未验证 提交 2bd66c11 编写于 作者: B Bin Lu 提交者: GitHub

Update feature_extraction.md

上级 8decca58
......@@ -113,14 +113,12 @@ python -m paddle.distributed.launch
推理过程包括两个步骤: 1) 导出推理模型; 2) 获取特征向量
### 4.4.1 导出推理模型
```
python tools/export_model -c xxx -o Global.pretrained_model = xxxx
python tools/export_model -c ppcls/configs/ResNet50_vd_SOP.yaml -o Global.pretrained_model = "output/ReModel/best_model"
```
生成的推理模型位于inference目录,名字为inference.pd*
### 4.4.2 获取特征向量
```
cd deploy
python python/inference_rec.py -c configs/ O rec_inference_model_dir: "../inference/inference"
```
输出
python python/inference_rec.py -c configs/inference_product.yaml -o rec_inference_model_dir="../inference/inference"
```
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册