YOLOv3v4-ModelCompression-MultidatasetTraining-Multibackbone

YOLOv3v4-ModelCompression-MultidatasetTraining-Multibackbone copied to clipboard

Timothy-105

Does it support scaledyolov4? How to use this for scaledyolov4?

bobbilichandu

test.py RuntimeError: CUDA error: device-side assert triggered

2

yangzhaojason

google drive

2

thanks you for your work but i couldn't download from baidu ,could you please tell me if there are any version from Doir or visdrone on google drive thanks again

omar-Fouad

YOLOv4

1

你好，请问如果我要先用yolov4训练自己的数据集，然后再使用您的repo来进行模型剪枝，那么这整体的步骤是什么呢，后续能否出个yolov4的教程什么的。是不是先用AlexAB的darknet训练自己的yolov4模型，然后将自己数据集的cfg和训练好的weights权重在您的repo去做稀疏化训练和剪枝，谢谢了。

CaiyuZhang

RuntimeError: Sizes of tensors must match except in dimension 2.

4

Hello, I have successfully finished the following steps: 1. Normal training using 1000 images from coco-2014 dataset with command: _python3 train.py --data data/coco2014.data --batch-size 16 -pt --weights yolov4.weights --cfg cfg/yolov4/yolov4.cfg_...

indersingh17188

Just to notify, the utils script need to be import in the models script. Best.

RRL-dev

UserWarning: Single-Process Multi-GPU is not the recommended mode for DDP. In this mode, each DDP instance operates on multiple devices and creates multiple module replicas within one process.

1

你好，我多卡训练的时候会报出这个提示，检查GPU发现，其中一个GPU的显存占用明显更多,这应该不太正常吧，请问知道怎么解决吗 

ly0303521

你好！ 请问一下，以compute_lost_KD3为例，以下代码为什么把objness也计算进softmax里面？ output_s_i = ps[..., 4:].view(-1, model.nc + 1) output_t_i = pt[..., 4:].view(-1, model.nc + 1) lcls += criterion_st(nn.functional.log_softmax(output_s_i / T, dim=1), nn.functional.softmax(output_t_i / T, dim=1)) * (T * T)...

zhangyilalala

while trying to prune after sparse training: sudo python3 regular_prune.py --cfg cfg/yolo-fastest-xl-test.cfg --data data/obj-boar.data --weights weights/last.pt --percent 0.95 --img-size 1280 Namespace(cfg='cfg/yolo-fastest-xl-test.cfg', data='data/obj-boar.data', img_size=1280, percent=0.95, weights='weights/last.pt') Warning: Unrecognized Layer Type: dropout...

mate-hegedus