XNAS 是一个轻量级高效的框架,囊括目前广泛使用的搜索空间(cell-based, chain-structure)以及搜索算法(darts),并且针对部件进行封装,让新手可以快速的上手开发自己的NAS搜索算法。
入口文件为 `train_darts.py` ` `以及 ` ` `configs/search/darts.yaml` ` `。 ` ` `train_darts.py` ` ` 从 ` ` `configs/search/darts.yaml`
当中读取相关超参数,来进行环境设置,数据加载以及优化器的选择以及构建。
具体运行方法为:
cd xx(项目地址); PYTHONPATH=./ python tools/train_darts.py --cfg configs/search/darts.yaml
假设我们需要对darts.yaml文件中的超参数进行覆盖,那么利用key value来进行额外指定。比如在下面的命令中,会将OUT_DIR这个参数由原来的/userhome/project/XNAS/experiment/darts/test 覆盖为 /userhome/project/XNAS/experiment/darts/test1, 其他参数也同理,由于使用了yacs,相关的库会自动进行解析,例如 key True 也是一个正确的用法。
cd /userhome/project/XNAS; PYTHONPATH=./ python tools/train_darts.py --cfg configs/search/darts.yaml OUT_DIR /userhome/project/XNAS/experiment/darts/test1
Dear OpenI User
Thank you for your continuous support to the Openl Qizhi Community AI Collaboration Platform. In order to protect your usage rights and ensure network security, we updated the Openl Qizhi Community AI Collaboration Platform Usage Agreement in January 2024. The updated agreement specifies that users are prohibited from using intranet penetration tools. After you click "Agree and continue", you can continue to use our services. Thank you for your cooperation and understanding.
For more agreement content, please refer to the《Openl Qizhi Community AI Collaboration Platform Usage Agreement》