/yamlpyconfig

Library for parsing yaml config files in ML training pipelines

Primary LanguagePython

yamlpyconfig

Library for parsing yaml config files in ML training pipelines. Based on Artur Kuzin talk "DL Pipelines tips & tricks".

Benefits:

  1. Config files have yaml format.
  2. All parameters in python code have form 'cfg.par_name'.
  3. Any parameter could be set in command line prompt.

There are 3 levels of configuration:

  1. Default config file is set during initialization:

    • yconf = YamlPyConfig('default_config', , conf_path='./configs')
    • Config file 'default_config.yaml' should be placed in './configs' folder.
    • python train_example.py

  2. Custom config file could be set with parameter --config:

    • python train_example.py --config custom_config

  3. Custom parameters could be set with appropriate keys:

    • python train_example.py --config custom_config --epochs 100 --lr 0.3

Installing Dependencies :

pip install -r requirements.txt