search

KaiyangZhou / deep-person-reid

Torchreid: Deep learning person re-identification in PyTorch.
https://kaiyangzhou.github.io/deep-person-reid/
MIT License
4.27k stars 1.14k forks source link

Error when loading DenseNet model : Missing key(s) in state_dict #23

Closed ghost closed 6 years ago

ghost commented 6 years ago

1.Error Description

I tried to perform test on the MARS dataset with the provided trained model densenet121_xent_htri_mars.pth.tar, using this command： python train_vid_model_xent_htri.py -d mars -a densenet121 --evaluate --resume saved-models/densenet121_xent_htri_mars.pth.tar --save-dir log/densenet121-xent-htri-mars --test-batch 2

And I got the following error info:

RuntimeError: Error(s) in loading state_dict for DenseNet121: Missing key(s) in state_dict: "base.denseblock1.denselayer1.norm1.running_var", ...

(for detailed info, please refer to the console log)

Some Analysis

1)Although I had already put the provided model file in the dir deep-person-reid/saved-models/,the console log shows that PyTorch still automatically downloaded a pre-trained model from "https://download.pytorch.org/models/densenet121-a639ec97.pth" to the path /home/user/.torch/models/densenet121-a639ec97.pth. After the download process had completed, pyTorch loaded the provided model densenet121_xent_htri_mars.pth.tar.(see the log for detailed info.)

It seems that the provided model densenet121_xent_htri_mars.pth.tar which contains the model's parameters ONLY is not consistent with the model auto downloaded from download.pytorch.org .

2)Why did PyTorch auto download pre-trained model from download.pytorch.org before it loaded densenet121_xent_htri_mars.pth.tar ? Because in __init__ function of DenseNet.py,"pretrained" is set to true. This Line: densenet121 = torchvision.models.densenet121(pretrained=True)

3)I tried to perform test on the MARS dataset with the ResNet model (resnet50_xent_mars.pth.tar). PyTorch again automatically downloaded another pre-trained model from download.pytorch.org, but this time, NO runtime error occurred. Command: python train_vid_model_xent.py -d mars -a resnet50 --evaluate --resume saved-models/resnet50_xent_mars.pth.tar --save-dir log/resnet50-xent-mars --test-batch 2

I have found a similar issue(#4), but it can't solve this problem. Does anyone know how to solve it? Thanks in advance.

2.Additional Info

Code Version

May 30, 2018

Model File Hash

File Name : densenet121_xent_htri_mars.pth.tar URL : http://www.eecs.qmul.ac.uk/~kz303/deep-person-reid/model-zoo/video-models/densenet121_xent_htri_mars.pth.tar MD5 : 544FFC7520B5719B2F63CAA44F412F49

Environment

Ubuntu 14.04 x64 Anaconda 2 4.4.0 x86_64 Python 2.7.13 PyTorch 0.4.0 torchvision-cpu 0.2.1 use CPU only (no CUDA installed)

ghost commented 6 years ago

Console log

/home/user/anaconda2/lib/python2.7/site-packages/torchvision-0.2.1-py2.7.egg/torchvision/models/densenet.py:212: UserWarning: nn.init.kaiming_normal is now deprecated in favor of nn.init.kaiming_normal_.
Downloading: "https://download.pytorch.org/models/densenet121-a639ec97.pth"
 to /home/user/.torch/models/densenet121-a639ec97.pth

0.0%
0.1%

...

99.9%
100.0%
==========
Args:Namespace(arch='densenet121', dataset='mars', eval_step=-1, evaluate=True, gamma=0.1, gpu_devices='0', height=256, htri_only=False, lr=0.0003, margin=0.3, max_epoch=500, num_instances=4, optim='adam', pool='avg', print_freq=10, resume='saved-models/densenet121_xent_htri_mars.pth.tar', root='data', save_dir='log/densenet121-xent-htri-mars', seed=1, seq_len=15, start_epoch=0, start_eval=0, stepsize=200, test_batch=2, train_batch=32, use_cpu=False, weight_decay=0.0005, width=128, workers=4)
==========
Currently using CPU (GPU is highly recommended)
Initializing dataset mars
=> MARS loaded
Dataset statistics:
  ------------------------------
  subset   | # ids | # tracklets
  ------------------------------
  train    |   625 |     8298
  query    |   626 |     1980
  gallery  |   622 |     9330
  ------------------------------
  total    |  1251 |    19608
  number of images per tracklet: 2 ~ 920, average 59.5
  ------------------------------
Initializing model: densenet121
Model size: 7.59448M
Loading checkpoint from 'saved-models/densenet121_xent_htri_mars.pth.tar'
Traceback (most recent call last):
  File "train_vid_model_xent_htri.py", line 322, in <module>
    main()
  File "train_vid_model_xent_htri.py", line 154, in main
    model.load_state_dict(checkpoint['state_dict'])
  File "/home/user/anaconda2/lib/python2.7/site-packages/torch/nn/modules/module.py", line 721, in load_state_dict
    self.__class__.__name__, "\n\t".join(error_msgs)))
RuntimeError: Error(s) in loading state_dict for DenseNet121:
    Missing key(s) in state_dict: "base.denseblock1.denselayer1.norm1.running_var", "base.denseblock1.denselayer1.norm1.bias", "base.denseblock1.denselayer1.norm1.weight", "base.denseblock1.denselayer1.norm1.running_mean", "base.denseblock1.denselayer1.conv1.weight", "base.denseblock1.denselayer1.norm2.running_var", "base.denseblock1.denselayer1.norm2.bias", "base.denseblock1.denselayer1.norm2.weight", "base.denseblock1.denselayer1.norm2.running_mean", "base.denseblock1.denselayer1.conv2.weight", "base.denseblock1.denselayer2.norm1.running_var", "base.denseblock1.denselayer2.norm1.bias", "base.denseblock1.denselayer2.norm1.weight", "base.denseblock1.denselayer2.norm1.running_mean", "base.denseblock1.denselayer2.conv1.weight", "base.denseblock1.denselayer2.norm2.running_var", "base.denseblock1.denselayer2.norm2.bias", "base.denseblock1.denselayer2.norm2.weight", "base.denseblock1.denselayer2.norm2.running_mean", "base.denseblock1.denselayer2.conv2.weight", "base.denseblock1.denselayer3.norm1.running_var", "base.denseblock1.denselayer3.norm1.bias", "base.denseblock1.denselayer3.norm1.weight", "base.denseblock1.denselayer3.norm1.running_mean", "base.denseblock1.denselayer3.conv1.weight", "base.denseblock1.denselayer3.norm2.running_var", "base.denseblock1.denselayer3.norm2.bias", "base.denseblock1.denselayer3.norm2.weight", "base.denseblock1.denselayer3.norm2.running_mean", "base.denseblock1.denselayer3.conv2.weight", "base.denseblock1.denselayer4.norm1.running_var", "base.denseblock1.denselayer4.norm1.bias", "base.denseblock1.denselayer4.norm1.weight", "base.denseblock1.denselayer4.norm1.running_mean", "base.denseblock1.denselayer4.conv1.weight", "base.denseblock1.denselayer4.norm2.running_var", "base.denseblock1.denselayer4.norm2.bias", "base.denseblock1.denselayer4.norm2.weight", "base.denseblock1.denselayer4.norm2.running_mean", "base.denseblock1.denselayer4.conv2.weight", "base.denseblock1.denselayer5.norm1.running_var", "base.denseblock1.denselayer5.norm1.bias", "base.denseblock1.denselayer5.norm1.weight", "base.denseblock1.denselayer5.norm1.running_mean", "base.denseblock1.denselayer5.conv1.weight", "base.denseblock1.denselayer5.norm2.running_var", "base.denseblock1.denselayer5.norm2.bias", "base.denseblock1.denselayer5.norm2.weight", "base.denseblock1.denselayer5.norm2.running_mean", "base.denseblock1.denselayer5.conv2.weight", "base.denseblock1.denselayer6.norm1.running_var", "base.denseblock1.denselayer6.norm1.bias", "base.denseblock1.denselayer6.norm1.weight", "base.denseblock1.denselayer6.norm1.running_mean", "base.denseblock1.denselayer6.conv1.weight", "base.denseblock1.denselayer6.norm2.running_var", "base.denseblock1.denselayer6.norm2.bias", "base.denseblock1.denselayer6.norm2.weight", "base.denseblock1.denselayer6.norm2.running_mean", "base.denseblock1.denselayer6.conv2.weight", "base.denseblock2.denselayer1.norm1.running_var", "base.denseblock2.denselayer1.norm1.bias", "base.denseblock2.denselayer1.norm1.weight", "base.denseblock2.denselayer1.norm1.running_mean", "base.denseblock2.denselayer1.conv1.weight", "base.denseblock2.denselayer1.norm2.running_var", "base.denseblock2.denselayer1.norm2.bias", "base.denseblock2.denselayer1.norm2.weight", "base.denseblock2.denselayer1.norm2.running_mean", "base.denseblock2.denselayer1.conv2.weight", "base.denseblock2.denselayer2.norm1.running_var", "base.denseblock2.denselayer2.norm1.bias", "base.denseblock2.denselayer2.norm1.weight", "base.denseblock2.denselayer2.norm1.running_mean", "base.denseblock2.denselayer2.conv1.weight", "base.denseblock2.denselayer2.norm2.running_var", "base.denseblock2.denselayer2.norm2.bias", "base.denseblock2.denselayer2.norm2.weight", "base.denseblock2.denselayer2.norm2.running_mean", "base.denseblock2.denselayer2.conv2.weight", "base.denseblock2.denselayer3.norm1.running_var", "base.denseblock2.denselayer3.norm1.bias", "base.denseblock2.denselayer3.norm1.weight", "base.denseblock2.denselayer3.norm1.running_mean", "base.denseblock2.denselayer3.conv1.weight", "base.denseblock2.denselayer3.norm2.running_var", "base.denseblock2.denselayer3.norm2.bias", "base.denseblock2.denselayer3.norm2.weight", "base.denseblock2.denselayer3.norm2.running_mean", "base.denseblock2.denselayer3.conv2.weight", "base.denseblock2.denselayer4.norm1.running_var", "base.denseblock2.denselayer4.norm1.bias", "base.denseblock2.denselayer4.norm1.weight", "base.denseblock2.denselayer4.norm1.running_mean", "base.denseblock2.denselayer4.conv1.weight", "base.denseblock2.denselayer4.norm2.running_var", "base.denseblock2.denselayer4.norm2.bias", "base.denseblock2.denselayer4.norm2.weight", "base.denseblock2.denselayer4.norm2.running_mean", "base.denseblock2.denselayer4.conv2.weight", "base.denseblock2.denselayer5.norm1.running_var", "base.denseblock2.denselayer5.norm1.bias", "base.denseblock2.denselayer5.norm1.weight", "base.denseblock2.denselayer5.norm1.running_mean", "base.denseblock2.denselayer5.conv1.weight", "base.denseblock2.denselayer5.norm2.running_var", "base.denseblock2.denselayer5.norm2.bias", "base.denseblock2.denselayer5.norm2.weight", "base.denseblock2.denselayer5.norm2.running_mean", "base.denseblock2.denselayer5.conv2.weight", "base.denseblock2.denselayer6.norm1.running_var", "base.denseblock2.denselayer6.norm1.bias", "base.denseblock2.denselayer6.norm1.weight", "base.denseblock2.denselayer6.norm1.running_mean", "base.denseblock2.denselayer6.conv1.weight", "base.denseblock2.denselayer6.norm2.running_var", "base.denseblock2.denselayer6.norm2.bias", "base.denseblock2.denselayer6.norm2.weight", "base.denseblock2.denselayer6.norm2.running_mean", "base.denseblock2.denselayer6.conv2.weight", "base.denseblock2.denselayer7.norm1.running_var", "base.denseblock2.denselayer7.norm1.bias", "base.denseblock2.denselayer7.norm1.weight", "base.denseblock2.denselayer7.norm1.running_mean", "base.denseblock2.denselayer7.conv1.weight", "base.denseblock2.denselayer7.norm2.running_var", "base.denseblock2.denselayer7.norm2.bias", "base.denseblock2.denselayer7.norm2.weight", "base.denseblock2.denselayer7.norm2.running_mean", "base.denseblock2.denselayer7.conv2.weight", "base.denseblock2.denselayer8.norm1.running_var", "base.denseblock2.denselayer8.norm1.bias", "base.denseblock2.denselayer8.norm1.weight", "base.denseblock2.denselayer8.norm1.running_mean", "base.denseblock2.denselayer8.conv1.weight", "base.denseblock2.denselayer8.norm2.running_var", "base.denseblock2.denselayer8.norm2.bias", "base.denseblock2.denselayer8.norm2.weight", "base.denseblock2.denselayer8.norm2.running_mean", "base.denseblock2.denselayer8.conv2.weight", "base.denseblock2.denselayer9.norm1.running_var", "base.denseblock2.denselayer9.norm1.bias", "base.denseblock2.denselayer9.norm1.weight", "base.denseblock2.denselayer9.norm1.running_mean", "base.denseblock2.denselayer9.conv1.weight", "base.denseblock2.denselayer9.norm2.running_var", "base.denseblock2.denselayer9.norm2.bias", "base.denseblock2.denselayer9.norm2.weight", "base.denseblock2.denselayer9.norm2.running_mean", "base.denseblock2.denselayer9.conv2.weight", "base.denseblock2.denselayer10.norm1.running_var", "base.denseblock2.denselayer10.norm1.bias", "base.denseblock2.denselayer10.norm1.weight", "base.denseblock2.denselayer10.norm1.running_mean", "base.denseblock2.denselayer10.conv1.weight", "base.denseblock2.denselayer10.norm2.running_var", "base.denseblock2.denselayer10.norm2.bias", "base.denseblock2.denselayer10.norm2.weight", "base.denseblock2.denselayer10.norm2.running_mean", "base.denseblock2.denselayer10.conv2.weight", "base.denseblock2.denselayer11.norm1.running_var", "base.denseblock2.denselayer11.norm1.bias", "base.denseblock2.denselayer11.norm1.weight", "base.denseblock2.denselayer11.norm1.running_mean", "base.denseblock2.denselayer11.conv1.weight", "base.denseblock2.denselayer11.norm2.running_var", "base.denseblock2.denselayer11.norm2.bias", "base.denseblock2.denselayer11.norm2.weight", "base.denseblock2.denselayer11.norm2.running_mean", "base.denseblock2.denselayer11.conv2.weight", "base.denseblock2.denselayer12.norm1.running_var", "base.denseblock2.denselayer12.norm1.bias", "base.denseblock2.denselayer12.norm1.weight", "base.denseblock2.denselayer12.norm1.running_mean", "base.denseblock2.denselayer12.conv1.weight", "base.denseblock2.denselayer12.norm2.running_var", "base.denseblock2.denselayer12.norm2.bias", "base.denseblock2.denselayer12.norm2.weight", "base.denseblock2.denselayer12.norm2.running_mean", "base.denseblock2.denselayer12.conv2.weight", "base.denseblock3.denselayer1.norm1.running_var", "base.denseblock3.denselayer1.norm1.bias", "base.denseblock3.denselayer1.norm1.weight", "base.denseblock3.denselayer1.norm1.running_mean", "base.denseblock3.denselayer1.conv1.weight", "base.denseblock3.denselayer1.norm2.running_var", "base.denseblock3.denselayer1.norm2.bias", "base.denseblock3.denselayer1.norm2.weight", "base.denseblock3.denselayer1.norm2.running_mean", "base.denseblock3.denselayer1.conv2.weight", "base.denseblock3.denselayer2.norm1.running_var", "base.denseblock3.denselayer2.norm1.bias", "base.denseblock3.denselayer2.norm1.weight", "base.denseblock3.denselayer2.norm1.running_mean", "base.denseblock3.denselayer2.conv1.weight", "base.denseblock3.denselayer2.norm2.running_var", "base.denseblock3.denselayer2.norm2.bias", "base.denseblock3.denselayer2.norm2.weight", "base.denseblock3.denselayer2.norm2.running_mean", "base.denseblock3.denselayer2.conv2.weight", "base.denseblock3.denselayer3.norm1.running_var", "base.denseblock3.denselayer3.norm1.bias", "base.denseblock3.denselayer3.norm1.weight", "base.denseblock3.denselayer3.norm1.running_mean", "base.denseblock3.denselayer3.conv1.weight", "base.denseblock3.denselayer3.norm2.running_var", "base.denseblock3.denselayer3.norm2.bias", "base.denseblock3.denselayer3.norm2.weight", "base.denseblock3.denselayer3.norm2.running_mean", "base.denseblock3.denselayer3.conv2.weight", "base.denseblock3.denselayer4.norm1.running_var", "base.denseblock3.denselayer4.norm1.bias", "base.denseblock3.denselayer4.norm1.weight", "base.denseblock3.denselayer4.norm1.running_mean", "base.denseblock3.denselayer4.conv1.weight", "base.denseblock3.denselayer4.norm2.running_var", "base.denseblock3.denselayer4.norm2.bias", "base.denseblock3.denselayer4.norm2.weight", "base.denseblock3.denselayer4.norm2.running_mean", "base.denseblock3.denselayer4.conv2.weight", "base.denseblock3.denselayer5.norm1.running_var", "base.denseblock3.denselayer5.norm1.bias", "base.denseblock3.denselayer5.norm1.weight", "base.denseblock3.denselayer5.norm1.running_mean", "base.denseblock3.denselayer5.conv1.weight", "base.denseblock3.denselayer5.norm2.running_var", "base.denseblock3.denselayer5.norm2.bias", "base.denseblock3.denselayer5.norm2.weight", "base.denseblock3.denselayer5.norm2.running_mean", "base.denseblock3.denselayer5.conv2.weight", "base.denseblock3.denselayer6.norm1.running_var", "base.denseblock3.denselayer6.norm1.bias", "base.denseblock3.denselayer6.norm1.weight", "base.denseblock3.denselayer6.norm1.running_mean", "base.denseblock3.denselayer6.conv1.weight", "base.denseblock3.denselayer6.norm2.running_var", "base.denseblock3.denselayer6.norm2.bias", "base.denseblock3.denselayer6.norm2.weight", "base.denseblock3.denselayer6.norm2.running_mean", "base.denseblock3.denselayer6.conv2.weight", "base.denseblock3.denselayer7.norm1.running_var", "base.denseblock3.denselayer7.norm1.bias", "base.denseblock3.denselayer7.norm1.weight", "base.denseblock3.denselayer7.norm1.running_mean", "base.denseblock3.denselayer7.conv1.weight", "base.denseblock3.denselayer7.norm2.running_var", "base.denseblock3.denselayer7.norm2.bias", "base.denseblock3.denselayer7.norm2.weight", "base.denseblock3.denselayer7.norm2.running_mean", "base.denseblock3.denselayer7.conv2.weight", "base.denseblock3.denselayer8.norm1.running_var", "base.denseblock3.denselayer8.norm1.bias", "base.denseblock3.denselayer8.norm1.weight", "base.denseblock3.denselayer8.norm1.running_mean", "base.denseblock3.denselayer8.conv1.weight", "base.denseblock3.denselayer8.norm2.running_var", "base.denseblock3.denselayer8.norm2.bias", "base.denseblock3.denselayer8.norm2.weight", "base.denseblock3.denselayer8.norm2.running_mean", "base.denseblock3.denselayer8.conv2.weight", "base.denseblock3.denselayer9.norm1.running_var", "base.denseblock3.denselayer9.norm1.bias", "base.denseblock3.denselayer9.norm1.weight", "base.denseblock3.denselayer9.norm1.running_mean", "base.denseblock3.denselayer9.conv1.weight", "base.denseblock3.denselayer9.norm2.running_var", "base.denseblock3.denselayer9.norm2.bias", "base.denseblock3.denselayer9.norm2.weight", "base.denseblock3.denselayer9.norm2.running_mean", "base.denseblock3.denselayer9.conv2.weight", "base.denseblock3.denselayer10.norm1.running_var", "base.denseblock3.denselayer10.norm1.bias", "base.denseblock3.denselayer10.norm1.weight", "base.denseblock3.denselayer10.norm1.running_mean", "base.denseblock3.denselayer10.conv1.weight", "base.denseblock3.denselayer10.norm2.running_var", "base.denseblock3.denselayer10.norm2.bias", "base.denseblock3.denselayer10.norm2.weight", "base.denseblock3.denselayer10.norm2.running_mean", "base.denseblock3.denselayer10.conv2.weight", "base.denseblock3.denselayer11.norm1.running_var", "base.denseblock3.denselayer11.norm1.bias", "base.denseblock3.denselayer11.norm1.weight", "base.denseblock3.denselayer11.norm1.running_mean", "base.denseblock3.denselayer11.conv1.weight", "base.denseblock3.denselayer11.norm2.running_var", "base.denseblock3.denselayer11.norm2.bias", "base.denseblock3.denselayer11.norm2.weight", "base.denseblock3.denselayer11.norm2.running_mean", "base.denseblock3.denselayer11.conv2.weight", "base.denseblock3.denselayer12.norm1.running_var", "base.denseblock3.denselayer12.norm1.bias", "base.denseblock3.denselayer12.norm1.weight", "base.denseblock3.denselayer12.norm1.running_mean", "base.denseblock3.denselayer12.conv1.weight", "base.denseblock3.denselayer12.norm2.running_var", "base.denseblock3.denselayer12.norm2.bias", "base.denseblock3.denselayer12.norm2.weight", "base.denseblock3.denselayer12.norm2.running_mean", "base.denseblock3.denselayer12.conv2.weight", "base.denseblock3.denselayer13.norm1.running_var", "base.denseblock3.denselayer13.norm1.bias", "base.denseblock3.denselayer13.norm1.weight", "base.denseblock3.denselayer13.norm1.running_mean", "base.denseblock3.denselayer13.conv1.weight", "base.denseblock3.denselayer13.norm2.running_var", "base.denseblock3.denselayer13.norm2.bias", "base.denseblock3.denselayer13.norm2.weight", "base.denseblock3.denselayer13.norm2.running_mean", "base.denseblock3.denselayer13.conv2.weight", "base.denseblock3.denselayer14.norm1.running_var", "base.denseblock3.denselayer14.norm1.bias", "base.denseblock3.denselayer14.norm1.weight", "base.denseblock3.denselayer14.norm1.running_mean", "base.denseblock3.denselayer14.conv1.weight", "base.denseblock3.denselayer14.norm2.running_var", "base.denseblock3.denselayer14.norm2.bias", "base.denseblock3.denselayer14.norm2.weight", "base.denseblock3.denselayer14.norm2.running_mean", "base.denseblock3.denselayer14.conv2.weight", "base.denseblock3.denselayer15.norm1.running_var", "base.denseblock3.denselayer15.norm1.bias", "base.denseblock3.denselayer15.norm1.weight", "base.denseblock3.denselayer15.norm1.running_mean", "base.denseblock3.denselayer15.conv1.weight", "base.denseblock3.denselayer15.norm2.running_var", "base.denseblock3.denselayer15.norm2.bias", "base.denseblock3.denselayer15.norm2.weight", "base.denseblock3.denselayer15.norm2.running_mean", "base.denseblock3.denselayer15.conv2.weight", "base.denseblock3.denselayer16.norm1.running_var", "base.denseblock3.denselayer16.norm1.bias", "base.denseblock3.denselayer16.norm1.weight", "base.denseblock3.denselayer16.norm1.running_mean", "base.denseblock3.denselayer16.conv1.weight", "base.denseblock3.denselayer16.norm2.running_var", "base.denseblock3.denselayer16.norm2.bias", "base.denseblock3.denselayer16.norm2.weight", "base.denseblock3.denselayer16.norm2.running_mean", "base.denseblock3.denselayer16.conv2.weight", "base.denseblock3.denselayer17.norm1.running_var", "base.denseblock3.denselayer17.norm1.bias", "base.denseblock3.denselayer17.norm1.weight", "base.denseblock3.denselayer17.norm1.running_mean", "base.denseblock3.denselayer17.conv1.weight", "base.denseblock3.denselayer17.norm2.running_var", "base.denseblock3.denselayer17.norm2.bias", "base.denseblock3.denselayer17.norm2.weight", "base.denseblock3.denselayer17.norm2.running_mean", "base.denseblock3.denselayer17.conv2.weight", "base.denseblock3.denselayer18.norm1.running_var", "base.denseblock3.denselayer18.norm1.bias", "base.denseblock3.denselayer18.norm1.weight", "base.denseblock3.denselayer18.norm1.running_mean", "base.denseblock3.denselayer18.conv1.weight", "base.denseblock3.denselayer18.norm2.running_var", "base.denseblock3.denselayer18.norm2.bias", "base.denseblock3.denselayer18.norm2.weight", "base.denseblock3.denselayer18.norm2.running_mean", "base.denseblock3.denselayer18.conv2.weight", "base.denseblock3.denselayer19.norm1.running_var", "base.denseblock3.denselayer19.norm1.bias", "base.denseblock3.denselayer19.norm1.weight", "base.denseblock3.denselayer19.norm1.running_mean", "base.denseblock3.denselayer19.conv1.weight", "base.denseblock3.denselayer19.norm2.running_var", "base.denseblock3.denselayer19.norm2.bias", "base.denseblock3.denselayer19.norm2.weight", "base.denseblock3.denselayer19.norm2.running_mean", "base.denseblock3.denselayer19.conv2.weight", "base.denseblock3.denselayer20.norm1.running_var", "base.denseblock3.denselayer20.norm1.bias", "base.denseblock3.denselayer20.norm1.weight", "base.denseblock3.denselayer20.norm1.running_mean", "base.denseblock3.denselayer20.conv1.weight", "base.denseblock3.denselayer20.norm2.running_var", "base.denseblock3.denselayer20.norm2.bias", "base.denseblock3.denselayer20.norm2.weight", "base.denseblock3.denselayer20.norm2.running_mean", "base.denseblock3.denselayer20.conv2.weight", "base.denseblock3.denselayer21.norm1.running_var", "base.denseblock3.denselayer21.norm1.bias", "base.denseblock3.denselayer21.norm1.weight", "base.denseblock3.denselayer21.norm1.running_mean", "base.denseblock3.denselayer21.conv1.weight", "base.denseblock3.denselayer21.norm2.running_var", "base.denseblock3.denselayer21.norm2.bias", "base.denseblock3.denselayer21.norm2.weight", "base.denseblock3.denselayer21.norm2.running_mean", "base.denseblock3.denselayer21.conv2.weight", "base.denseblock3.denselayer22.norm1.running_var", "base.denseblock3.denselayer22.norm1.bias", "base.denseblock3.denselayer22.norm1.weight", "base.denseblock3.denselayer22.norm1.running_mean", "base.denseblock3.denselayer22.conv1.weight", "base.denseblock3.denselayer22.norm2.running_var", "base.denseblock3.denselayer22.norm2.bias", "base.denseblock3.denselayer22.norm2.weight", "base.denseblock3.denselayer22.norm2.running_mean", "base.denseblock3.denselayer22.conv2.weight", "base.denseblock3.denselayer23.norm1.running_var", "base.denseblock3.denselayer23.norm1.bias", "base.denseblock3.denselayer23.norm1.weight", "base.denseblock3.denselayer23.norm1.running_mean", "base.denseblock3.denselayer23.conv1.weight", "base.denseblock3.denselayer23.norm2.running_var", "base.denseblock3.denselayer23.norm2.bias", "base.denseblock3.denselayer23.norm2.weight", "base.denseblock3.denselayer23.norm2.running_mean", "base.denseblock3.denselayer23.conv2.weight", "base.denseblock3.denselayer24.norm1.running_var", "base.denseblock3.denselayer24.norm1.bias", "base.denseblock3.denselayer24.norm1.weight", "base.denseblock3.denselayer24.norm1.running_mean", "base.denseblock3.denselayer24.conv1.weight", "base.denseblock3.denselayer24.norm2.running_var", "base.denseblock3.denselayer24.norm2.bias", "base.denseblock3.denselayer24.norm2.weight", "base.denseblock3.denselayer24.norm2.running_mean", "base.denseblock3.denselayer24.conv2.weight", "base.denseblock4.denselayer1.norm1.running_var", "base.denseblock4.denselayer1.norm1.bias", "base.denseblock4.denselayer1.norm1.weight", "base.denseblock4.denselayer1.norm1.running_mean", "base.denseblock4.denselayer1.conv1.weight", "base.denseblock4.denselayer1.norm2.running_var", "base.denseblock4.denselayer1.norm2.bias", "base.denseblock4.denselayer1.norm2.weight", "base.denseblock4.denselayer1.norm2.running_mean", "base.denseblock4.denselayer1.conv2.weight", "base.denseblock4.denselayer2.norm1.running_var", "base.denseblock4.denselayer2.norm1.bias", "base.denseblock4.denselayer2.norm1.weight", "base.denseblock4.denselayer2.norm1.running_mean", "base.denseblock4.denselayer2.conv1.weight", "base.denseblock4.denselayer2.norm2.running_var", "base.denseblock4.denselayer2.norm2.bias", "base.denseblock4.denselayer2.norm2.weight", "base.denseblock4.denselayer2.norm2.running_mean", "base.denseblock4.denselayer2.conv2.weight", "base.denseblock4.denselayer3.norm1.running_var", "base.denseblock4.denselayer3.norm1.bias", "base.denseblock4.denselayer3.norm1.weight", "base.denseblock4.denselayer3.norm1.running_mean", "base.denseblock4.denselayer3.conv1.weight", "base.denseblock4.denselayer3.norm2.running_var", "base.denseblock4.denselayer3.norm2.bias", "base.denseblock4.denselayer3.norm2.weight", "base.denseblock4.denselayer3.norm2.running_mean", "base.denseblock4.denselayer3.conv2.weight", "base.denseblock4.denselayer4.norm1.running_var", "base.denseblock4.denselayer4.norm1.bias", "base.denseblock4.denselayer4.norm1.weight", "base.denseblock4.denselayer4.norm1.running_mean", "base.denseblock4.denselayer4.conv1.weight", "base.denseblock4.denselayer4.norm2.running_var", "base.denseblock4.denselayer4.norm2.bias", "base.denseblock4.denselayer4.norm2.weight", "base.denseblock4.denselayer4.norm2.running_mean", "base.denseblock4.denselayer4.conv2.weight", "base.denseblock4.denselayer5.norm1.running_var", "base.denseblock4.denselayer5.norm1.bias", "base.denseblock4.denselayer5.norm1.weight", "base.denseblock4.denselayer5.norm1.running_mean", "base.denseblock4.denselayer5.conv1.weight", "base.denseblock4.denselayer5.norm2.running_var", "base.denseblock4.denselayer5.norm2.bias", "base.denseblock4.denselayer5.norm2.weight", "base.denseblock4.denselayer5.norm2.running_mean", "base.denseblock4.denselayer5.conv2.weight", "base.denseblock4.denselayer6.norm1.running_var", "base.denseblock4.denselayer6.norm1.bias", "base.denseblock4.denselayer6.norm1.weight", "base.denseblock4.denselayer6.norm1.running_mean", "base.denseblock4.denselayer6.conv1.weight", "base.denseblock4.denselayer6.norm2.running_var", "base.denseblock4.denselayer6.norm2.bias", "base.denseblock4.denselayer6.norm2.weight", "base.denseblock4.denselayer6.norm2.running_mean", "base.denseblock4.denselayer6.conv2.weight", "base.denseblock4.denselayer7.norm1.running_var", "base.denseblock4.denselayer7.norm1.bias", "base.denseblock4.denselayer7.norm1.weight", "base.denseblock4.denselayer7.norm1.running_mean", "base.denseblock4.denselayer7.conv1.weight", "base.denseblock4.denselayer7.norm2.running_var", "base.denseblock4.denselayer7.norm2.bias", "base.denseblock4.denselayer7.norm2.weight", "base.denseblock4.denselayer7.norm2.running_mean", "base.denseblock4.denselayer7.conv2.weight", "base.denseblock4.denselayer8.norm1.running_var", "base.denseblock4.denselayer8.norm1.bias", "base.denseblock4.denselayer8.norm1.weight", "base.denseblock4.denselayer8.norm1.running_mean", "base.denseblock4.denselayer8.conv1.weight", "base.denseblock4.denselayer8.norm2.running_var", "base.denseblock4.denselayer8.norm2.bias", "base.denseblock4.denselayer8.norm2.weight", "base.denseblock4.denselayer8.norm2.running_mean", "base.denseblock4.denselayer8.conv2.weight", "base.denseblock4.denselayer9.norm1.running_var", "base.denseblock4.denselayer9.norm1.bias", "base.denseblock4.denselayer9.norm1.weight", "base.denseblock4.denselayer9.norm1.running_mean", "base.denseblock4.denselayer9.conv1.weight", "base.denseblock4.denselayer9.norm2.running_var", "base.denseblock4.denselayer9.norm2.bias", "base.denseblock4.denselayer9.norm2.weight", "base.denseblock4.denselayer9.norm2.running_mean", "base.denseblock4.denselayer9.conv2.weight", "base.denseblock4.denselayer10.norm1.running_var", "base.denseblock4.denselayer10.norm1.bias", "base.denseblock4.denselayer10.norm1.weight", "base.denseblock4.denselayer10.norm1.running_mean", "base.denseblock4.denselayer10.conv1.weight", "base.denseblock4.denselayer10.norm2.running_var", "base.denseblock4.denselayer10.norm2.bias", "base.denseblock4.denselayer10.norm2.weight", "base.denseblock4.denselayer10.norm2.running_mean", "base.denseblock4.denselayer10.conv2.weight", "base.denseblock4.denselayer11.norm1.running_var", "base.denseblock4.denselayer11.norm1.bias", "base.denseblock4.denselayer11.norm1.weight", "base.denseblock4.denselayer11.norm1.running_mean", "base.denseblock4.denselayer11.conv1.weight", "base.denseblock4.denselayer11.norm2.running_var", "base.denseblock4.denselayer11.norm2.bias", "base.denseblock4.denselayer11.norm2.weight", "base.denseblock4.denselayer11.norm2.running_mean", "base.denseblock4.denselayer11.conv2.weight", "base.denseblock4.denselayer12.norm1.running_var", "base.denseblock4.denselayer12.norm1.bias", "base.denseblock4.denselayer12.norm1.weight", "base.denseblock4.denselayer12.norm1.running_mean", "base.denseblock4.denselayer12.conv1.weight", "base.denseblock4.denselayer12.norm2.running_var", "base.denseblock4.denselayer12.norm2.bias", "base.denseblock4.denselayer12.norm2.weight", "base.denseblock4.denselayer12.norm2.running_mean", "base.denseblock4.denselayer12.conv2.weight", "base.denseblock4.denselayer13.norm1.running_var", "base.denseblock4.denselayer13.norm1.bias", "base.denseblock4.denselayer13.norm1.weight", "base.denseblock4.denselayer13.norm1.running_mean", "base.denseblock4.denselayer13.conv1.weight", "base.denseblock4.denselayer13.norm2.running_var", "base.denseblock4.denselayer13.norm2.bias", "base.denseblock4.denselayer13.norm2.weight", "base.denseblock4.denselayer13.norm2.running_mean", "base.denseblock4.denselayer13.conv2.weight", "base.denseblock4.denselayer14.norm1.running_var", "base.denseblock4.denselayer14.norm1.bias", "base.denseblock4.denselayer14.norm1.weight", "base.denseblock4.denselayer14.norm1.running_mean", "base.denseblock4.denselayer14.conv1.weight", "base.denseblock4.denselayer14.norm2.running_var", "base.denseblock4.denselayer14.norm2.bias", "base.denseblock4.denselayer14.norm2.weight", "base.denseblock4.denselayer14.norm2.running_mean", "base.denseblock4.denselayer14.conv2.weight", "base.denseblock4.denselayer15.norm1.running_var", "base.denseblock4.denselayer15.norm1.bias", "base.denseblock4.denselayer15.norm1.weight", "base.denseblock4.denselayer15.norm1.running_mean", "base.denseblock4.denselayer15.conv1.weight", "base.denseblock4.denselayer15.norm2.running_var", "base.denseblock4.denselayer15.norm2.bias", "base.denseblock4.denselayer15.norm2.weight", "base.denseblock4.denselayer15.norm2.running_mean", "base.denseblock4.denselayer15.conv2.weight", "base.denseblock4.denselayer16.norm1.running_var", "base.denseblock4.denselayer16.norm1.bias", "base.denseblock4.denselayer16.norm1.weight", "base.denseblock4.denselayer16.norm1.running_mean", "base.denseblock4.denselayer16.conv1.weight", "base.denseblock4.denselayer16.norm2.running_var", "base.denseblock4.denselayer16.norm2.bias", "base.denseblock4.denselayer16.norm2.weight", "base.denseblock4.denselayer16.norm2.running_mean", "base.denseblock4.denselayer16.conv2.weight". 
    Unexpected key(s) in state_dict: "base.denseblock1.denselayer1.norm.1.weight", "base.denseblock1.denselayer1.norm.1.bias", "base.denseblock1.denselayer1.norm.1.running_mean", "base.denseblock1.denselayer1.norm.1.running_var", "base.denseblock1.denselayer1.conv.1.weight", "base.denseblock1.denselayer1.norm.2.weight", "base.denseblock1.denselayer1.norm.2.bias", "base.denseblock1.denselayer1.norm.2.running_mean", "base.denseblock1.denselayer1.norm.2.running_var", "base.denseblock1.denselayer1.conv.2.weight", "base.denseblock1.denselayer2.norm.1.weight", "base.denseblock1.denselayer2.norm.1.bias", "base.denseblock1.denselayer2.norm.1.running_mean", "base.denseblock1.denselayer2.norm.1.running_var", "base.denseblock1.denselayer2.conv.1.weight", "base.denseblock1.denselayer2.norm.2.weight", "base.denseblock1.denselayer2.norm.2.bias", "base.denseblock1.denselayer2.norm.2.running_mean", "base.denseblock1.denselayer2.norm.2.running_var", "base.denseblock1.denselayer2.conv.2.weight", "base.denseblock1.denselayer3.norm.1.weight", "base.denseblock1.denselayer3.norm.1.bias", "base.denseblock1.denselayer3.norm.1.running_mean", "base.denseblock1.denselayer3.norm.1.running_var", "base.denseblock1.denselayer3.conv.1.weight", "base.denseblock1.denselayer3.norm.2.weight", "base.denseblock1.denselayer3.norm.2.bias", "base.denseblock1.denselayer3.norm.2.running_mean", "base.denseblock1.denselayer3.norm.2.running_var", "base.denseblock1.denselayer3.conv.2.weight", "base.denseblock1.denselayer4.norm.1.weight", "base.denseblock1.denselayer4.norm.1.bias", "base.denseblock1.denselayer4.norm.1.running_mean", "base.denseblock1.denselayer4.norm.1.running_var", "base.denseblock1.denselayer4.conv.1.weight", "base.denseblock1.denselayer4.norm.2.weight", "base.denseblock1.denselayer4.norm.2.bias", "base.denseblock1.denselayer4.norm.2.running_mean", "base.denseblock1.denselayer4.norm.2.running_var", "base.denseblock1.denselayer4.conv.2.weight", "base.denseblock1.denselayer5.norm.1.weight", "base.denseblock1.denselayer5.norm.1.bias", "base.denseblock1.denselayer5.norm.1.running_mean", "base.denseblock1.denselayer5.norm.1.running_var", "base.denseblock1.denselayer5.conv.1.weight", "base.denseblock1.denselayer5.norm.2.weight", "base.denseblock1.denselayer5.norm.2.bias", "base.denseblock1.denselayer5.norm.2.running_mean", "base.denseblock1.denselayer5.norm.2.running_var", "base.denseblock1.denselayer5.conv.2.weight", "base.denseblock1.denselayer6.norm.1.weight", "base.denseblock1.denselayer6.norm.1.bias", "base.denseblock1.denselayer6.norm.1.running_mean", "base.denseblock1.denselayer6.norm.1.running_var", "base.denseblock1.denselayer6.conv.1.weight", "base.denseblock1.denselayer6.norm.2.weight", "base.denseblock1.denselayer6.norm.2.bias", "base.denseblock1.denselayer6.norm.2.running_mean", "base.denseblock1.denselayer6.norm.2.running_var", "base.denseblock1.denselayer6.conv.2.weight", "base.denseblock2.denselayer1.norm.1.weight", "base.denseblock2.denselayer1.norm.1.bias", "base.denseblock2.denselayer1.norm.1.running_mean", "base.denseblock2.denselayer1.norm.1.running_var", "base.denseblock2.denselayer1.conv.1.weight", "base.denseblock2.denselayer1.norm.2.weight", "base.denseblock2.denselayer1.norm.2.bias", "base.denseblock2.denselayer1.norm.2.running_mean", "base.denseblock2.denselayer1.norm.2.running_var", "base.denseblock2.denselayer1.conv.2.weight", "base.denseblock2.denselayer2.norm.1.weight", "base.denseblock2.denselayer2.norm.1.bias", "base.denseblock2.denselayer2.norm.1.running_mean", "base.denseblock2.denselayer2.norm.1.running_var", "base.denseblock2.denselayer2.conv.1.weight", "base.denseblock2.denselayer2.norm.2.weight", "base.denseblock2.denselayer2.norm.2.bias", "base.denseblock2.denselayer2.norm.2.running_mean", "base.denseblock2.denselayer2.norm.2.running_var", "base.denseblock2.denselayer2.conv.2.weight", "base.denseblock2.denselayer3.norm.1.weight", "base.denseblock2.denselayer3.norm.1.bias", "base.denseblock2.denselayer3.norm.1.running_mean", "base.denseblock2.denselayer3.norm.1.running_var", "base.denseblock2.denselayer3.conv.1.weight", "base.denseblock2.denselayer3.norm.2.weight", "base.denseblock2.denselayer3.norm.2.bias", "base.denseblock2.denselayer3.norm.2.running_mean", "base.denseblock2.denselayer3.norm.2.running_var", "base.denseblock2.denselayer3.conv.2.weight", "base.denseblock2.denselayer4.norm.1.weight", "base.denseblock2.denselayer4.norm.1.bias", "base.denseblock2.denselayer4.norm.1.running_mean", "base.denseblock2.denselayer4.norm.1.running_var", "base.denseblock2.denselayer4.conv.1.weight", "base.denseblock2.denselayer4.norm.2.weight", "base.denseblock2.denselayer4.norm.2.bias", "base.denseblock2.denselayer4.norm.2.running_mean", "base.denseblock2.denselayer4.norm.2.running_var", "base.denseblock2.denselayer4.conv.2.weight", "base.denseblock2.denselayer5.norm.1.weight", "base.denseblock2.denselayer5.norm.1.bias", "base.denseblock2.denselayer5.norm.1.running_mean", "base.denseblock2.denselayer5.norm.1.running_var", "base.denseblock2.denselayer5.conv.1.weight", "base.denseblock2.denselayer5.norm.2.weight", "base.denseblock2.denselayer5.norm.2.bias", "base.denseblock2.denselayer5.norm.2.running_mean", "base.denseblock2.denselayer5.norm.2.running_var", "base.denseblock2.denselayer5.conv.2.weight", "base.denseblock2.denselayer6.norm.1.weight", "base.denseblock2.denselayer6.norm.1.bias", "base.denseblock2.denselayer6.norm.1.running_mean", "base.denseblock2.denselayer6.norm.1.running_var", "base.denseblock2.denselayer6.conv.1.weight", "base.denseblock2.denselayer6.norm.2.weight", "base.denseblock2.denselayer6.norm.2.bias", "base.denseblock2.denselayer6.norm.2.running_mean", "base.denseblock2.denselayer6.norm.2.running_var", "base.denseblock2.denselayer6.conv.2.weight", "base.denseblock2.denselayer7.norm.1.weight", "base.denseblock2.denselayer7.norm.1.bias", "base.denseblock2.denselayer7.norm.1.running_mean", "base.denseblock2.denselayer7.norm.1.running_var", "base.denseblock2.denselayer7.conv.1.weight", "base.denseblock2.denselayer7.norm.2.weight", "base.denseblock2.denselayer7.norm.2.bias", "base.denseblock2.denselayer7.norm.2.running_mean", "base.denseblock2.denselayer7.norm.2.running_var", "base.denseblock2.denselayer7.conv.2.weight", "base.denseblock2.denselayer8.norm.1.weight", "base.denseblock2.denselayer8.norm.1.bias", "base.denseblock2.denselayer8.norm.1.running_mean", "base.denseblock2.denselayer8.norm.1.running_var", "base.denseblock2.denselayer8.conv.1.weight", "base.denseblock2.denselayer8.norm.2.weight", "base.denseblock2.denselayer8.norm.2.bias", "base.denseblock2.denselayer8.norm.2.running_mean", "base.denseblock2.denselayer8.norm.2.running_var", "base.denseblock2.denselayer8.conv.2.weight", "base.denseblock2.denselayer9.norm.1.weight", "base.denseblock2.denselayer9.norm.1.bias", "base.denseblock2.denselayer9.norm.1.running_mean", "base.denseblock2.denselayer9.norm.1.running_var", "base.denseblock2.denselayer9.conv.1.weight", "base.denseblock2.denselayer9.norm.2.weight", "base.denseblock2.denselayer9.norm.2.bias", "base.denseblock2.denselayer9.norm.2.running_mean", "base.denseblock2.denselayer9.norm.2.running_var", "base.denseblock2.denselayer9.conv.2.weight", "base.denseblock2.denselayer10.norm.1.weight", "base.denseblock2.denselayer10.norm.1.bias", "base.denseblock2.denselayer10.norm.1.running_mean", "base.denseblock2.denselayer10.norm.1.running_var", "base.denseblock2.denselayer10.conv.1.weight", "base.denseblock2.denselayer10.norm.2.weight", "base.denseblock2.denselayer10.norm.2.bias", "base.denseblock2.denselayer10.norm.2.running_mean", "base.denseblock2.denselayer10.norm.2.running_var", "base.denseblock2.denselayer10.conv.2.weight", "base.denseblock2.denselayer11.norm.1.weight", "base.denseblock2.denselayer11.norm.1.bias", "base.denseblock2.denselayer11.norm.1.running_mean", "base.denseblock2.denselayer11.norm.1.running_var", "base.denseblock2.denselayer11.conv.1.weight", "base.denseblock2.denselayer11.norm.2.weight", "base.denseblock2.denselayer11.norm.2.bias", "base.denseblock2.denselayer11.norm.2.running_mean", "base.denseblock2.denselayer11.norm.2.running_var", "base.denseblock2.denselayer11.conv.2.weight", "base.denseblock2.denselayer12.norm.1.weight", "base.denseblock2.denselayer12.norm.1.bias", "base.denseblock2.denselayer12.norm.1.running_mean", "base.denseblock2.denselayer12.norm.1.running_var", "base.denseblock2.denselayer12.conv.1.weight", "base.denseblock2.denselayer12.norm.2.weight", "base.denseblock2.denselayer12.norm.2.bias", "base.denseblock2.denselayer12.norm.2.running_mean", "base.denseblock2.denselayer12.norm.2.running_var", "base.denseblock2.denselayer12.conv.2.weight", "base.denseblock3.denselayer1.norm.1.weight", "base.denseblock3.denselayer1.norm.1.bias", "base.denseblock3.denselayer1.norm.1.running_mean", "base.denseblock3.denselayer1.norm.1.running_var", "base.denseblock3.denselayer1.conv.1.weight", "base.denseblock3.denselayer1.norm.2.weight", "base.denseblock3.denselayer1.norm.2.bias", "base.denseblock3.denselayer1.norm.2.running_mean", "base.denseblock3.denselayer1.norm.2.running_var", "base.denseblock3.denselayer1.conv.2.weight", "base.denseblock3.denselayer2.norm.1.weight", "base.denseblock3.denselayer2.norm.1.bias", "base.denseblock3.denselayer2.norm.1.running_mean", "base.denseblock3.denselayer2.norm.1.running_var", "base.denseblock3.denselayer2.conv.1.weight", "base.denseblock3.denselayer2.norm.2.weight", "base.denseblock3.denselayer2.norm.2.bias", "base.denseblock3.denselayer2.norm.2.running_mean", "base.denseblock3.denselayer2.norm.2.running_var", "base.denseblock3.denselayer2.conv.2.weight", "base.denseblock3.denselayer3.norm.1.weight", "base.denseblock3.denselayer3.norm.1.bias", "base.denseblock3.denselayer3.norm.1.running_mean", "base.denseblock3.denselayer3.norm.1.running_var", "base.denseblock3.denselayer3.conv.1.weight", "base.denseblock3.denselayer3.norm.2.weight", "base.denseblock3.denselayer3.norm.2.bias", "base.denseblock3.denselayer3.norm.2.running_mean", "base.denseblock3.denselayer3.norm.2.running_var", "base.denseblock3.denselayer3.conv.2.weight", "base.denseblock3.denselayer4.norm.1.weight", "base.denseblock3.denselayer4.norm.1.bias", "base.denseblock3.denselayer4.norm.1.running_mean", "base.denseblock3.denselayer4.norm.1.running_var", "base.denseblock3.denselayer4.conv.1.weight", "base.denseblock3.denselayer4.norm.2.weight", "base.denseblock3.denselayer4.norm.2.bias", "base.denseblock3.denselayer4.norm.2.running_mean", "base.denseblock3.denselayer4.norm.2.running_var", "base.denseblock3.denselayer4.conv.2.weight", "base.denseblock3.denselayer5.norm.1.weight", "base.denseblock3.denselayer5.norm.1.bias", "base.denseblock3.denselayer5.norm.1.running_mean", "base.denseblock3.denselayer5.norm.1.running_var", "base.denseblock3.denselayer5.conv.1.weight", "base.denseblock3.denselayer5.norm.2.weight", "base.denseblock3.denselayer5.norm.2.bias", "base.denseblock3.denselayer5.norm.2.running_mean", "base.denseblock3.denselayer5.norm.2.running_var", "base.denseblock3.denselayer5.conv.2.weight", "base.denseblock3.denselayer6.norm.1.weight", "base.denseblock3.denselayer6.norm.1.bias", "base.denseblock3.denselayer6.norm.1.running_mean", "base.denseblock3.denselayer6.norm.1.running_var", "base.denseblock3.denselayer6.conv.1.weight", "base.denseblock3.denselayer6.norm.2.weight", "base.denseblock3.denselayer6.norm.2.bias", "base.denseblock3.denselayer6.norm.2.running_mean", "base.denseblock3.denselayer6.norm.2.running_var", "base.denseblock3.denselayer6.conv.2.weight", "base.denseblock3.denselayer7.norm.1.weight", "base.denseblock3.denselayer7.norm.1.bias", "base.denseblock3.denselayer7.norm.1.running_mean", "base.denseblock3.denselayer7.norm.1.running_var", "base.denseblock3.denselayer7.conv.1.weight", "base.denseblock3.denselayer7.norm.2.weight", "base.denseblock3.denselayer7.norm.2.bias", "base.denseblock3.denselayer7.norm.2.running_mean", "base.denseblock3.denselayer7.norm.2.running_var", "base.denseblock3.denselayer7.conv.2.weight", "base.denseblock3.denselayer8.norm.1.weight", "base.denseblock3.denselayer8.norm.1.bias", "base.denseblock3.denselayer8.norm.1.running_mean", "base.denseblock3.denselayer8.norm.1.running_var", "base.denseblock3.denselayer8.conv.1.weight", "base.denseblock3.denselayer8.norm.2.weight", "base.denseblock3.denselayer8.norm.2.bias", "base.denseblock3.denselayer8.norm.2.running_mean", "base.denseblock3.denselayer8.norm.2.running_var", "base.denseblock3.denselayer8.conv.2.weight", "base.denseblock3.denselayer9.norm.1.weight", "base.denseblock3.denselayer9.norm.1.bias", "base.denseblock3.denselayer9.norm.1.running_mean", "base.denseblock3.denselayer9.norm.1.running_var", "base.denseblock3.denselayer9.conv.1.weight", "base.denseblock3.denselayer9.norm.2.weight", "base.denseblock3.denselayer9.norm.2.bias", "base.denseblock3.denselayer9.norm.2.running_mean", "base.denseblock3.denselayer9.norm.2.running_var", "base.denseblock3.denselayer9.conv.2.weight", "base.denseblock3.denselayer10.norm.1.weight", "base.denseblock3.denselayer10.norm.1.bias", "base.denseblock3.denselayer10.norm.1.running_mean", "base.denseblock3.denselayer10.norm.1.running_var", "base.denseblock3.denselayer10.conv.1.weight", "base.denseblock3.denselayer10.norm.2.weight", "base.denseblock3.denselayer10.norm.2.bias", "base.denseblock3.denselayer10.norm.2.running_mean", "base.denseblock3.denselayer10.norm.2.running_var", "base.denseblock3.denselayer10.conv.2.weight", "base.denseblock3.denselayer11.norm.1.weight", "base.denseblock3.denselayer11.norm.1.bias", "base.denseblock3.denselayer11.norm.1.running_mean", "base.denseblock3.denselayer11.norm.1.running_var", "base.denseblock3.denselayer11.conv.1.weight", "base.denseblock3.denselayer11.norm.2.weight", "base.denseblock3.denselayer11.norm.2.bias", "base.denseblock3.denselayer11.norm.2.running_mean", "base.denseblock3.denselayer11.norm.2.running_var", "base.denseblock3.denselayer11.conv.2.weight", "base.denseblock3.denselayer12.norm.1.weight", "base.denseblock3.denselayer12.norm.1.bias", "base.denseblock3.denselayer12.norm.1.running_mean", "base.denseblock3.denselayer12.norm.1.running_var", "base.denseblock3.denselayer12.conv.1.weight", "base.denseblock3.denselayer12.norm.2.weight", "base.denseblock3.denselayer12.norm.2.bias", "base.denseblock3.denselayer12.norm.2.running_mean", "base.denseblock3.denselayer12.norm.2.running_var", "base.denseblock3.denselayer12.conv.2.weight", "base.denseblock3.denselayer13.norm.1.weight", "base.denseblock3.denselayer13.norm.1.bias", "base.denseblock3.denselayer13.norm.1.running_mean", "base.denseblock3.denselayer13.norm.1.running_var", "base.denseblock3.denselayer13.conv.1.weight", "base.denseblock3.denselayer13.norm.2.weight", "base.denseblock3.denselayer13.norm.2.bias", "base.denseblock3.denselayer13.norm.2.running_mean", "base.denseblock3.denselayer13.norm.2.running_var", "base.denseblock3.denselayer13.conv.2.weight", "base.denseblock3.denselayer14.norm.1.weight", "base.denseblock3.denselayer14.norm.1.bias", "base.denseblock3.denselayer14.norm.1.running_mean", "base.denseblock3.denselayer14.norm.1.running_var", "base.denseblock3.denselayer14.conv.1.weight", "base.denseblock3.denselayer14.norm.2.weight", "base.denseblock3.denselayer14.norm.2.bias", "base.denseblock3.denselayer14.norm.2.running_mean", "base.denseblock3.denselayer14.norm.2.running_var", "base.denseblock3.denselayer14.conv.2.weight", "base.denseblock3.denselayer15.norm.1.weight", "base.denseblock3.denselayer15.norm.1.bias", "base.denseblock3.denselayer15.norm.1.running_mean", "base.denseblock3.denselayer15.norm.1.running_var", "base.denseblock3.denselayer15.conv.1.weight", "base.denseblock3.denselayer15.norm.2.weight", "base.denseblock3.denselayer15.norm.2.bias", "base.denseblock3.denselayer15.norm.2.running_mean", "base.denseblock3.denselayer15.norm.2.running_var", "base.denseblock3.denselayer15.conv.2.weight", "base.denseblock3.denselayer16.norm.1.weight", "base.denseblock3.denselayer16.norm.1.bias", "base.denseblock3.denselayer16.norm.1.running_mean", "base.denseblock3.denselayer16.norm.1.running_var", "base.denseblock3.denselayer16.conv.1.weight", "base.denseblock3.denselayer16.norm.2.weight", "base.denseblock3.denselayer16.norm.2.bias", "base.denseblock3.denselayer16.norm.2.running_mean", "base.denseblock3.denselayer16.norm.2.running_var", "base.denseblock3.denselayer16.conv.2.weight", "base.denseblock3.denselayer17.norm.1.weight", "base.denseblock3.denselayer17.norm.1.bias", "base.denseblock3.denselayer17.norm.1.running_mean", "base.denseblock3.denselayer17.norm.1.running_var", "base.denseblock3.denselayer17.conv.1.weight", "base.denseblock3.denselayer17.norm.2.weight", "base.denseblock3.denselayer17.norm.2.bias", "base.denseblock3.denselayer17.norm.2.running_mean", "base.denseblock3.denselayer17.norm.2.running_var", "base.denseblock3.denselayer17.conv.2.weight", "base.denseblock3.denselayer18.norm.1.weight", "base.denseblock3.denselayer18.norm.1.bias", "base.denseblock3.denselayer18.norm.1.running_mean", "base.denseblock3.denselayer18.norm.1.running_var", "base.denseblock3.denselayer18.conv.1.weight", "base.denseblock3.denselayer18.norm.2.weight", "base.denseblock3.denselayer18.norm.2.bias", "base.denseblock3.denselayer18.norm.2.running_mean", "base.denseblock3.denselayer18.norm.2.running_var", "base.denseblock3.denselayer18.conv.2.weight", "base.denseblock3.denselayer19.norm.1.weight", "base.denseblock3.denselayer19.norm.1.bias", "base.denseblock3.denselayer19.norm.1.running_mean", "base.denseblock3.denselayer19.norm.1.running_var", "base.denseblock3.denselayer19.conv.1.weight", "base.denseblock3.denselayer19.norm.2.weight", "base.denseblock3.denselayer19.norm.2.bias", "base.denseblock3.denselayer19.norm.2.running_mean", "base.denseblock3.denselayer19.norm.2.running_var", "base.denseblock3.denselayer19.conv.2.weight", "base.denseblock3.denselayer20.norm.1.weight", "base.denseblock3.denselayer20.norm.1.bias", "base.denseblock3.denselayer20.norm.1.running_mean", "base.denseblock3.denselayer20.norm.1.running_var", "base.denseblock3.denselayer20.conv.1.weight", "base.denseblock3.denselayer20.norm.2.weight", "base.denseblock3.denselayer20.norm.2.bias", "base.denseblock3.denselayer20.norm.2.running_mean", "base.denseblock3.denselayer20.norm.2.running_var", "base.denseblock3.denselayer20.conv.2.weight", "base.denseblock3.denselayer21.norm.1.weight", "base.denseblock3.denselayer21.norm.1.bias", "base.denseblock3.denselayer21.norm.1.running_mean", "base.denseblock3.denselayer21.norm.1.running_var", "base.denseblock3.denselayer21.conv.1.weight", "base.denseblock3.denselayer21.norm.2.weight", "base.denseblock3.denselayer21.norm.2.bias", "base.denseblock3.denselayer21.norm.2.running_mean", "base.denseblock3.denselayer21.norm.2.running_var", "base.denseblock3.denselayer21.conv.2.weight", "base.denseblock3.denselayer22.norm.1.weight", "base.denseblock3.denselayer22.norm.1.bias", "base.denseblock3.denselayer22.norm.1.running_mean", "base.denseblock3.denselayer22.norm.1.running_var", "base.denseblock3.denselayer22.conv.1.weight", "base.denseblock3.denselayer22.norm.2.weight", "base.denseblock3.denselayer22.norm.2.bias", "base.denseblock3.denselayer22.norm.2.running_mean", "base.denseblock3.denselayer22.norm.2.running_var", "base.denseblock3.denselayer22.conv.2.weight", "base.denseblock3.denselayer23.norm.1.weight", "base.denseblock3.denselayer23.norm.1.bias", "base.denseblock3.denselayer23.norm.1.running_mean", "base.denseblock3.denselayer23.norm.1.running_var", "base.denseblock3.denselayer23.conv.1.weight", "base.denseblock3.denselayer23.norm.2.weight", "base.denseblock3.denselayer23.norm.2.bias", "base.denseblock3.denselayer23.norm.2.running_mean", "base.denseblock3.denselayer23.norm.2.running_var", "base.denseblock3.denselayer23.conv.2.weight", "base.denseblock3.denselayer24.norm.1.weight", "base.denseblock3.denselayer24.norm.1.bias", "base.denseblock3.denselayer24.norm.1.running_mean", "base.denseblock3.denselayer24.norm.1.running_var", "base.denseblock3.denselayer24.conv.1.weight", "base.denseblock3.denselayer24.norm.2.weight", "base.denseblock3.denselayer24.norm.2.bias", "base.denseblock3.denselayer24.norm.2.running_mean", "base.denseblock3.denselayer24.norm.2.running_var", "base.denseblock3.denselayer24.conv.2.weight", "base.denseblock4.denselayer1.norm.1.weight", "base.denseblock4.denselayer1.norm.1.bias", "base.denseblock4.denselayer1.norm.1.running_mean", "base.denseblock4.denselayer1.norm.1.running_var", "base.denseblock4.denselayer1.conv.1.weight", "base.denseblock4.denselayer1.norm.2.weight", "base.denseblock4.denselayer1.norm.2.bias", "base.denseblock4.denselayer1.norm.2.running_mean", "base.denseblock4.denselayer1.norm.2.running_var", "base.denseblock4.denselayer1.conv.2.weight", "base.denseblock4.denselayer2.norm.1.weight", "base.denseblock4.denselayer2.norm.1.bias", "base.denseblock4.denselayer2.norm.1.running_mean", "base.denseblock4.denselayer2.norm.1.running_var", "base.denseblock4.denselayer2.conv.1.weight", "base.denseblock4.denselayer2.norm.2.weight", "base.denseblock4.denselayer2.norm.2.bias", "base.denseblock4.denselayer2.norm.2.running_mean", "base.denseblock4.denselayer2.norm.2.running_var", "base.denseblock4.denselayer2.conv.2.weight", "base.denseblock4.denselayer3.norm.1.weight", "base.denseblock4.denselayer3.norm.1.bias", "base.denseblock4.denselayer3.norm.1.running_mean", "base.denseblock4.denselayer3.norm.1.running_var", "base.denseblock4.denselayer3.conv.1.weight", "base.denseblock4.denselayer3.norm.2.weight", "base.denseblock4.denselayer3.norm.2.bias", "base.denseblock4.denselayer3.norm.2.running_mean", "base.denseblock4.denselayer3.norm.2.running_var", "base.denseblock4.denselayer3.conv.2.weight", "base.denseblock4.denselayer4.norm.1.weight", "base.denseblock4.denselayer4.norm.1.bias", "base.denseblock4.denselayer4.norm.1.running_mean", "base.denseblock4.denselayer4.norm.1.running_var", "base.denseblock4.denselayer4.conv.1.weight", "base.denseblock4.denselayer4.norm.2.weight", "base.denseblock4.denselayer4.norm.2.bias", "base.denseblock4.denselayer4.norm.2.running_mean", "base.denseblock4.denselayer4.norm.2.running_var", "base.denseblock4.denselayer4.conv.2.weight", "base.denseblock4.denselayer5.norm.1.weight", "base.denseblock4.denselayer5.norm.1.bias", "base.denseblock4.denselayer5.norm.1.running_mean", "base.denseblock4.denselayer5.norm.1.running_var", "base.denseblock4.denselayer5.conv.1.weight", "base.denseblock4.denselayer5.norm.2.weight", "base.denseblock4.denselayer5.norm.2.bias", "base.denseblock4.denselayer5.norm.2.running_mean", "base.denseblock4.denselayer5.norm.2.running_var", "base.denseblock4.denselayer5.conv.2.weight", "base.denseblock4.denselayer6.norm.1.weight", "base.denseblock4.denselayer6.norm.1.bias", "base.denseblock4.denselayer6.norm.1.running_mean", "base.denseblock4.denselayer6.norm.1.running_var", "base.denseblock4.denselayer6.conv.1.weight", "base.denseblock4.denselayer6.norm.2.weight", "base.denseblock4.denselayer6.norm.2.bias", "base.denseblock4.denselayer6.norm.2.running_mean", "base.denseblock4.denselayer6.norm.2.running_var", "base.denseblock4.denselayer6.conv.2.weight", "base.denseblock4.denselayer7.norm.1.weight", "base.denseblock4.denselayer7.norm.1.bias", "base.denseblock4.denselayer7.norm.1.running_mean", "base.denseblock4.denselayer7.norm.1.running_var", "base.denseblock4.denselayer7.conv.1.weight", "base.denseblock4.denselayer7.norm.2.weight", "base.denseblock4.denselayer7.norm.2.bias", "base.denseblock4.denselayer7.norm.2.running_mean", "base.denseblock4.denselayer7.norm.2.running_var", "base.denseblock4.denselayer7.conv.2.weight", "base.denseblock4.denselayer8.norm.1.weight", "base.denseblock4.denselayer8.norm.1.bias", "base.denseblock4.denselayer8.norm.1.running_mean", "base.denseblock4.denselayer8.norm.1.running_var", "base.denseblock4.denselayer8.conv.1.weight", "base.denseblock4.denselayer8.norm.2.weight", "base.denseblock4.denselayer8.norm.2.bias", "base.denseblock4.denselayer8.norm.2.running_mean", "base.denseblock4.denselayer8.norm.2.running_var", "base.denseblock4.denselayer8.conv.2.weight", "base.denseblock4.denselayer9.norm.1.weight", "base.denseblock4.denselayer9.norm.1.bias", "base.denseblock4.denselayer9.norm.1.running_mean", "base.denseblock4.denselayer9.norm.1.running_var", "base.denseblock4.denselayer9.conv.1.weight", "base.denseblock4.denselayer9.norm.2.weight", "base.denseblock4.denselayer9.norm.2.bias", "base.denseblock4.denselayer9.norm.2.running_mean", "base.denseblock4.denselayer9.norm.2.running_var", "base.denseblock4.denselayer9.conv.2.weight", "base.denseblock4.denselayer10.norm.1.weight", "base.denseblock4.denselayer10.norm.1.bias", "base.denseblock4.denselayer10.norm.1.running_mean", "base.denseblock4.denselayer10.norm.1.running_var", "base.denseblock4.denselayer10.conv.1.weight", "base.denseblock4.denselayer10.norm.2.weight", "base.denseblock4.denselayer10.norm.2.bias", "base.denseblock4.denselayer10.norm.2.running_mean", "base.denseblock4.denselayer10.norm.2.running_var", "base.denseblock4.denselayer10.conv.2.weight", "base.denseblock4.denselayer11.norm.1.weight", "base.denseblock4.denselayer11.norm.1.bias", "base.denseblock4.denselayer11.norm.1.running_mean", "base.denseblock4.denselayer11.norm.1.running_var", "base.denseblock4.denselayer11.conv.1.weight", "base.denseblock4.denselayer11.norm.2.weight", "base.denseblock4.denselayer11.norm.2.bias", "base.denseblock4.denselayer11.norm.2.running_mean", "base.denseblock4.denselayer11.norm.2.running_var", "base.denseblock4.denselayer11.conv.2.weight", "base.denseblock4.denselayer12.norm.1.weight", "base.denseblock4.denselayer12.norm.1.bias", "base.denseblock4.denselayer12.norm.1.running_mean", "base.denseblock4.denselayer12.norm.1.running_var", "base.denseblock4.denselayer12.conv.1.weight", "base.denseblock4.denselayer12.norm.2.weight", "base.denseblock4.denselayer12.norm.2.bias", "base.denseblock4.denselayer12.norm.2.running_mean", "base.denseblock4.denselayer12.norm.2.running_var", "base.denseblock4.denselayer12.conv.2.weight", "base.denseblock4.denselayer13.norm.1.weight", "base.denseblock4.denselayer13.norm.1.bias", "base.denseblock4.denselayer13.norm.1.running_mean", "base.denseblock4.denselayer13.norm.1.running_var", "base.denseblock4.denselayer13.conv.1.weight", "base.denseblock4.denselayer13.norm.2.weight", "base.denseblock4.denselayer13.norm.2.bias", "base.denseblock4.denselayer13.norm.2.running_mean", "base.denseblock4.denselayer13.norm.2.running_var", "base.denseblock4.denselayer13.conv.2.weight", "base.denseblock4.denselayer14.norm.1.weight", "base.denseblock4.denselayer14.norm.1.bias", "base.denseblock4.denselayer14.norm.1.running_mean", "base.denseblock4.denselayer14.norm.1.running_var", "base.denseblock4.denselayer14.conv.1.weight", "base.denseblock4.denselayer14.norm.2.weight", "base.denseblock4.denselayer14.norm.2.bias", "base.denseblock4.denselayer14.norm.2.running_mean", "base.denseblock4.denselayer14.norm.2.running_var", "base.denseblock4.denselayer14.conv.2.weight", "base.denseblock4.denselayer15.norm.1.weight", "base.denseblock4.denselayer15.norm.1.bias", "base.denseblock4.denselayer15.norm.1.running_mean", "base.denseblock4.denselayer15.norm.1.running_var", "base.denseblock4.denselayer15.conv.1.weight", "base.denseblock4.denselayer15.norm.2.weight", "base.denseblock4.denselayer15.norm.2.bias", "base.denseblock4.denselayer15.norm.2.running_mean", "base.denseblock4.denselayer15.norm.2.running_var", "base.denseblock4.denselayer15.conv.2.weight", "base.denseblock4.denselayer16.norm.1.weight", "base.denseblock4.denselayer16.norm.1.bias", "base.denseblock4.denselayer16.norm.1.running_mean", "base.denseblock4.denselayer16.norm.1.running_var", "base.denseblock4.denselayer16.conv.1.weight", "base.denseblock4.denselayer16.norm.2.weight", "base.denseblock4.denselayer16.norm.2.bias", "base.denseblock4.denselayer16.norm.2.running_mean", "base.denseblock4.denselayer16.norm.2.running_var", "base.denseblock4.denselayer16.conv.2.weight".
KaiyangZhou commented 6 years ago

The error is caused by the mismatch in keys, e.g. layers were named 'norm.1', 'conv.1', but are now named 'norm1', 'conv1' (I trained the model with the old torchvision). You can either manually update the keys in the state_dict or train a new one (try the default param setting).

ghost commented 6 years ago

Thank you for your prompt and useful reply! I modified "train_vid_model_xent_htri.py",and the problem has been solved!

The following code is for upgrading old TorchVision models to the latest version (For DenseNet model ONLY). In "train_vid_model_xent_htri.py":

import re

...

    if args.resume:
        print("Loading checkpoint from '{}'".format(args.resume))
        checkpoint = torch.load(args.resume)

#modify:
        # '.'s are no longer allowed in module names, but pervious _DenseLayer
        # has keys 'norm.1', 'relu.1', 'conv.1', 'norm.2', 'relu.2', 'conv.2'.
        # They are also in the checkpoints in model_urls. 
        # This pattern is used to find such keys.
        pattern = re.compile(
            r'^(.*denselayer\d+\.(?:norm|relu|conv))\.((?:[12])\.(?:weight|bias|running_mean|running_var))$')
        state_dict = checkpoint['state_dict']
        for key in list(state_dict.keys()):
            res = pattern.match(key)
            if res:
                new_key = res.group(1) + res.group(2)
                state_dict[new_key] = state_dict[key]
                del state_dict[key]
        model.load_state_dict(state_dict)

        #model.load_state_dict(checkpoint['state_dict'])
        start_epoch = checkpoint['epoch']

code reference: torchvision/densenet.py