"""Sybil model configuration"""

from transformers import PretrainedConfig
from typing import Optional, List, Dict
import json


class SybilConfig(PretrainedConfig):
    """
    This is the configuration class to store the configuration of a [`SybilForRiskPrediction`].
    It is used to instantiate a Sybil model according to the specified arguments, defining the model architecture.

    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs.

    Args:
        hidden_dim (`int`, *optional*, defaults to 512):
            Dimensionality of the hidden representations.
        dropout (`float`, *optional*, defaults to 0.0):
            The dropout probability for all fully connected layers.
        max_followup (`int`, *optional*, defaults to 6):
            Maximum number of years for risk prediction.
        num_images (`int`, *optional*, defaults to 208):
            Number of CT scan slices to process.
        img_size (`List[int]`, *optional*, defaults to `[512, 512]`):
            Size of input images after preprocessing.
        voxel_spacing (`List[float]`, *optional*, defaults to `[0.703125, 0.703125, 2.5]`):
            Target voxel spacing for CT scans (row, column, slice thickness).
        censoring_distribution (`str`, *optional*, defaults to "weibull"):
            Distribution used for censoring in survival analysis.
        ensemble_size (`int`, *optional*, defaults to 5):
            Number of models in the ensemble.
        calibrator_data (`Dict`, *optional*):
            Calibration data for risk score adjustment.
    """

    model_type = "sybil"

    def __init__(
        self,
        hidden_dim: int = 512,
        dropout: float = 0.0,
        max_followup: int = 6,
        num_images: int = 208,
        img_size: List[int] = None,
        voxel_spacing: List[float] = None,
        censoring_distribution: str = "weibull",
        ensemble_size: int = 5,
        calibrator_data: Optional[Dict] = None,
        initializer_range: float = 0.02,
        **kwargs
    ):
        super().__init__(**kwargs)

        self.hidden_dim = hidden_dim
        self.dropout = dropout
        self.max_followup = max_followup
        self.num_images = num_images
        self.img_size = img_size if img_size is not None else [512, 512]
        self.voxel_spacing = voxel_spacing if voxel_spacing is not None else [0.703125, 0.703125, 2.5]
        self.censoring_distribution = censoring_distribution
        self.ensemble_size = ensemble_size
        self.calibrator_data = calibrator_data
        self.initializer_range = initializer_range