Xinyuan Chang

Thank you for your great works. https://github.com/xmu-xiaoma666/External-Attention-pytorch/blob/4f21eba1ae338dd0be6bd2a57656da55e2504580/attention/ShuffleAttention.py#L59 There is a problem with this line of code

Hi :) nice job, can you provide a pretrained model when args.layer='layer4'? thanks.

你好，为什么使用resnet101的pretrain模型时，报错： `pickle.UnpicklingError: A load persistent id instruction was encountered, but no persistent_load function was specified.`

Hello, nice work, but during the reproduction process, I found a lot of bugs in the code (the loss does not converge in the iccv version script, and there are...

hello~，看到作者都是国人，就用中文提问了。 aaai论文中的结果，是使用了PON的划分吗，没有看到文中提及？另外文中提到了使用的是6 camera，但对比的方法都是1 camera，是否公平？ 另外，iccv论文扩充的期刊中（https://arxiv.org/pdf/2402.08207），ablation study有提及default 配置与new的配置，差异较大，主要是new split下数据是non-overlapping的，然而PON也是non-overlapping的，为何在PON配置下，指标仍然较高？此处default配置是PON配置吗？（看起来更像是NuScene原始划分？）如果不是PON划分，与sota的对比是否公平？ 期待解答，祝好~