base.py

from yacs.config import CfgNode

_C = CfgNode()

_C.DATA = CfgNode()
_C.DATA.IMAGE_SIZE = 256
_C.DATA.IN_CHANS = 3
_C.DATA.NUM_CLASSES = 0
_C.DATA.CLASS_FREQ = None
_C.DATA.MEAN = 0.
_C.DATA.STD = 1.

_C.MODEL = CfgNode()

_C.MODEL.ENCODER = CfgNode()
_C.MODEL.ENCODER.TYPE = None
_C.MODEL.ENCODER.NAME = None
_C.MODEL.ENCODER.PATCH_SIZE = 4
_C.MODEL.ENCODER.WINDOW_SIZE = None
_C.MODEL.ENCODER.GROUP_SIZE = None
_C.MODEL.ENCODER.SPLIT_SIZE = None
_C.MODEL.ENCODER.DEPTHS = None
_C.MODEL.ENCODER.NUM_PATHS = None
_C.MODEL.ENCODER.NUM_HEADS = None
_C.MODEL.ENCODER.EMBED_DIM = None
_C.MODEL.ENCODER.MLP_RATIO = None
_C.MODEL.ENCODER.SR_RATIOS = None
_C.MODEL.ENCODER.KV_SCALE = None
_C.MODEL.ENCODER.QK_SCALE = None
_C.MODEL.ENCODER.QKV_BIAS = None
_C.MODEL.ENCODER.DROP_RATE = 0.0
_C.MODEL.ENCODER.DROP_PATH_RATE = 0.1
_C.MODEL.ENCODER.ATTN_DROP_RATE = 0.0
_C.MODEL.ENCODER.USE_ABS_POS_EMBED = False
_C.MODEL.ENCODER.USE_REL_POS_EMBED = False
_C.MODEL.ENCODER.USE_BN_IN_ATTN = False
_C.MODEL.ENCODER.USE_MULTI_MERGE = False
_C.MODEL.ENCODER.USE_GHOST_FFN = False
_C.MODEL.ENCODER.USE_IRFFN = False
_C.MODEL.ENCODER.USE_LPU = False

_C.MODEL.DECODER = CfgNode()
_C.MODEL.DECODER.TYPE = None
_C.MODEL.DECODER.NAME = None
_C.MODEL.DECODER.DEPTH = None
_C.MODEL.DECODER.FUSE_OP = None
_C.MODEL.DECODER.IN_FEATURES = None
_C.MODEL.DECODER.EMBED_DIM = None

_C.TRAIN = CfgNode()
_C.TRAIN.EPOCHS = 100
_C.TRAIN.WARMUP_EPOCHS = 3
_C.TRAIN.WEIGHT_DECAY = 1e-2
_C.TRAIN.LR = 3e-4
_C.TRAIN.LR_DECAY = 0.1
_C.TRAIN.LR_STEP = 1
_C.TRAIN.LR_SCHEDULER = 'poly'
_C.TRAIN.CLIP_GRAD = 1.0