剪枝后mAP直接降为0

[sbbug]

你好，楼主，为何剪枝完map直接为0？

[sbbug]

[ZJU-lishuang]

模型调教的原因 可以参考示例复现

[Moon404]

你好，楼主，为何剪枝完map直接为0？

你好，请问你是使用楼主的剪枝代码及命令完成减枝的吗？

[sbbug]

你好，楼主，为何剪枝完map直接为0？

你好，请问你是使用楼主的剪枝代码及命令完成减枝的吗？

是的，基本是楼主的代码

[Moon404]

主的代码

我运行楼主的剪枝就会报错，你能帮我看看吗？vx：txy1292757414

[sbbug]

主的代码

我运行楼主的剪枝就会报错，你能帮我看看吗？vx：txy1292757414

你可以把错误贴出来

[Moon404]

主的代码

我运行楼主的剪枝就会报错，你能帮我看看吗？vx：txy1292757414

你可以把错误贴出来

Traceback (most recent call last): File "slim_prune_yolov5s.py", line 355, in origin_model_metric = eval_model(model) File "slim_prune_yolov5s.py", line 350, in eval_model = lambda model:test(model=model,cfg=opt.cfg, data=opt.data, batch_size=16, img_size=img_size) File "D:\Gitee\yolov5_prune-4\test.py", line 181, in test inf_out, train_out = model(imgs) # inference and training outputs File "C:\Users\TXY\anaconda3\envs\yolo_v5\lib\site-packages\torch\nn\modules\module.py", line 727, in _call_impl result = self.forward(*input, *kwargs) File "D:\Gitee\yolov5_prune-4\modelsori.py", line 293, in forward x = module(x, img_size) File "C:\Users\TXY\anaconda3\envs\yolo_v5\lib\site-packages\torch\nn\modules\module.py", line 727, in _call_impl result = self.forward(input, **kwargs) File "D:\Gitee\yolov5_prune-4\modelsori.py", line 178, in forward p = p.view(bs, self.na, self.nc + 5, self.ny, self.nx).permute(0, 1, 3, 4, 2).contiguous() # prediction RuntimeError: shape '[16, 3, 85, 40, 80]' is invalid for input of size 1075200

[sbbug]

主的代码

我运行楼主的剪枝就会报错，你能帮我看看吗？vx：txy1292757414

你可以把错误贴出来

Traceback (most recent call last): File "slim_prune_yolov5s.py", line 355, in origin_model_metric = eval_model(model) File "slim_prune_yolov5s.py", line 350, in eval_model = lambda model:test(model=model,cfg=opt.cfg, data=opt.data, batch_size=16, img_size=img_size) File "D:\Gitee\yolov5_prune-4\test.py", line 181, in test inf_out, train_out = model(imgs) # inference and training outputs File "C:\Users\TXY\anaconda3\envs\yolo_v5\lib\site-packages\torch\nn\modules\module.py", line 727, in _call_impl result = self.forward(*input, *kwargs) File "D:\Gitee\yolov5_prune-4\modelsori.py", line 293, in forward x = module(x, img_size) File "C:\Users\TXY\anaconda3\envs\yolo_v5\lib\site-packages\torch\nn\modules\module.py", line 727, in _call_impl result = self.forward(input, **kwargs) File "D:\Gitee\yolov5_prune-4\modelsori.py", line 178, in forward p = p.view(bs, self.na, self.nc + 5, self.ny, self.nx).permute(0, 1, 3, 4, 2).contiguous() # prediction RuntimeError: shape '[16, 3, 85, 40, 80]' is invalid for input of size 1075200

模型配置文件需要修改，classes 、filters

[sbbug]

模型调教的原因 可以参考示例复现

楼主，你剪枝代码有点复杂，能不能精简一下，，

[Moon404]

你好，楼主，为何剪枝完map直接为0？

根据你说的调整classes后成功剪枝，出现和你相同的问题，map为0，且进行下一步微调时

python prune_finetune.py --img 640 --batch 4 --epochs 5 --data data/myvoc.yaml --cfg ./cfg/prune_0.8_keep_0.01_yolov5s_v4.cfg --weights ./weights/best
.pt

报错 Traceback (most recent call last): File "prune_finetune.py", line 532, in train(hyp, opt, device, tb_writer, wandb) File "prune_finetune.py", line 84, in train state_dict = ckpt['model'].float().state_dict() # to FP32 AttributeError: 'collections.OrderedDict' object has no attribute 'float' 请问你是否也是这样呢？

[Moon404]

你好，楼主，为何剪枝完map直接为0？

请问你解决剪枝的问题了吗？

[ZJU-lishuang]

如果是开源数据集，请提供下载路径，用来帮助修复bug

[dada1437903138]

主的代码

我运行楼主的剪枝就会报错，你能帮我看看吗？vx：txy1292757414

你可以把错误贴出来

Traceback (most recent call last): File "slim_prune_yolov5s.py", line 355, in origin_model_metric = eval_model(model) File "slim_prune_yolov5s.py", line 350, in eval_model = lambda model:test(model=model,cfg=opt.cfg, data=opt.data, batch_size=16, img_size=img_size) File "D:\Gitee\yolov5_prune-4\test.py", line 181, in test inf_out, train_out = model(imgs) # inference and training outputs File "C:\Users\TXY\anaconda3\envs\yolo_v5\lib\site-packages\torch\nn\modules\module.py", line 727, in _call_impl result = self.forward(*input, *kwargs) File "D:\Gitee\yolov5_prune-4\modelsori.py", line 293, in forward x = module(x, img_size) File "C:\Users\TXY\anaconda3\envs\yolo_v5\lib\site-packages\torch\nn\modules\module.py", line 727, in _call_impl result = self.forward(input, **kwargs) File "D:\Gitee\yolov5_prune-4\modelsori.py", line 178, in forward p = p.view(bs, self.na, self.nc + 5, self.ny, self.nx).permute(0, 1, 3, 4, 2).contiguous() # prediction RuntimeError: shape '[16, 3, 85, 40, 80]' is invalid for input of size 1075200

模型配置文件需要修改，classes 、filters

你好，你们剪枝的时候--data xxxxx.data文件从何而来呢，yolov5网络训练的时候用的yaml文件呀

[haoxue1215]

这个问题解决了吗？我的也是，而且会显示剪枝前的map也为0

[ZJU-lishuang]

更新了剪枝代码验证模型的函数，和原版保持一致