YOLOv3-complete-pruning icon indicating copy to clipboard operation
YOLOv3-complete-pruning copied to clipboard

提供对YOLOv3及Tiny的多种剪枝版本,以适应不同的需求。

Results 97 YOLOv3-complete-pruning issues
Sort by recently updated
recently updated
newest added

如题,数据生成程序converter.py在哪?

Hello. When I am doing quantizing with the dataset "oxford hand", _python3 train.py --data data/oxfordhand.data --batch-size 32 --accumulate 1 --weights weights/yolov3.weights --cfg cfg/yolov3-quantize-hand.cfg_ the prediction of the model is always...

Number of channels has been reduced from 10592 to 3178 Prune ratio: 0.700 mAP of the pruned model is 0.0002 layer index: 0 total channel: 32 remaining channel: 11 layer...

代码:238.行,我加了 if mtype in ['convolutional', 'upsample', 'maxpool']: print(self.module_list[i]) print(x.shape) x = module(x) error:最相关的最后几行 Sequential( (Conv2d): Conv2d(343, 99, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False) (BatchNorm2d): BatchNorm2d(99, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)...

现在想对v3tiny做量化,但不知道是否支持,我将默认的16bit改为了8bit目前来看使用目标代码可以运行,但量化生成的模型为.weight,模型大小也未发生变化,是不是哪里我操作失误?希望可以为我答疑

When I'm running python3 normal_prune.py I get this error TypeError: 'MaxPool2d' object is not subscriptable can someone help? please

Traceback (most recent call last): File "D:/python projects/YOLOv3-complete-pruning/regular_prune.py", line 233, in pruned_model = prune_model_keep_size(model, prune_idx, CBL_idx, CBLidx2mask) File "D:\python projects\YOLOv3-complete-pruning\utils\prune_utils.py", line 221, in prune_model_keep_size next_conv = pruned_model.module_list[next_idx][0] TypeError: 'MaxPool2d' object...

@coldlarry, 您好!感谢您的分享!我运行了您的量化代码(python3 train.py --data data/oxfordhand.data --batch-size 8 --accumulate 1 --weights weights/yolov3.weights --cfg cfg/yolov3-quantize-hand.cfg --epochs 200),已经训练结束,请问训练之后的量化文件保存在哪里?是什么格式的文件?以及如何利用这些量化文件对测试数据集进行测试? 期望得到您的回复,非常感谢! 祝好! blueeagle ![2020-03-04 00-14-14屏幕截图](https://user-images.githubusercontent.com/24905967/75795405-3baadc00-5dad-11ea-9546-72d4f94aee14.png)

大牛,我想问一个问题,我觉的论文《Learning Efficient Convolutional Networks Through Network Slimming》中给出的损失函数是针对需要剪枝的BN层的,而网络的最后层的损失函数还是经典的yolov3的损失函数,可以这样理解吗?根据代码的意思,最后的loss依然是经典的yolov3的损失函数值,没有加入L1正则的损失值 期待您的回复。十分感谢