打扰一下，训练代码报错

[haotlong]

错误详情：
Traceback (most recent call last):
File "train.py", line 444, in
device=device, )
File "train.py", line 320, in train
loss, loss_xy, loss_wh, loss_obj, loss_cls, loss_l2 = criterion(bboxes_pred, bboxes)
File "E:\anaconda\envs\python3.6\lib\site-packages\torch\nn\modules\module.py", line 550, in call
result = self.forward(*input, **kwargs)
File "train.py", line 204, in forward
pred[..., 0] += self.grid_x[output_id]
RuntimeError: The size of tensor a (8) must match the size of tensor b (76) at non-singleton dimension 3
python3.6版本 pytorch1.50 cpu版

train.txt文件样例：
00000016.jpg 260,252,424,379,3 594,236,663,316,3 602,181,650,229,3
训练语句：
python train.py -l 0.001 -pretrained ./yolov4.conv.137.pth -classes 11 -dir J:\data\mio\MIO-TCD-Localization1\MIO-TCD-Localization\train

未修改其他部分，还望指出错误在哪，打扰作者了，谢谢

[hnqin-xdu]

I have met similar problem, did you solve it?

[wallcuber]

train.py line 93
The image_size is locked. If you want to train your own dataset, you should modify this line.
You need to add a "img_size" parameter, and write
"self.img_size = img_size
#image_size = 608"

also in line 110" fsize = image_size // self.strides[i]" should be modified.
line 300 "criterion = Yolo_loss(device=device, batch=config.batch // config.subdivisions,n_classes=config.classes)" should also add an parameter"img_size".

[bochuanwu]

train.py line 93
The image_size is locked. If you want to train your own dataset, you should modify this line.
You need to add a "img_size" parameter, and write
"self.img_size = img_size
#image_size = 608"

also in line 110" fsize = image_size // self.strides[i]" should be modified.
line 300 "criterion = Yolo_loss(device=device, batch=config.batch // config.subdivisions,n_classes=config.classes)" should also add an parameter"img_size".

great answer!

[Rich2333]

What if the image size is 640*480?