refine doc for transfer learning (#3642)

* refine doc for transfer learning

docs/GETTING_STARTED.md

@@ -70,7 +70,10 @@ python -u tools/train.py -c configs/faster_rcnn_r50_1x.yml \
 
 - Fine-tune other task
 
-When using pre-trained model to fine-tune other task, the excluded pre-trained parameters can be set by finetune_exclude_pretrained_params in YAML config or -o finetune_exclude_pretrained_params in the arguments.
+When using pre-trained model to fine-tune other task, two methods can be used:
+
+1. The excluded pre-trained parameters can be set by `finetune_exclude_pretrained_params` in YAML config
+2. Set -o finetune_exclude_pretrained_params in the arguments.
 
 ```bash
 export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7

docs/GETTING_STARTED_cn.md

@@ -71,7 +71,10 @@ python -u tools/train.py -c configs/faster_rcnn_r50_1x.yml \
 
 - Fine-tune其他任务
 
-使用预训练模型fine-tune其他任务时，在YAML配置文件中设置`finetune_exclude_pretrained_params`或在命令行中添加`-o finetune_exclude_pretrained_params`对预训练模型进行选择性加载。
+使用预训练模型fine-tune其他任务时，可采用如下两种方式：
+
+1. 在YAML配置文件中设置`finetune_exclude_pretrained_params`
+2. 在命令行中添加-o finetune_exclude_pretrained_params对预训练模型进行选择性加载。
 
 ```bash
 export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7

docs/TRANSFER_LEARNING.md

@@ -6,7 +6,10 @@ In transfer learning, if different dataset and the number of classes is used, th
 
 ## Transfer Learning in PaddleDetection
 
-In transfer learning, it's needed to load pretrained model selectively. Set `finetune_exclude_pretrained_params` in YAML configuration files or set `-o finetune_exclude_pretrained_params` in command line.
+In transfer learning, it's needed to load pretrained model selectively. The following two methods can be used:
+
+1. Set `finetune_exclude_pretrained_params` in YAML configuration files. Please refer to [configure file](../configs/yolov3_mobilenet_v1_fruit.yml#L15)
+2. Set -o finetune_exclude_pretrained_params in command line. For example:
 
 ```python
 export PYTHONPATH=$PYTHONPATH:.

docs/TRANSFER_LEARNING_cn.md

@@ -6,7 +6,10 @@
 
 ## PaddleDetection进行迁移学习
 
-在迁移学习中，对预训练模型进行选择性加载，可通过在 YMAL 配置文件中通过设置 finetune_exclude_pretrained_params字段，也可通过在 train.py的启动参数中设置 -o finetune_exclude_pretrained_params。
+在迁移学习中，对预训练模型进行选择性加载，可通过如下两种方式实现：
+
+1. 在 YMAL 配置文件中通过设置`finetune_exclude_pretrained_params`字段。可参考[配置文件](../configs/yolov3_mobilenet_v1_fruit.yml#L15)
+2. 在 train.py的启动参数中设置 -o finetune_exclude_pretrained_params。例如：
 
 ```python
 export PYTHONPATH=$PYTHONPATH:.

ppdet/utils/checkpoint.py

@@ -86,6 +86,8 @@ def load_params(exe, prog, path, ignore_params=[]):
         prog (fluid.Program): load weight to which Program object.
         path (string): URL string or loca model path.
         ignore_params (bool): ignore variable to load when finetuning.
+            It can be specified by finetune_exclude_pretrained_params 
+            and the usage can refer to docs/TRANSFER_LEARNING.md
     """
 
     if is_url(path):