import time
from pathlib import Path
import torch
import torch.nn as nn
from torch.utils.data import DataLoader
from unlearning.Strategy import Strategy
from architectures.Model import Model, Architecture

class Retrain(Strategy):
    """
    Implements the Exact Unlearning Baseline by re-instantiating a fresh, 
    pre-trained instance of the specific architecture and training it from scratch 
    on the retain set using the Model's internal train function.
    """
    def __init__(self, target_class_index: int, arch: Architecture, size: int, 
                 lr: float = 0.001, epochs: int = 5):
        super().__init__(target_class_index)
        self.arch = arch
        self.size = size
        self.lr = lr
        self.epochs = epochs

    def _run(self, model: nn.Module, forget_loader: DataLoader, retain_loader: DataLoader) -> nn.Module:
        # 1. Determine the active execution device from the running sandbox
        device = next(model.parameters()).device
        
        print(f">> Triggering Exact Unlearning Baseline (Retraining {self.arch.name} from pristine state)...")
        
        # a new model with default params is created
        fresh_meat = Model.create(self.arch, device, self.size)
        
        # we train it with retain set
        fresh_meat.train(
            epochs=self.epochs, 
            loader=retain_loader, 
            rate=self.lr, 
            mode="retrain"
        )
        
        # 4. Extract the trained nn.Module parameter state dict
        # In-place copy onto the existing sandbox model structure to seamlessly retain downstream evaluations
        model.load_state_dict(fresh_meat.model.state_dict())
        
        print(">> Retraining pipeline finished. Pristine baseline weights successfully established.")
        return model

    def _split_data(self, dataset):
        from sets.Data import get_unlearning_loaders
        return get_unlearning_loaders(
            dataset=dataset,
            forget_class_idx=self.target_class_index,
            batch_size=32
        )