Yolov3训练时报错RuntimeError: CUDA error: device-side assert triggered

报错

Traceback (most recent call last):
  File "train.py", line 122, in <module>
    loss, outputs = model(imgs, targets)
  File "/home/pinery/anaconda3/envs/tf3/lib/python3.7/site-packages/torch/nn/modules/module.py", line 532, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/pinery/code/YOLO/PyTorch-YOLOv3/models.py", line 263, in forward
    x, layer_loss = module[0](x, targets, img_dim)
  File "/home/pinery/anaconda3/envs/tf3/lib/python3.7/site-packages/torch/nn/modules/module.py", line 532, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/pinery/code/YOLO/PyTorch-YOLOv3/models.py", line 188, in forward
    ignore_thres=self.ignore_thres,
  File "/home/pinery/code/YOLO/PyTorch-YOLOv3/utils/utils.py", line 319, in build_targets
    noobj_mask[b[i], anchor_ious > ignore_thres, gj[i], gi[i]] = 0
RuntimeError: CUDA error: device-side assert triggered

解决方法

原因：真实边界框越界，需要调整一下。
1.打开utils/utils.py
2.增添代码如下，放置在上图位置中

gi[gi < 0] = 0
gj[gj < 0] = 0
gi[gi > nG - 1] = nG - 1
gj[gj > nG - 1] = nG - 1

参考资料：https://github.com/eriklindernoren/PyTorch-YOLOv3/issues/157

本文地址：https://blog.csdn.net/qq_35407318/article/details/110938466