Fix training_params deprecation

[Louis-Dupont]

Reproduce

from omegaconf import OmegaConf
from hydra.utils import instantiate

from super_gradients import Trainer
from super_gradients.training import models
from super_gradients.common.object_names import Models
from super_gradients.common.environment.cfg_utils import load_recipe
from super_gradients.training.dataloaders import coco2017_train, coco2017_val

train_loader = coco2017_train(dataset_params={"max_num_samples": 400})
val_loader = coco2017_val(dataset_params={"max_num_samples": 400})

yolonas_recipe = load_recipe(config_name="coco2017_yolo_nas_s", overrides=[f"arch_params.num_classes=80", "training_hyperparams.max_epochs=1", "training_hyperparams.mixed_precision=False"])
yolonas_recipe = OmegaConf.to_container(instantiate(yolonas_recipe))
training_params = yolonas_recipe['training_hyperparams']

trainer = Trainer(experiment_name="yolonas_training_with_adapter", ckpt_root_dir="./")
model = models.get(model_name=Models.YOLO_NAS_S, num_classes=80, pretrained_weights="coco")

trainer.train(model=model, training_params=training_params, train_loader=train_loader, valid_loader=val_loader)

Explanation

It looks like criterion_params still exists with this code, but is empty. this may be due to default in yamls.
Instead of checking if criterion_params exists, I changed the code to check if criterion_params is empty

[BloodAxe]

LGTM