_your_pretrained/'" on the website UI interface.
# Add other parameters on the website UI interface.
# (2) Prepare model code.
# (3) Upload or copy your pretrained model to S3 bucket if you want to finetune.
# (4) Upload the original mnist_data dataset to S3 bucket.
# (5) Set the code directory to "/path/lenet" on the website UI interface.
# (6) Set the startup file to "train.py" on the website UI interface.
# (7) Set the "Dataset path" and "Output file path" and "Job log path" to your path on the website UI interface.
# (8) Create your job.
#
# Train 1p with Ascend
# (1) Perform a or b.
# a. Set "enable_modelarts=True" on default_config.yaml file.
# Set "data_path='/cache/data'" on default_config.yaml file.
# Set "ckpt_path='/cache/data'" on default_config.yaml file.
# (optional)Set "checkpoint_url='s3://dir_to_your_pretrained/'" on default_config.yaml file.
# Set other parameters on default_config.yaml file you need.
# b. Add "enable_modelarts=True" on the website UI interface.
# Add "data_path='/cache/data'" on the website UI interface.
# Add "ckpt_path='/cache/data'" on the website UI interface.
# (optional)Add "checkpoint_url='s3://dir_to_your_pretrained/'" on the website UI interface.
# Add other parameters on the website UI interface.
# (2) Prepare model code.
# (3) Upload or copy your pretrained model to S3 bucket if you want to finetune.
# (4) Upload the original mnist_data dataset to S3 bucket.
# (5) Set the code directory to "/path/lenet" on the website UI interface.
# (6) Set the startup file to "train.py" on the website UI interface.
# (7) Set the "Dataset path" and "Output file path" and "Job log path" to your path on the website UI interface.
# (8) Create your job.
#
# Eval 1p with Ascend
# (1) Perform a or b.
# a. Set "enable_modelarts=True" on default_config.yaml file.
# Set "data_path='/cache/data'" on default_config.yaml file.
# Set "checkpoint_url='s3://dir_to_your_trained_model/'" on base_config.yaml file.
# Set "ckpt_file='/cache/data/checkpoint_lenet-10_1875.ckpt'" on default_config.yaml file.
# Set other parameters on default_config.yaml file you need.
# b. Add "enable_modelarts=True" on the website UI interface.
# Add "data_path='/cache/data'" on the website UI interface.
# Add "checkpoint_url='s3://dir_to_your_trained_model/'" on the website UI interface.
# Add "ckpt_file='/cache/data/checkpoint_lenet-10_1875.ckpt'" on the website UI interface.
# Add other parameters on the website UI interface.
# (2) Prepare model code.
# (3) Upload or copy your trained model to S3 bucket.
# (4) Upload the original mnist_data dataset to S3 bucket.
# (5) Set the code directory to "/path/lenet" on the website UI interface.
# (6) Set the startup file to "eval.py" on the website UI interface.
# (7) Set the "Dataset path" and "Output file path" and "Job log path" to your path on the website UI interface.
# (8) Create your job.
```
## [Script Description](#contents)
### [Script and Sample Code](#contents)
@ -115,7 +181,7 @@ sh run_standalone_eval_ascend.sh [DATA_PATH] [CKPT_NAME]
## [Script Parameters](#contents)
```python
Major parameters in train.py and config.py as follows:
Major parameters in train.py and default_config.yaml as follows:
--data_path: The absolute full path to the train and evaluation datasets.
--epoch_size: Total training epochs.
@ -134,7 +200,7 @@ Major parameters in train.py and config.py as follows:
### Training
```bash
python train.py --data_path Data --ckpt_path ckpt > log.txt 2>&1 &
python train.py --config_path CONFIG_PATH --data_path Data --ckpt_path ckpt > log.txt 2>&1 &
# or enter script dir, and run the script
sh run_standalone_train_ascend.sh Data ckpt
```
@ -160,7 +226,7 @@ The model checkpoint will be saved in the current directory.
Before running the command below, please check the checkpoint path used for evaluation.
```bash
python eval.py --data_path Data --ckpt_path ckpt/checkpoint_lenet-1_1875.ckpt > log.txt 2>&1 &
python eval.py --config_path CONFIG_PATH --data_path Data --ckpt_path ckpt/checkpoint_lenet-1_1875.ckpt > log.txt 2>&1 &
# or enter script dir, and run the script
sh run_standalone_eval_ascend.sh Data ckpt/checkpoint_lenet-1_1875.ckpt
```
@ -177,7 +243,7 @@ You can view the results through the file "log.txt". The accuracy of the test da
### Export MindIR
```shell
python export.py --ckpt_file [CKPT_PATH] --file_name [FILE_NAME] --file_format [FILE_FORMAT]
python export.py --config_path [CONFIG_PATH] --ckpt_file [CKPT_PATH] --file_name [FILE_NAME] --file_format [FILE_FORMAT]
```
The ckpt_file parameter is required,

119
model_zoo/official/cv/lenet/README_CN.md
View File

@ -82,6 +82,69 @@ sh run_standalone_train_ascend.sh [DATA_PATH] [CKPT_SAVE_PATH]
sh run_standalone_eval_ascend.sh [DATA_PATH] [CKPT_NAME]
```
- 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/))
```bash
# 在 ModelArts 上使用8卡训练
# (1) 执行a或者b
# a. 在 default_config.yaml 文件中设置 "enable_modelarts=True"
# 在 default_config.yaml 文件中设置 "distribute=True"
# 在 default_config.yaml 文件中设置 "data_path='/cache/data'"
# 在 default_config.yaml 文件中设置 "ckpt_path='/cache/data'"
# (可选)在 default_config.yaml 文件中设置 "checkpoint_url='s3://dir_to_your_pretrained/'"
# 在 default_config.yaml 文件中设置 其他参数
# b. 在网页上设置 "enable_modelarts=True"
# 在网页上设置 "distribute=True"
# 在网页上设置 "data_path=/cache/data"
# 在网页上设置 "ckpt_path=/cache/data"
# (可选)在网页上设置 "checkpoint_url='s3://dir_to_your_pretrained/'"
# 在网页上设置 其他参数
# (3) 如果选择微调您的模型，请上传你的预训练模型到 S3 桶上
# (4) 上传原始 mnist_data 数据集到 S3 桶上。
# (5) 在网页上设置你的代码路径为 "/path/lenet"
# (6) 在网页上设置启动文件为 "train.py"
# (7) 在网页上设置"训练数据集"、"训练输出文件路径"、"作业日志路径"等
# (8) 创建训练作业
#
# 在 ModelArts 上使用单卡训练
# (1) 执行a或者b
# a. 在 default_config.yaml 文件中设置 "enable_modelarts=True"
# 在 default_config.yaml 文件中设置 "data_path='/cache/data'"
# 在 default_config.yaml 文件中设置 "ckpt_path='/cache/data'"
# (可选)在 default_config.yaml 文件中设置 "checkpoint_url='s3://dir_to_your_pretrained/'"
# 在 default_config.yaml 文件中设置 其他参数
# b. 在网页上设置 "enable_modelarts=True"
# 在网页上设置 "data_path='/cache/data'"
# 在网页上设置 "ckpt_path='/cache/data'"
# (可选)在网页上设置 "checkpoint_url='s3://dir_to_your_pretrained/'"
# 在网页上设置 其他参数
# (3) 如果选择微调您的模型，上传你的预训练模型到 S3 桶上
# (4) 上传原始 mnist_data 数据集到 S3 桶上。
# (5) 在网页上设置你的代码路径为 "/path/lenet"
# (6) 在网页上设置启动文件为 "train.py"
# (7) 在网页上设置"训练数据集"、"训练输出文件路径"、"作业日志路径"等
# (8) 创建训练作业
#
# 在 ModelArts 上使用单卡验证
# (1) 执行a或者b
# a. 在 default_config.yaml 文件中设置 "enable_modelarts=True"
# 在 default_config.yaml 文件中设置 "checkpoint_url='s3://dir_to_your_trained_model/'"
# 在 default_config.yaml 文件中设置 "data_path='/cache/data'"
# 在 default_config.yaml 文件中设置 "ckpt_file='/cache/data/checkpoint_lenet-10_1875.ckpt'"
# 在 default_config.yaml 文件中设置 其他参数
# b. 在网页上设置 "enable_modelarts=True"
# 在网页上设置 "checkpoint_url='s3://dir_to_your_trained_model/'"
# 在网页上设置 "data_path='/cache/data'"
# 在网页上设置 "ckpt_file='/cache/data/checkpoint_lenet-10_1875.ckpt'"
# 在网页上设置 其他参数
# (3) 上传你训练好的模型到 S3 桶上
# (4) 上传原始 mnist_data 数据集到 S3 桶上。
# (5) 在网页上设置你的代码路径为 "/path/lenet"
# (6) 在网页上设置启动文件为 "eval.py"
# (7) 在网页上设置"训练数据集"、"训练输出文件路径"、"作业日志路径"等
# (8) 创建训练作业
```
## 脚本说明
### 脚本及样例代码
@ -119,7 +182,7 @@ sh run_standalone_eval_ascend.sh [DATA_PATH] [CKPT_NAME]
## 脚本参数
```python
train.py和config.py中主要参数如下：
train.py和default_config.yaml中主要参数如下：
--data_path: 到训练和评估数据集的绝对全路径
--epoch_size: 训练轮次数
@ -136,7 +199,7 @@ train.py和config.py中主要参数如下：
### 训练
```bash
python train.py --data_path Data --ckpt_path ckpt > log.txt 2>&1 &
python train.py --config_path CONFIG_PATH --data_path Data --ckpt_path ckpt > log.txt 2>&1 &
# or enter script dir, and run the script
sh run_standalone_train_ascend.sh Data ckpt
```
@ -162,7 +225,7 @@ epoch:1 step:1538, loss is 1.0221305
在运行以下命令之前，请检查用于评估的检查点路径。
```bash
python eval.py --data_path Data --ckpt_path ckpt/checkpoint_lenet-1_1875.ckpt > log.txt 2>&1 &
python eval.py --config_path CONFIG_PATH --data_path Data --ckpt_path ckpt/checkpoint_lenet-1_1875.ckpt > log.txt 2>&1 &
# or enter script dir, and run the script
sh run_standalone_eval_ascend.sh Data ckpt/checkpoint_lenet-1_1875.ckpt
```
@ -179,7 +242,7 @@ sh run_standalone_eval_ascend.sh Data ckpt/checkpoint_lenet-1_1875.ckpt
### 导出MindIR
```shell
python export.py --ckpt_file [CKPT_PATH] --file_name [FILE_NAME] --file_format [FILE_FORMAT]
python export.py --config_path [CONFIG_PATH] --ckpt_file [CKPT_PATH] --file_name [FILE_NAME] --file_format [FILE_FORMAT]
```
参数ckpt_file为必填项，
@ -213,33 +276,33 @@ bash run_infer_310.sh [MINDIR_PATH] [DATA_PATH] [DVPP] [DEVICE_ID]
### 评估性能
| 参数 | LeNet |
| -------------------------- | ----------------------------------------------------------- |
| 资源 | Ascend 910；CPU 2.60GHz，192核；内存 755G；系统 Euler2.8 |
| 上传日期 | 2020-06-09 |
| MindSpore版本 | 0.5.0-beta |
| 数据集 | MNIST |
| 训练参数 | epoch=10, steps=1875, batch_size = 32, lr=0.01 |
| 优化器 | Momentum |
| 损失函数 | Softmax交叉熵 |
| 输出 | 概率 |
| 损失 | 0.002 |
| 速度 | 1.70毫秒/步 |
| 总时长 | 43.1秒 | |
| 微调检查点 | 482k (.ckpt文件) |
| 脚本 | [LeNet脚本](https://gitee.com/mindspore/mindspore/tree/master/model_zoo/official/cv/lenet) |
| -------------------- | ------------------------------------------------------- |
| 资源 | Ascend 910；CPU 2.60GHz，192核；内存 755G；系统 Euler2.8|
| 上传日期 | 2020-06-09 |
| MindSpore版本 | 0.5.0-beta |
| 数据集 | MNIST |
| 训练参数 | epoch=10, steps=1875, batch_size = 32, lr=0.01 |
| 优化器 | Momentum |
| 损失函数 | Softmax交叉熵 |
| 输出 | 概率 |
| 损失 | 0.002 |
| 速度 | 1.70毫秒/步 |
| 总时长 | 43.1秒 |
| 微调检查点 | 482k (.ckpt文件) |
| 脚本 | [LeNet脚本](https://gitee.com/mindspore/mindspore/tree/master/model_zoo/official/cv/lenet) |
### 推理性能
| 参数 | Ascend |
| ------------- | ----------------------------|
| 模型版本 | LeNet |
| 资源 | Ascend 310；系统 CentOS 3.10 |
| 上传日期 | 2021-05-07 |
| MindSpore版本 | 1.2.0 |
| 数据集 | Mnist |
| batch_size | 1 |
| 输出 | Accuracy |
| 准确率 | Accuracy=0.9843 |
| 参数 | Ascend |
| --------------- | -----------------------------|
| 模型版本 | LeNet |
| 资源 | Ascend 310；系统 CentOS 3.10 |
| 上传日期 | 2021-05-07 |
| MindSpore版本 | 1.2.0 |
| 数据集 | Mnist |
| batch_size | 1 |
| 输出 | Accuracy |
| 准确率 | Accuracy=0.9843 |
| 推理模型 | 482K（.ckpt文件） |
## 随机情况说明

117
model_zoo/official/cv/maskrcnn/README.md
View File

@ -175,6 +175,121 @@ bash run_eval.sh [VALIDATION_JSON_FILE] [CHECKPOINT_PATH] [DATA_PATH]
bash run_infer_310.sh [MINDIR_PATH] [DATA_PATH] [ANN_FILE] [DEVICE_ID]
```
- Running on [ModelArts](https://support.huaweicloud.com/modelarts/)
```bash
# Train 8p with Ascend
# (1) Perform a or b.
# a. Set "enable_modelarts=True" on default_config.yaml file.
# Set "distribute=True" on default_config.yaml file.
# Set "need_modelarts_dataset_unzip=True" on default_config.yaml file.
# Set "modelarts_dataset_unzip_name='cocodataset'" on default_config.yaml file.
# Set "base_lr=0.02" on default_config.yaml file.
# Set "mindrecord_dir='./MindRecord_COCO'" on default_config.yaml file.
# Set "data_path='/cache/data'" on default_config.yaml file.
# Set "ann_file='./annotations/instances_val2017.json'" on default_config.yaml file.
# Set "epoch_size=12" on default_config.yaml file.
# Set "ckpt_path='./ckpt_maskrcnn/mask_rcnn-12_7393.ckpt'" on default_config.yaml file.
# (optional)Set "checkpoint_url='s3://dir_to_your_pretrained/'" on default_config.yaml file.
# Set other parameters on default_config.yaml file you need.
# b. Add "enable_modelarts=True" on the website UI interface.
# Add "need_modelarts_dataset_unzip=True" on the website UI interface.
# Add "modelarts_dataset_unzip_name='cocodataset'" on the website UI interface.
# Add "distribute=True" on the website UI interface.
# Add "base_lr=0.02" on the website UI interface.
# Add "mindrecord_dir='./MindRecord_COCO'" on the website UI interface.
# Add "data_path='/cache/data'" on the website UI interface.
# Add "ann_file='./annotations/instances_val2017.json'" on the website UI interface.
# Add "epoch_size=12" on the website UI interface.
# Set "ckpt_path='./ckpt_maskrcnn/mask_rcnn-12_7393.ckpt'" on default_config.yaml file.
# (optional)Add "checkpoint_url='s3://dir_to_your_pretrained/'" on the website UI interface.
# Add other parameters on the website UI interface.
# (2) Prepare model code
# (3) Upload or copy your pretrained model to S3 bucket if you want to finetune.
# (4) Perform a or b. (suggested option a)
# a. First, run "train.py" like the following to create MindRecord dataset locally from coco2017.
# "python train.py --only_create_dataset=True --mindrecord_dir=$MINDRECORD_DIR --data_path=$DATA_PATH --ann_file=$ANNO_PATH"
# Second, zip MindRecord dataset to one zip file.
# Finally, Upload your zip dataset to S3 bucket.(you could also upload the origin mindrecord dataset, but it can be so slow.)
# b. Upload the original coco dataset to S3 bucket.
# (Data set conversion occurs during training process and costs a lot of time. it happens every time you train.)
# (5) Set the code directory to "/path/maskrcnn" on the website UI interface.
# (6) Set the startup file to "train.py" on the website UI interface.
# (7) Set the "Dataset path" and "Output file path" and "Job log path" to your path on the website UI interface.
# (8) Create your job.
#
# Train 1p with Ascend
# (1) Perform a or b.
# a. Set "enable_modelarts=True" on default_config.yaml file.
# Set "need_modelarts_dataset_unzip=True" on default_config.yaml file.
# Set "modelarts_dataset_unzip_name='cocodataset'" on default_config.yaml file.
# Set "mindrecord_dir='./MindRecord_COCO'" on default_config.yaml file.
# Set "data_path='/cache/data'" on default_config.yaml file.
# Set "ann_file='./annotations/instances_val2017.json'" on default_config.yaml file.
# Set "epoch_size=12" on default_config.yaml file.
# Set "ckpt_path='./ckpt_maskrcnn/mask_rcnn-12_7393.ckpt'" on default_config.yaml file.
# (optional)Set "checkpoint_url='s3://dir_to_your_pretrained/'" on default_config.yaml file.
# Set other parameters on default_config.yaml file you need.
# b. Add "enable_modelarts=True" on the website UI interface.
# Add "need_modelarts_dataset_unzip=True" on the website UI interface.
# Add "modelarts_dataset_unzip_name='cocodataset'" on the website UI interface.
# Add "mindrecord_dir='./MindRecord_COCO'" on the website UI interface.
# Add "data_path='/cache/data'" on the website UI interface.
# Add "ann_file='./annotations/instances_val2017.json'" on the website UI interface.
# Add "epoch_size=12" on the website UI interface.
# Set "ckpt_path='./ckpt_maskrcnn/mask_rcnn-12_7393.ckpt'" on default_config.yaml file.
# (optional)Add "checkpoint_url='s3://dir_to_your_pretrained/'" on the website UI interface.
# Add other parameters on the website UI interface.
# (2) Prepare model code
# (3) Upload or copy your pretrained model to S3 bucket if you want to finetune.
# (4) Perform a or b. (suggested option a)
# a. First, run "train.py" like the following to create MindRecord dataset locally from coco2017.
# "python train.py --only_create_dataset=True --mindrecord_dir=$MINDRECORD_DIR --data_path=$DATA_PATH --ann_file=$ANNO_PATH"
# Second, zip MindRecord dataset to one zip file.
# Finally, Upload your zip dataset to S3 bucket.(you could also upload the origin mindrecord dataset, but it can be so slow.)
# b. Upload the original coco dataset to S3 bucket.
# (Data set conversion occurs during training process and costs a lot of time. it happens every time you train.)
# (5) Set the code directory to "/path/maskrcnn" on the website UI interface.
# (6) Set the startup file to "train.py" on the website UI interface.
# (7) Set the "Dataset path" and "Output file path" and "Job log path" to your path on the website UI interface.
# (8) Create your job.
#
# Eval 1p with Ascend
# (1) Perform a or b.
# a. Set "enable_modelarts=True" on default_config.yaml file.
# Set "need_modelarts_dataset_unzip=True" on default_config.yaml file.
# Set "modelarts_dataset_unzip_name='cocodataset'" on default_config.yaml file.
# Set "checkpoint_url='s3://dir_to_your_trained_model/'" on base_config.yaml file.
# Set "checkpoint_path='./ckpt_maskrcnn/mask_rcnn-12_7393.ckpt'" on default_config.yaml file.
# Set "mindrecord_file='/cache/data/cocodataset/MindRecord_COCO'" on default_config.yaml file.
# Set "data_path='/cache/data'" on default_config.yaml file.
# Set "ann_file='./annotations/instances_val2017.json'" on default_config.yaml file.
# Set other parameters on default_config.yaml file you need.
# b. Add "enable_modelarts=True" on the website UI interface.
# Add "need_modelarts_dataset_unzip=True" on the website UI interface.
# Add "modelarts_dataset_unzip_name='cocodataset'" on the website UI interface.
# Add "checkpoint_url='s3://dir_to_your_trained_model/'" on the website UI interface.
# Add "checkpoint_path='./ckpt_maskrcnn/mask_rcnn-12_7393.ckpt'" on the website UI interface.
# Set "mindrecord_file='/cache/data/cocodataset/MindRecord_COCO'" on default_config.yaml file.
# Add "data_path='/cache/data'" on the website UI interface.
# Set "ann_file='./annotations/instances_val2017.json'" on default_config.yaml file.
# Add other parameters on the website UI interface.
# (2) Prepare model code
# (3) Upload or copy your trained model to S3 bucket.
# (4) Perform a or b. (suggested option a)
# a. First, run "eval.py" like the following to create MindRecord dataset locally from coco2017.
# "python eval.py --only_create_dataset=True --mindrecord_dir=$MINDRECORD_DIR --data_path=$DATA_PATH --ann_file=$ANNO_PATH \
# --checkpoint_path=$CHECKPOINT_PATH"
# Second, zip MindRecord dataset to one zip file.
# Finally, Upload your zip dataset to S3 bucket.(you could also upload the origin mindrecord dataset, but it can be so slow.)
# b. Upload the original coco dataset to S3 bucket.
# (Data set conversion occurs during training process and costs a lot of time. it happens every time you train.)
# (5) Set the code directory to "/path/maskrcnn" on the website UI interface.
# (6) Set the startup file to "eval.py" on the website UI interface.
# (7) Set the "Dataset path" and "Output file path" and "Job log path" to your path on the website UI interface.
# (8) Create your job.
```
# [Script Description](#contents)
## [Script and Sample Code](#contents)
@ -503,7 +618,7 @@ Accumulating evaluation results...
## Model Export
```shell
python export.py --ckpt_file [CKPT_PATH] --device_target [DEVICE_TARGET] --file_format[EXPORT_FORMAT]
python export.py --config_path [CONFIG_FILE] --ckpt_file [CKPT_PATH] --device_target [DEVICE_TARGET] --file_format[EXPORT_FORMAT]
```
`EXPORT_FORMAT` should be in ["AIR", "MINDIR"]

115
model_zoo/official/cv/maskrcnn/README_CN.md
View File

@ -169,6 +169,119 @@ bash run_distribute_train.sh [RANK_TABLE_FILE] [PRETRAINED_CKPT]
bash run_eval.sh [VALIDATION_JSON_FILE] [CHECKPOINT_PATH]
```
- 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/))
```bash
# 在 ModelArts 上使用8卡训练
# (1) 执行a或者b
# a. 在 default_config.yaml 文件中设置 "enable_modelarts=True"
# 在 default_config.yaml 文件中设置 "distribute=True"
# 在 default_config.yaml 文件中设置 "need_modelarts_dataset_unzip=True"
# 在 default_config.yaml 文件中设置 "modelarts_dataset_unzip_name='cocodataset'"
# 在 default_config.yaml 文件中设置 "base_lr=0.02"
# 在 default_config.yaml 文件中设置 "mindrecord_dir='./MindRecord_COCO'"
# 在 default_config.yaml 文件中设置 "data_path='/cache/data'"
# 在 default_config.yaml 文件中设置 "ann_file='./annotations/instances_val2017.json'"
# 在 default_config.yaml 文件中设置 "epoch_size=12"
# 在 default_config.yaml 文件中设置 "ckpt_path='./ckpt_maskrcnn/mask_rcnn-12_7393.ckpt'"
# (可选)在 default_config.yaml 文件中设置 "checkpoint_url='s3://dir_to_your_pretrained/'"
# 在 default_config.yaml 文件中设置 其他参数
# b. 在网页上设置 "enable_modelarts=True"
# 在网页上设置 "need_modelarts_dataset_unzip=True"
# 在网页上设置 "modelarts_dataset_unzip_name='cocodataset'"
# 在网页上设置 "distribute=True"
# 在网页上设置 "base_lr=0.02"
# 在网页上设置 "mindrecord_dir='./MindRecord_COCO'"
# 在网页上设置 "data_path='/cache/data'"
# 在网页上设置 "ann_file='./annotations/instances_val2017.json'"
# 在网页上设置 "epoch_size=12"
# 在网页上设置 "ckpt_path='./ckpt_maskrcnn/mask_rcnn-12_7393.ckpt'"
# (可选)在网页上设置 "checkpoint_url='s3://dir_to_your_pretrained/'"
# 在网页上设置 其他参数
# (2) 如果选择微调您的模型，请上传你的预训练模型到 S3 桶上
# (3) 执行a或者b (推荐选择 a)
# a. 第一, 根据以下方式在本地运行 "train.py" 脚本来生成 MindRecord 格式的数据集。
# "python train.py --only_create_dataset=True --mindrecord_dir=$MINDRECORD_DIR --data_path=$DATA_PATH --ann_file=$ANNO_PATH"
# 第二, 将该数据集压缩为一个 ".zip" 文件。
# 最后, 上传你的压缩数据集到 S3 桶上 (你也可以上传未压缩的数据集，但那可能会很慢。)
# b. 上传原始 coco 数据集到 S3 桶上。
# (数据集转换发生在训练过程中，需要花费较多的时间。每次训练的时候都会重新进行转换。)
# (4) 在网页上设置你的代码路径为 "/path/maskrcnn"
# (5) 在网页上设置启动文件为 "train.py"
# (6) 在网页上设置"训练数据集"、"训练输出文件路径"、"作业日志路径"等
# (7) 创建训练作业
#
# 在 ModelArts 上使用单卡训练
# (1) 执行a或者b
# a. 在 default_config.yaml 文件中设置 "enable_modelarts=True"
# 在 default_config.yaml 文件中设置 "need_modelarts_dataset_unzip=True"
# 在 default_config.yaml 文件中设置 "modelarts_dataset_unzip_name='cocodataset'"
# 在 default_config.yaml 文件中设置 "mindrecord_dir='./MindRecord_COCO'"
# 在 default_config.yaml 文件中设置 "data_path='/cache/data'"
# 在 default_config.yaml 文件中设置 "ann_file='./annotations/instances_val2017.json'"
# 在 default_config.yaml 文件中设置 "epoch_size=12"
# 在 default_config.yaml 文件中设置 "ckpt_path='./ckpt_maskrcnn/mask_rcnn-12_7393.ckpt'"
# (可选)在 default_config.yaml 文件中设置 "checkpoint_url='s3://dir_to_your_pretrained/'"
# 在 default_config.yaml 文件中设置 其他参数
# b. 在网页上设置 "enable_modelarts=True"
# 在网页上设置 "need_modelarts_dataset_unzip=True"
# 在网页上设置 "modelarts_dataset_unzip_name='cocodataset'"
# 在网页上设置 "mindrecord_dir='./MindRecord_COCO'"
# 在网页上设置 "data_path='/cache/data'"
# 在网页上设置 "ann_file='./annotations/instances_val2017.json'"
# 在网页上设置 "epoch_size=12"
# 在网页上设置 "ckpt_path='./ckpt_maskrcnn/mask_rcnn-12_7393.ckpt'"
# (可选)在网页上设置 "checkpoint_url='s3://dir_to_your_pretrained/'"
# 在网页上设置 其他参数
# (2) 如果选择微调您的模型，上传你的预训练模型到 S3 桶上
# (3) 执行a或者b (推荐选择 a)
# a. 第一, 根据以下方式在本地运行 "train.py" 脚本来生成 MindRecord 格式的数据集。
# "python train.py --only_create_dataset=True --mindrecord_dir=$MINDRECORD_DIR --data_path=$DATA_PATH --ann_file=$ANNO_PATH"
# 第二, 将该数据集压缩为一个 ".zip" 文件。
# 最后, 上传你的压缩数据集到 S3 桶上 (你也可以上传未压缩的数据集，但那可能会很慢。)
# b. 上传原始 coco 数据集到 S3 桶上。
# (数据集转换发生在训练过程中，需要花费较多的时间。每次训练的时候都会重新进行转换。)
# (4) 在网页上设置你的代码路径为 "/path/maskrcnn"
# (5) 在网页上设置启动文件为 "train.py"
# (6) 在网页上设置"训练数据集"、"训练输出文件路径"、"作业日志路径"等
# (7) 创建训练作业
#
# 在 ModelArts 上使用单卡验证
# (1) 执行a或者b
# a. 在 default_config.yaml 文件中设置 "enable_modelarts=True"
# 在 default_config.yaml 文件中设置 "need_modelarts_dataset_unzip=True"
# 在 default_config.yaml 文件中设置 "modelarts_dataset_unzip_name='cocodataset'"
# 在 default_config.yaml 文件中设置 "checkpoint_url='s3://dir_to_your_trained_model/'"
# 在 default_config.yaml 文件中设置 "checkpoint_path='./ckpt_maskrcnn/mask_rcnn-12_7393.ckpt'"
# 在 default_config.yaml 文件中设置 "mindrecord_file='/cache/data/cocodataset/MindRecord_COCO'"
# 在 default_config.yaml 文件中设置 "data_path='/cache/data'"
# 在 default_config.yaml 文件中设置 "ann_file='./annotations/instances_val2017.json'"
# 在 default_config.yaml 文件中设置 其他参数
# b. 在网页上设置 "enable_modelarts=True"
# 在网页上设置 "need_modelarts_dataset_unzip=True"
# 在网页上设置 "modelarts_dataset_unzip_name='cocodataset'"
# 在网页上设置 "checkpoint_url='s3://dir_to_your_trained_model/'"
# 在网页上设置 "checkpoint_path='./ckpt_maskrcnn/mask_rcnn-12_7393.ckpt'"
# 在网页上设置 "mindrecord_file='/cache/data/cocodataset/MindRecord_COCO'"
# 在网页上设置 "data_path='/cache/data'"
# 在网页上设置 "ann_file='./annotations/instances_val2017.json'"
# 在网页上设置 其他参数
# (2) 上传你训练好的模型到 S3 桶上
# (3) 执行a或者b (推荐选择 a)
# a. 第一, 根据以下方式在本地运行 "eval.py" 脚本来生成 MindRecord 格式的数据集。
# "python eval.py --only_create_dataset=True --mindrecord_dir=$MINDRECORD_DIR --data_path=$DATA_PATH --ann_file=$ANNO_PATH \
# --checkpoint_path=$CHECKPOINT_PATH"
# 第二, 将该数据集压缩为一个 ".zip" 文件。
# 最后, 上传你的压缩数据集到 S3 桶上 (你也可以上传未压缩的数据集，但那可能会很慢。)
# b. 上传原始 coco 数据集到 S3 桶上。
# (数据集转换发生在训练过程中，需要花费较多的时间。每次训练的时候都会重新进行转换。)
# (4) 在网页上设置你的代码路径为 "/path/maskrcnn"
# (5) 在网页上设置启动文件为 "eval.py"
# (6) 在网页上设置"训练数据集"、"训练输出文件路径"、"作业日志路径"等
# (7) 创建训练作业
```
# 脚本说明
## 脚本和样例代码
@ -501,7 +614,7 @@ sh run_eval.sh [VALIDATION_ANN_FILE_JSON] [CHECKPOINT_PATH]
## 模型导出
```shell
python export.py --ckpt_file [CKPT_PATH] --device_target [DEVICE_TARGET] --file_format[EXPORT_FORMAT]
python export.py --config_path [CONFIG_PATH] --ckpt_file [CKPT_PATH] --device_target [DEVICE_TARGET] --file_format[EXPORT_FORMAT]
```
`EXPORT_FORMAT` 选项 ["AIR", "MINDIR"]

27
model_zoo/official/gnn/gat/README.md
View File

@ -114,6 +114,29 @@ After installing MindSpore via the official website and Dataset is correctly gen
sh run_train_ascend.sh [DATASET_NAME]
```
- Running on [ModelArts](https://support.huaweicloud.com/modelarts/)
```bash
# Train/eval 1p with Ascend
# (1) Perform a or b.
# a. Set "enable_modelarts=True" on default_config.yaml file.
# Set "data_dir='/cache/data'" on default_config.yaml file.
# (optional)Set "checkpoint_url='s3://dir_to_your_pretrained/'" on default_config.yaml file.
# Set other parameters on default_config.yaml file you need.
# b. Add "enable_modelarts=True" on the website UI interface.
# Add "data_dir='/cache/data'" on the website UI interface.
# (optional)Add "checkpoint_url='s3://dir_to_your_pretrained/'" on the website UI interface.
# Add other parameters on the website UI interface.
# (3) Upload or copy your pretrained model to S3 bucket if you want to finetune.
# (4) Upload the original Cora/Citeseer dataset to S3 bucket.
# (5) Set the code directory to "/path/gat" on the website UI interface.
# (6) Set the startup file to "train.py" on the website UI interface.
# (7) Set the "Dataset path" and "Output file path" and "Job log path" to your path on the website UI interface.
# (8) Create your job.
#
```
## [Script Description](#contents)
## [Script and Sample Code](#contents)
@ -142,7 +165,7 @@ After installing MindSpore via the official website and Dataset is correctly gen
## [Script Parameters](#contents)
Parameters for both training and evaluation can be set in config.py.
Parameters for both training and evaluation can be set in default_config.yaml.
- config for GAT, CORA dataset
@ -191,7 +214,7 @@ Parameters for both training and evaluation can be set in config.py.
### [Export MindIR](#contents)
```shell
python export.py --ckpt_file [CKPT_PATH] --file_name [FILE_NAME] --file_format [FILE_FORMAT]
python export.py --config_path [CONFIG_PATH]--ckpt_file [CKPT_PATH] --file_name [FILE_NAME] --file_format [FILE_FORMAT]
```
The ckpt_file parameter is required,

21
model_zoo/official/gnn/gat/README_CN.md
View File

@ -112,6 +112,27 @@
sh run_train_ascend.sh [DATASET_NAME]
```
- 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/))
```bash
# 在 ModelArts 上使用单卡训练验证
# (1) 执行a或者b
# a. 在 default_config.yaml 文件中设置 "enable_modelarts=True"
# 在 default_config.yaml 文件中设置 "data_dir='/cache/data'"
# (可选)在 default_config.yaml 文件中设置 "checkpoint_url='s3://dir_to_your_pretrained/'"
# 在 default_config.yaml 文件中设置 其他参数
# b. 在网页上设置 "enable_modelarts=True"
# 在网页上设置 "data_dir='/cache/data'"
# (可选)在网页上设置 "checkpoint_url='s3://dir_to_your_pretrained/'"
# 在网页上设置 其他参数
# (3) 如果选择微调您的模型，上传你的预训练模型到 S3 桶上
# (4) 上传原始 Cora/Citeseer 数据集到 S3 桶上。
# (5) 在网页上设置你的代码路径为 "/path/gat"
# (6) 在网页上设置启动文件为 "train.py"
# (7) 在网页上设置"训练数据集"、"训练输出文件路径"、"作业日志路径"等
# (8) 创建训练作业
```
## 脚本说明
### 脚本及样例代码