def get_transforms(*, data):
    
    if data == 'train':
        return Compose([
            Resize(CFG.size, CFG.size),
            albumentations.HorizontalFlip(p=0.5),
            albumentations.RandomBrightness(limit=0.2, p=0.75),
            albumentations.RandomContrast(limit=0.2, p=0.75),
            HorizontalFlip(p=0.5),
            albumentations.HueSaturationValue(hue_shift_limit=40, sat_shift_limit=40, val_shift_limit=0, p=0.75),
            albumentations.ShiftScaleRotate(shift_limit=0.2, scale_limit=0.3, rotate_limit=5, border_mode=0, p=0.75),
            CutoutV2(max_h_size=int(CFG.size * 0.2), max_w_size=int(CFG.size * 0.2), num_holes=1, p=0.75),# VerticalFlip(p=0.5),
            # ShiftScaleRotate(p=0.5),
            Normalize(
                mean=[0.485, 0.456, 0.406],
                std=[0.229, 0.224, 0.225],
            ),
            ToTensorV2(),
        ])

    elif data == 'valid':
        return Compose([
            Resize(CFG.size, CFG.size),
            Normalize(
                mean=[0.485, 0.456, 0.406],
                std=[0.229, 0.224, 0.225],
            ),
            ToTensorV2(),
        ])

class CFG:
    debug=False
    apex=False
    print_freq=100
    num_workers=8
    model_name='swin_large_patch4_window12_384'
    size=384
    scheduler='CosineAnnealingWarmRestarts' # ['ReduceLROnPlateau', 'CosineAnnealingLR', 'CosineAnnealingWarmRestarts']
    epochs=8
    #factor=0.2 # ReduceLROnPlateau
    #patience=4 # ReduceLROnPlateau
    #eps=1e-6 # ReduceLROnPlateau
    #T_max=10 # CosineAnnealingLR
    T_0=10 # CosineAnnealingWarmRestarts
    lr=1e-4
    min_lr=1e-6
    batch_size=12
    weight_decay=1e-6
    gradient_accumulation_steps=1
    max_grad_norm=1000
    seed=42
    target_size=6
    target_col='class_6'
    n_fold=5
    trn_fold=[0, 1, 2, 3, 4]
    train=True
    inference=False

https://www.kaggle.com/yasufuminakama/cassava-resnext50-32x4d-starter-training