yolov8模型稀疏化剪枝

1. 模型稀疏化训练（增加权重稀疏化的代码进行训练）
--ultralytics/engine/trainer.py
  约在345行:
    # Backward
    self.scaler.scale(self.loss).backward()

    # ========== 新增 ==========
    l1_lambda = 1e-2 * (1 - 0.9 * epoch / self.epochs)
    for k, m in self.model.named_modules():
        if isinstance(m, nn.BatchNorm2d):
            m.weight.grad.data.add_(l1_lambda * torch.sign(m.weight.data))
            m.bias.grad.data.add_(1e-2 * torch.sign(m.bias.data))
    # ========== 新增 ==========

    # Optimize - https://pytorch.org/docs/master/notes/amp_examples.html
2. 开始训练
--ultralytics-main/train.py
    训练代码如下：
        from ultralytics import YOLO
        import torch

        model = YOLO('yolov8s.yaml'