This is the project template for tensorflow-2.0 based projects.
pipeline如下：
模型定义，数据读取，optimizer，训练过程，metric，validate
文件分布如下：
- utils
- config.py: process configs | 从json文件处理为dict，或者bunch
- dirs.py: process dirs | 处理和dir相关的
- logger.py: wrapper of tensorboard writer |
- utils.py: process args
- mains
- main.py: train & validate the model
- models
- model.py: model definition
- trainers
- trainer.py: define train_step and train_epoch
- configs
- config.json
- data_loader
- data_generator.py
