Wangbenzhi

I have the same question. Do you have resolved it?

Maybe it is a bug in detection? The backbone(PVT) initialize in Detector.weight_init()

So sad

你好，请问你这里不会报参数数量不正确的错误吗？

> > 你好，请问你这里不会报参数数量不正确的错误吗？ > > 你好，我也遇到了相同的报错，请问你是怎么解决的呢？ https://github.com/CaptainEven/MOTEvaluate

> For 1.3: > > * error: ‘torch::ExpandingArray torch::nn::ConvOptions::stride_’ is private within this context > torch::nn::Conv2dOptions conv_options( > > * conv_options.stride_ = stride; > * conv_options.padding_ = padding; > *...

I solve this issue by setting drop_last=False in the dataloder.