yolov8断点续训教程

最新推荐文章于 2024-05-21 14:24:06 发布

我真不想CV

最新推荐文章于 2024-05-21 14:24:06 发布

阅读量580

点赞数

文章标签： YOLO pytorch 深度学习

本文链接：https://blog.csdn.net/xlxun/article/details/133747915

版权

1.打开搜索功能，搜素check_resume()函数
将resume = self.args.resume注释掉
2.找到resume_training()函数，ckpt =torch.load(上次中断生成的last.pt的路径)
ckpt = torch.load(‘runs/detect/train160/weights/last.pt’)
3.将以下文件赋值导自己的py文件上，执行即可

from ultralytics import YOLO
import torch
# Load a model
model = YOLO('ultralytics/cfg/models/v8/yolov8-C2f-DySnakeConv-GAM.yaml')  # build a new model from YAML   默认是yolov8n/
model = YOLO('runs/detect/train160/weights/last.pt')
model = YOLO('ultralytics/cfg/models/v8/yolov8-C2f-DySnakeConv-GAM.yaml').load('runs/detect/train160/weights/last.pt')  # build from YAML and transfer weights

if __name__ == '__main__':
# Train the model    #your code
    print(torch.cuda.is_available())
    print(torch.cuda.device_count())
    model.train(data='ultralytics/cfg/datasets/custom_data.yaml', epochs=400, imgsz=640, resume=True)