configuration_tips.py

"""TIPSv2 model configuration."""

from transformers import PretrainedConfig


class TIPSv2ImageConfig(PretrainedConfig):
    """Configuration for TIPSv2 vision-language model."""

    model_type = "tipsv2"

    def __init__(
        self,
        model_variant="base",
        hidden_size=768,
        patch_size=14,
        image_size=448,
        ffn_layer="mlp",
        init_values=1.0,
        num_register_tokens=1,
        **kwargs,
    ):
        super().__init__(**kwargs)
        self.model_variant = model_variant
        self.hidden_size = hidden_size
        self.patch_size = patch_size
        self.image_size = image_size
        self.ffn_layer = ffn_layer
        self.init_values = init_values
        self.num_register_tokens = num_register_tokens