How to finetune with charged dataset #1423

Hi, i found two parts mentions about finetune with charge state.
first is:
python run_train.py
--train_file=data.xyz
--embedding_specs='{"charge": {"embed_type": "continuous", "min": -5, "max": 5}, "spin": {"embed_type": "continuous", "min": 0, "max": 4}, "temperature": {"embed_type": "continuous", "min": 0, "max": 3000}}'
--use_embedding_readout
--model=MACE

Second is:
parser.add_argument(
"--total_charge_key",
help="Key of total charge in training xyz",
type=str,
default=DefaultKeys.TOTAL_CHARGE.value,

I am not quite sure what's the difference. Which one i should use. What i am trying is:
python ./mace/scripts/run_train.py --name="MACE_on_big_data"--train_file="train.xyz" --valid_file="valid.xyz" --test_file="test.xyz" --foundation_model="mace-mpa-0-medium.model" --batch_size=32 --max_num_epochs=100 --stage_two --start_stage_two=60 --ema --ema_decay=0.99 --amsgrad --error_table='PerAtomMAE' --device=cuda --seed=123 --wandb --wandb_project="mace_defect" --wandb_name="MACE_on_big_data" --energy_key="REF_energy" --forces_key="REF_forces" --stress_key="REF_stress" --total_charge_key="charge" --E0s="estimated"
is it right? or i need to add "--embedding_specs='{"charge":{"per":"graph"}}'" and remove --total_charge_key

Thank you so much if you could help!!