YOLOv5-LITE train训练报错总结

问题：

Traceback (most recent call last):
File “train.py”, line 544, in <module>
train(hyp, opt, device, tb_writer)
File “train.py”, line 305, in train
loss, loss_items = compute_loss(pred, targets.to(device)) # loss scaled by batch_size
File “/home/**/YOLOv5-Lite-master/utils/loss.py”, line 122, in call
b = b.long()
UnboundLocalError: local variable ‘b’ referenced before assignment

（1）小白训练一直出现报错，一开始是utils里面的datasets.py general.py里面的数据类型和版本对应不上，把这两个文件里面的np.int修改成np.int_

例如：bi = np.floor(np.arange(n) / batch_size).astype(np.int)修改成

           bi = np.floor(np.arange(n) / batch_size).astype(np.int_)

（2）然后再次报错出现了loss.py中的数据类型问题，修改成红色部分

# Define

b, c = t[:, :2].long().T # image, class

gxy = t[:, 2:4] # grid xy

gwh = t[:, 4:6] # grid wh

gij = (gxy - offsets).to(torch.int64)

gi, gj = gij.T # grid xy indices

# Append

a = t[:, 6].long() # anchor indices

indices.append((b, a, gj.to(torch.int64).clamp(0, gain[3] - 1), gi.to(torch.int64).clamp(0, gain[2] - 1))) # image, anchor, grid indices

tbox.append(torch.cat((gxy - gij, gwh), 1)) # box

anch.append(anchors[a]) # anchors

tcls.append(c) # class

（3）最后出现了一个报错我无法解决，以我这种小白就在网上各种找资料，最后修改了

Traceback (most recent call last):
File “train.py”, line 544, in <module>
train(hyp, opt, device, tb_writer)
File “train.py”, line 305, in train
loss, loss_items = compute_loss(pred, targets.to(device)) # loss scaled by batch_size
File “/home/**/YOLOv5-Lite-master/utils/loss.py”, line 122, in call
b = b.long()
UnboundLocalError: local variable ‘b’ referenced before assignment

把loss.py中的这一行

gain = torch.ones(7, device=targets.device) 修改成

gain = torch.ones(7, device=targets.device).long()

问题解决是转载https://blog.csdn.net/weixin_41012767/article/details/125779972

虽然我也不太懂为什么是这么修改，但是最后真的跑通了，感谢大佬！