from transformers import PretrainedConfig

class PATHViTConfig(PretrainedConfig):
    model_type = "vit-b16"

    def __init__(
        self,
        img_size=224,
        patch_size=16,
        in_chans=3,
        num_classes=80,
        embed_dim=768,
        depth=12,
        num_heads=12,
        mlp_ratio=4.0,
        qkv_bias=True,
        drop_path_rate=0.1,
        norm_layer=None,
        norm_layer_eps=1e-6,
        window=True,
        use_abs_pos_emb=True,
        interval=3,
        test_pos_mode=False,
        task_sp_list=(),
        neck_sp_list=(),
        learnable_pos=False,
        rel_pos_spatial=False,
        lms_checkpoint_train=False,
        prompt=None,
        pad_attn_mask=False,
        freeze_iters=0,
        act_layer="GELU",
        pre_ln=False,
        mask_input=False,
        ending_norm=True,
        round_padding=False,
        compat=False,
        use_cls_token=False,
        **kwargs,
    ):
        super().__init__(**kwargs)
        self.img_size = img_size
        self.patch_size = patch_size
        self.in_chans = in_chans
        self.num_classes = num_classes
        self.embed_dim = embed_dim
        self.depth = depth
        self.num_heads = num_heads
        self.mlp_ratio = mlp_ratio
        self.qkv_bias = qkv_bias
        self.drop_path_rate = drop_path_rate
        # NOTE: norm_layer is not used for building the model
        self.norm_layer = norm_layer
        self.norm_layer_eps = norm_layer_eps
        
        self.window = window
        self.use_abs_pos_emb = use_abs_pos_emb
        self.interval = interval
        self.test_pos_mode = test_pos_mode
        self.task_sp_list = task_sp_list
        self.neck_sp_list = neck_sp_list
        self.learnable_pos = learnable_pos
        self.rel_pos_spatial = rel_pos_spatial
        self.lms_checkpoint_train = lms_checkpoint_train
        self.prompt = prompt
        self.pad_attn_mask = pad_attn_mask
        self.freeze_iters = freeze_iters
        self.act_layer = act_layer
        self.pre_ln = pre_ln
        self.mask_input = mask_input
        self.ending_norm = ending_norm
        self.round_padding = round_padding
        self.compat = compat
        self.use_cls_token = use_cls_token