DJames23的博客

参考代码：https://github.com/eriklindernoren/PyTorch-YOLOv31.训练时loss出现nan原因：默认学习率learning_rate=0.001过大导致梯度爆炸，出现inf与nan值解决办法：在yolov3.cfg中降低学习率：参考博客：网络训练时出现loss为nan的情况(已解决)神经网络训练时，出现NaN loss...

报错：InvalidArchiveError("Error with archive /anaconda3/pkgs/pytorch-1.2.0-py3.7_cuda10.0.130_cudnn7.6.2_0.tar.bz2. You probably need to delete and re-download or re-create this file. Message from libarchive was:\n\nFailed to open '/anaconda3/pkgs/pytorch

yolov3中的bbox_wh_iou代码如下：def bbox_wh_iou(wh1, wh2): wh2 = wh2.t() # w1是当前anchor的w，h1是当前anchor的h w1, h1 = wh1[0], wh1[1] # w2是所有目标的w,h2是所有目标的h w2, h2 = wh2[0], wh2[1] inter_area = torch.min(w1, w2) * torch.min(h1, h2) union_area

参考博客：PyTorch中[…, 0]的使用pytorch中[…, 0]的用法>>> a= torch.rand(1,2,2,2,2)>>> atensor([[[[[0.2528, 0.9359], [0.1876, 0.4189]], [[0.5546, 0.1931], [0.1389, 0.7482]]], [[[0.6169, 0.2977],

DLA网络详解CenterNet的骨干网络之DLASeg

RuntimeError: CUDA error: device-side assert triggered/pytorch/aten/src/ATen/native/cuda/IndexKernel.cu:60: lambda [](int)->auto::operator()(int)->auto: block: [0,0,0], thread: [2,0,0] Assertion `index >= -sizes[i] && index < sizes[i]

conda create -n 虚拟环境名称（这一步可以指定python版本）conda activate虚拟环境名称conda install pytorch1.2.0 torchvision0.4.0 cudatoolkit=10.0 -c pytorch（这一步指定pytorch版本为1.2.0，torchvision0.4.0，CUDA10.0）以FairMOT为例：conda create -n FairMOT python=3.7conda activate FairMOTconda

在PyTorch中Tensor的查找和筛选例子>>> x = torch.arange(5) >>> xtensor([0, 1, 2, 3, 4])>>> torch.gt(x,1) # 大于tensor([0, 0, 1, 1, 1], dtype=torch.uint8)>>> x>1 # 大于...