super-gradients/src/super_gradients/recipes/variable_setup.yaml

# Varaible setup for shortcuts and setting the hydra output directory.
# Any SG Recipe should set this yaml file as a default, after _self_, i.e at the top of your recipe file:
#
#  defaults:
#    - training_hyperparams: my_train_params
#    - dataset_params: my_dataset_params
#    - arch_params: my_arch_params
#    - checkpoint_params: my_checkpoint_params
#    - _self_
#    - variable_setup
#
#
#  Interpolates the shortcuts defined below, with their aliases (see comments near each parameter).
#  When any of the above are not set, they will be populated with the original values (for example
#      config.lr will be set with config.training_hyperparams.initial_lr) for clarity in logs.
#
#  In other words, the following training launch commands are equivalent:
#
#    python train_from_recipe --config-name=recipe lr=0.003
#
#    python train_from_recipe --config-name=recipe config.training_hyperparams.initial_lr=0.003
#
# Note that interpolation is done by triggering RecipeShortcutsCallbackm which is a Hydra Callback (see http://hydra.cc/docs/experimental/callbacks/)
#  so interpolation of these in other yaml configuration files won't be present.

lr: # config.training_hyperparams.initial_lr

batch_size: # config.dataset_params.train_dataloader_params.batch_size

val_batch_size: # config.dataset_params.val_dataloader_params.batch_size

ema: # config.training_hyperparams.ema

epochs: # config.training_hyperparams.max_epochs

resume: # config.training_hyperparams.resume

num_workers: # config.dataset_params.train_dataloader_params.num_workers and config.dataset_params.val_dataloader_params.num_workers

ckpt_root_dir:
# THE FOLLOWING PARAMS ARE DIRECTLY USED BY HYDRA
hydra:
  callbacks:
    shortcuts_cb:
      _target_: super_gradients.common.environment.omegaconf_utils.RecipeShortcutsCallback
  run:
    # Set the output directory (i.e. where .hydra folder that logs all the input params will be generated)
    dir: ${hydra_output_dir:${ckpt_root_dir}, ${experiment_name}}