resume_net未定义以及测试加载的模型

[jamesbondzhou]

您好，我有两个问题想请教一下。
第一个是train.py第114行，torch.load(resume_net)这里的resume_net没定义吧
第二个问题是test.py运行时报错RuntimeError: Error(s) in loading state_dict for FSSD:size mismatch for conf.0.bias: copying a param with shape torch.Size([126]) from checkpoint, the shape in current model is torch.Size([486])........
这里net.load_state_dict(new_state_dict)加载fssd_voc_79_74.pth导致了size不匹配

[dlyldxwl]

1. resume_net is defined in line 36;
   parser.add_argument('--resume_net', default=None, help='resume net for retraining')
2. I believe you utilize the pruned model while not modify the version on the test.py when testing.
   parser.add_argument('-v', '--version', default='FSSD_VGG', help='version.')

[jamesbondzhou]

非常感谢回答!
关于第二个问题。是要把parser.add_argument('-v', '--version', default='FSSD_VGG', help='version.')default这里的FSSD_VGG换成FSSD_VGG_prune么，我试过了，错的更加离谱啦。
这是FSSD_VGG下完整的报错
RuntimeError: Error(s) in loading state_dict for FSSD:
size mismatch for conf.0.bias: copying a param with shape torch.Size([126]) from checkpoint, the shape in current model is torch.Size([486]).
size mismatch for conf.0.weight: copying a param with shape torch.Size([126, 512, 3, 3]) from checkpoint, the shape in current model is torch.Size([486, 512, 3, 3]).
size mismatch for conf.1.bias: copying a param with shape torch.Size([126]) from checkpoint, the shape in current model is torch.Size([486]).
size mismatch for conf.1.weight: copying a param with shape torch.Size([126, 512, 3, 3]) from checkpoint, the shape in current model is torch.Size([486, 512, 3, 3]).
size mismatch for conf.2.bias: copying a param with shape torch.Size([126]) from checkpoint, the shape in current model is torch.Size([486]).
size mismatch for conf.2.weight: copying a param with shape torch.Size([126, 256, 3, 3]) from checkpoint, the shape in current model is torch.Size([486, 256, 3, 3]).
size mismatch for conf.3.bias: copying a param with shape torch.Size([126]) from checkpoint, the shape in current model is torch.Size([486]).
size mismatch for conf.3.weight: copying a param with shape torch.Size([126, 256, 3, 3]) from checkpoint, the shape in current model is torch.Size([486, 256, 3, 3]).
size mismatch for conf.4.bias: copying a param with shape torch.Size([84]) from checkpoint, the shape in current model is torch.Size([324]).
size mismatch for conf.4.weight: copying a param with shape torch.Size([84, 256, 3, 3]) from checkpoint, the shape in current model is torch.Size([324, 256, 3, 3]).
size mismatch for conf.5.bias: copying a param with shape torch.Size([84]) from checkpoint, the shape in current model is torch.Size([324]).
size mismatch for conf.5.weight: copying a param with shape torch.Size([84, 256, 3, 3]) from checkpoint, the shape in current model is torch.Size([324, 256, 3, 3]).

[dlyldxwl]

126 = 21×6
486= 81×6
I guess you use the class number of coco(81) rather than VOC(21)

[jamesbondzhou]

嗯嗯，极有可能！感谢！！

[dlyldxwl]

OK, if you think this repo is useful for you, you can give it a "star", -.-
and then, I will release a project about stronger one-stage detector based on ssd /yolo v3, which integrate much tricks, for example, mix up, GIOU and so on. If you are interested in it, please give more attention.

[jamesbondzhou]

持续关注大佬中！！！

[ZCCDL]

我想问一下你是怎么加的BN,为什么我加的时候出错，在加入到VGG主干网络中怎么和原预训练模型匹配呢？谢谢，能交流一下吗？qq2497431867