Can update a config_dict from yaml/json file?

[Chutlhu]

Hi all, thank you for this nice library.

I am using config_dicts in my ML pipeline as explained in the README.
I currently use the CLI interface to run different ML models with different parameters.
Instead, is it possible to update a ConfigDict using a yaml config file generated via ConfigDict().to_yaml()?

[xk-huang]

I use OmegaConf to achieve this:

import os.path as osp
from omegaconf import OmegaConf

FLAGS_DEF = {}
condig_root_name="src.config"
for flag_values in FLAGS.get_flags_for_module(condig_root_name):
    # NOTE xk: flag_values.name is the key, flag_values.value is the ConfigDict, which has `to_dict`, `to_yaml` methods.
    FLAGS_DEF[flag_values.name] = flag_values.value

config = OmegaConf.create({k: v.to_dict() for k, v in FLAGS_DEF.items()})

# save config
config_path = osp.join(config.main.ckpt_dir, "config.yaml")
OmegaConf.save(config=config, f=config_path)

[Conchylicultor]

You can simply pass the json dict to the ConfigDict constructor:

cfg = ConfigDict(json.loads(data))