RepViT

RepViT copied to clipboard

Hi, Thank you for your great work! I have some questions about exporting the repvit-sam decoder onnx model: 1. How to set point_labels? 2. How to distinguish between point and...

zongzhesun

22wei22

Hi, I'm currently experimenting with the RepViT segmentation model. In the repository, the latest checkpoint weights provided are repvit_m2_3. Could you update me with the configuration and weight, as well...

ddung2c

我使用该仓库中repvit_m0_9的预训练参数和timm中repvit_m0_9预训练参数，在我的任务上进行微调时，结果是不同的，请问这是为什么呢？

xryami

thank you for your great work! how can i export the segmentation model as onnx? can you provide the code

hilaryliang

I see that the repository currently supports only single‑node, multi‑GPU training. Even with a single node equipped with six L40 GPUs, the process still takes at least two days. Is...

limengzhaolihai

The implementation for Residual() is incorrect in repvit.py. if isinstance(self.m, Conv2d_BN): m = self.m.fuse() assert(m.groups == m.in_channels) identity = torch.ones(m.weight.shape[0], m.weight.shape[1], 1, 1) identity = torch.nn.functional.pad(identity, [1,1,1,1]) m.weight += identity.to(m.weight.device)...

kensun619

下游任务，比如检测时，怎么使用replace_batchnorm函数？我在代码里没有看到。。

puyiwen

Does the repvit classification experiment only have training sets and test sets, no val sets?

2

weldingCode

使用xcode测试速度不符，进行了多次测试，取最后一次稳定的结果如下图所示： 这里速度明显过慢，我发现执行多次推理的过程中第一次推理的时间过长，甚至达到10ms，把这个时间用来计算均值显然不合理，所以在xcode中如何避免这一问题，文中又是怎么做的呢

chairman-lu