cli_args_examples.sh

# Three groups of parameters:
#     1. trainer_params
#     2. dataset_params
#     3. model_params

# python run_omega.py \
# mode='train,val,test' \
# trainer_params.gpus=[5]


# When you specify model_type, it sets
#     model_params.model_config
# to the corresponding dict in Config.model_type_lookup()

# python run_omega.py \
# mode='train,val,test'\
# model_type='memory' \
# trainer_params.gpus=[5] \
# model_params.lr=1e-5 \


# You can change model_params.model_config values individually 
# after specifying a model_type (or leaving it as baseline)

# python run_omega.py \
# mode='train,val,test'\
# model_type='memory' \
# trainer_params.gpus=[5] \
# model_params.lr=1e-5 \
# model_params.model_config.dropout=0.02


# dataset_params are changed indidually for each 'mode'

# python run_omega.py \
# mode='train,val,test'\
# model_type='memory' \
# trainer_params.gpus=[5] \
# dataset_params.train.shuffle=True
# dataset_params.val.shuffle=False


# You can add arguments that don't have defaults in Config
# (min_epochs is a Trainer param)

python run_omega.py \
mode='train,val,test' \
trainer_params.min_epochs=5