aws / aws/sagemaker-python-sdk
Make cli handling more useful
未关闭
type: feature request
- 主要语言
- Python
- 星标
- 2.3k
- 派生
- 1.3k
- 平均合并
- 1 天 22 小时
- 30 天内合并 PR
- 35
描述
Right now, all the training and inference cli is handled through `argparse` which is in battery-included default python but is hard to work with for slightly more complicated cmd args, such as subcommands, lack of type checking etc. There're some nice cli packages in Python ecosystem. One of them is [click](https://click.palletsprojects.com/en/7.x/) that handles the mentioned shortcomings nicely, however, simply replacing an argparse cli with a click equivalent doesn't work in SageMaker Estimator hyperparameters, which is enforcing all the passed args to be string more or less (even booleans don't work by default).
So it'd great to support click.
贡献指南
评估
这个 Issue 还没有评估数据。