WindVChen comments

Results 160 comments of


                                            WindVChen

ImportError att_grid_generator_cuda error

@vb123er951 @guofeng12sigma @haidog-yaqub @Aminkhormali Hi guys, I think I have figured out how to run the setup.py. You maybe interested to read that here #24

Questions for classifier-free guidance on ImageNet

@GiannisPikoulis I held the same assumption previously. However, one question is that, if the model is not trained with unconditional mode, i.e., empty string, can it really handle the empty...

Questions for classifier-free guidance on ImageNet

Hi everyone, I've found this figure in the [Stable Diffusion Repository](https://github.com/CompVis/stable-diffusion): ![image](https://user-images.githubusercontent.com/56162412/203264024-34445a2a-1dca-4f7c-98f6-e986de92bb70.png) where it seems that only v1.3 and v1.4 used dropping in the training, but all the versions used...

I cannot run DRENet. Because has a bug: RuntimeError: The size of tensor a (160) must match the size of tensor b (256) at non-singleton dimension 1

Hi @vtise-github, It appears that you're encountering a similar issue to #4 regarding input resolution. You can find potential solutions by referencing that issue. If you're using detect.py and our...

在yolov5-5.0中只使用CRMA模块

哈喽 @Potato640 ，参数设置看起来没啥问题。从报错信息看，训练能够正常训练，但是测试过程好像是因为输入尺寸原因崩了，可能需要设断点检查下下面这一块代码传入的img的尺寸，以及augment参数是否为False. ![image](https://user-images.githubusercontent.com/56162412/226658439-a5f7991c-5e73-4029-a853-08e009d6cfdc.png)

在yolov5-5.0中只使用CRMA模块

👍👍👍

Cannot run detect.py

Hi @tommylatranvu, I think this may help https://github.com/WindVChen/DRENet/issues/4#issuecomment-1255846664.

performance decrease during training

There seems a gradient explosion (or something else) that lead to a NAN loss value. What about turning down the learning rate, or clip the gradient before _optimizer.step()_ ?

运行detect.py出错，您好我把C3ResAtnMHSA加入了yolov7中，出现下面的错误

哈喽呀 @denghuimin1 从上面的报错信息来看，大概率是输入尺寸的问题。需要注意的是，如果图像尺寸!=512，需要对配置文件进行修改。具体可以参考下issue #4 和 #9 .

实验

你好呀 @liuz25 ，抱歉回复晚了，我们的代码里没有包含对比方法的代码/配置文件，但你可以利用下面的链接对结果进行复现： YOLOv3: [所用代码库](https://github.com/ultralytics/yolov3), 因为作者跟YOLOv5是同一批，所以配置跟本代码库基本一致 Retinanet: [所用代码库](https://github.com/yhenon/pytorch-retinanet), 配置为ResNet50 SSD： [所用的源码库](https://github.com/lufficc/SSD), [所用的配置文件](https://github.com/lufficc/SSD/blob/master/configs/vgg_ssd512_coco_trainval35k.yaml) FasterRCNN: [所用代码库](https://github.com/jwyang/faster-rcnn.pytorch), backbone为VGG16, 配置方式请见其代码库README.md EfficientDet: [所用代码库](https://github.com/zylo117/Yet-Another-EfficientDet-Pytorch), 采用D0和D2, 配置方式请见其代码库README.md FCOS: [所用代码库](https://github.com/tianzhi0549/FCOS), backbone为ResNet50, 配置方式请见其代码库README.md CenterNet: [所用代码库](https://github.com/xingyizhou/CenterNet), backbone为Hourglass104, 配置方式请见其代码库README.md HSFNet,...