YOLOv5报错：RuntimeError: adaptive_avg_pool2d_backward_cuda does not have a deterministic implementatio

YOLOv5添加注意力机制后报错：RuntimeError: adaptive_avg_pool2d_backward_cuda does not have a deterministic implementation, but you set ‘torch.use_deterministic_algorithms(True)’. You can turn off determinism just for this operation, or you can use the ‘warn_only=True’ option, if that’s acceptable for your application. You can also file an issue at https://github.com/pytorch/pytorch/issues to help us prioritize adding deterministic support for this operation.

解决办法

一、关闭确定性算法模式

在train.py321行（ scaler.scale(loss).backward()，代码前）添加下面一行代码：

torch.use_deterministic_algorithms(False)

若找不到可在train.py中使用Ctrl+F打开查找，输入scaler.scale(loss).backward()查找即可！

二、忽视错误（不关闭确定性算法模式）

找到use_deterministic_algorithms(mode, *, warn_only=False):函数，将warn_only=False改为：warn_only=True。

寻找方法：

可在train.py中随机位置输入torch.use_deterministic_algorithms(False)，选中use_deterministic_algorithms，Ctrl+鼠标左键跳转。

注意：跳转修改后将torch.use_deterministic_algorithms(False)删除。

两者有何区别，请自行实验。有问题可私聊或评论，看到后会尽快回复；若有错误，望指正！