YinggangLi

in my situation, using torch.nn.conv2d instead of F.conv2d . It seems that when using nn.conv2d, you need to initialize in_channels and out_channels，the onnxparser knows how to deal with it. But...

> in my situation, using torch.nn.conv2d instead of F.conv2d . It seems that when using nn.conv2d, you need to initialize in_channels and out_channels，the onnxparser knows how to deal with it....

i solved the problem when i run in Linux with python3.7 & TensorFlow1.14

我的perlayer_bn_weights基本没有变化 从epoch的bn分布来看确实如此 跟你差不多 你用的稀疏训练的版本是哪一个？ 交流一下

yolov5m.cfg该怎么生成？ 我看你这里有s模型的.cfg，不同版本的.cfg又是怎么生成的？

Try to replace softnms with nms. ---Original--- From: ***@***.***> Date: Tue, Oct 31, 2023 18:45 PM To: ***@***.***>; Cc: ***@***.******@***.***>; Subject: Re: [open-mmlab/mmdeploy] [Bug] (Issue #2487) Hello , any updates...

NMS works for me and it is a common operator. Softnms is not implemented in the mmdeploy plugin. ---Original--- From: ***@***.***> Date: Fri, Nov 3, 2023 22:56 PM To: ***@***.***>;...

> Indeed NMS is a common operator. But for CO-DETR, I dont know why It does not work If I modify sof_nms with nms. When I made this modification, the...

I try to download the Adobe-5k-dataset. But it i s too big for me and costs a long time.

same error