71 lines
1.7 KiB
Markdown
71 lines
1.7 KiB
Markdown
# ModYml
|
|
|
|
## 📦 Project Introduction
|
|
|
|
ModYml is a lightweight Python configuration management library designed for modular and reusable configurations. With YAML file support, it enables configuration inheritance, parameter overriding, and alias systems, making complex configuration management simple and intuitive.
|
|
|
|
## ✨ Key Features
|
|
|
|
- **Modular Configs**: Split configurations into reusable modules
|
|
- **Dynamic Inheritance**: Reference other configs using `$module.key` syntax
|
|
- **Parameter Overriding**: Modify parameters at runtime with `$module.key(param=value)`
|
|
- **Alias System**: Create short aliases for config items `key(alias)`
|
|
- **Namespace Isolation**: Automatic namespace management for config items
|
|
- **Dot-access**: Returns `EasyDict` objects for attribute-style access
|
|
|
|
## ⚙️ Installation
|
|
|
|
```bash
|
|
pip install modyml
|
|
```
|
|
|
|
## 🚀 Quick Start
|
|
|
|
1. Create module config (`modules/optimizers.yml`):
|
|
|
|
```yaml
|
|
adam:
|
|
type: Adam
|
|
params:
|
|
learning_rate(lr): 0.001
|
|
weight_decay(wd): 0.0001
|
|
```
|
|
|
|
2. Create experiment config (`experiments/E01.yml`):
|
|
|
|
```yaml
|
|
training:
|
|
base_optimizer: $optimizers.adam
|
|
custom_optimizer: $optimizers.adam(lr=0.002)
|
|
special_optimizer($optimizers.adam):
|
|
params:
|
|
wd: 0.01
|
|
```
|
|
|
|
3. Load configuration:
|
|
|
|
```python
|
|
from modyml import load
|
|
|
|
config = load(
|
|
"experiments/E01.yml",
|
|
module_files=["modules/optimizers.yml"],
|
|
base_dir="configs/"
|
|
)
|
|
|
|
print(config.training.base_optimizer.params.lr) # Output: 0.001
|
|
print(config.training.custom_optimizer.params.lr) # Output: 0.002
|
|
print(config.training.special_optimizer.params.wd) # Output: 0.01
|
|
```
|
|
|
|
## 🧪 Testing
|
|
|
|
```bash
|
|
pytest tests/
|
|
```
|
|
|
|
---
|
|
|
|
**Author**: Yuyao Huang (Sam)
|
|
**License**: MIT
|