Hello, ice-learn! 你好！ice-learn！
ice-learn is a modular deep learning training framework that wraps pytorch-based workflows for ease of use. The main goal of ice-learn is to minimize boilerplate code, maximize scalability, readability, experimental reproducibility, and facilitate rapid development.
ice-learn 是一个模块化的深度学习训练框架，她对基于 pytorch 的工作流作了上层的包装，以方便使用。ice-learn 的主要目标是最小化辅助代码，最大化可扩展性、可读性、实验可复现性，帮助快速开发。
Quick Start 快速开始
Here, a short example will be included to show you the basic concepts and usage of the ice-learn program.
在这里，将包含一个简短的示例，向您说明 ice-learn 程序的基本概念和使用方法。
User Guide 用户指南
- How to easily use command line parameters? 如何快捷使用命令行参数？
- How to easily find global variables like global steps? 如何快捷使用具有全局意义的变量（例如总步长）？
- How to customize learning rate strategy? 如何设置学习率策略？
- How to extend Pytorch using CUDA in ice-learn? 在ice中如何用 cuda 扩展 Pytorch ?
API Reference Manual 参考手册
Here are the lower-level code-level comments, you can search for the signature and usage of a specific function. 这里是较为底层的代码级注释，您可以搜索具体某个函数的签名和用法。