Wrong format of pre-trained models

[SourceKim]

I can't unarchive them by tar or 7zip.
Here are the log:

7z.exe a mobilenetv2_total_with_prior_channel.tar

7-Zip 19.00 (x64) : Copyright (c) 1999-2018 Igor Pavlov : 2019-02-21

Open archive: mobilenetv2_total_with_prior_channel.tar
ERROR: mobilenetv2_total_with_prior_channel.tar
mobilenetv2_total_with_prior_channel.tar
Open ERROR: Can not open the file as [tar] archive

ERRORS:
Is not archive

[ypw1996]

Hello， I meet the same problem. Have you fixed it?

[Novaal]

You don't need to extract them. Just rename them accordingly to the used config file: e.g myexp/mobilenetv2_video_total/single_224_without_group/model_best.pth.tar.

[NhatRio]

Hi @dong-x16 , Thanks a lot for your impressive work.
I got an error while loading the pre-trained model. Specifically, it was at "checkpoint_video = torch.load(bestModelFile)" in the file VideoTest.ipynb. Here is the error:

    53 if os.path.isfile(bestModelFile):
---> 54     checkpoint_video = torch.load(bestModelFile)
     55     netmodel_video.load_state_dict(checkpoint_video['state_dict'])
     56     print ("minLoss: ", checkpoint_video['minLoss'], checkpoint_video['epoch'])

1 frames
/usr/local/lib/python3.6/dist-packages/torch/serialization.py in _legacy_load(f, map_location, pickle_module, **pickle_load_args)
    763     unpickler = pickle_module.Unpickler(f, **pickle_load_args)
    764     unpickler.persistent_load = persistent_load
--> 765     result = unpickler.load()
    766 
    767     deserialized_storage_keys = pickle_module.load(f, **pickle_load_args)

UnicodeDecodeError: 'utf-8' codec can't decode byte 0xc0 in position 0: invalid start byte   

Do you think this problem caused by the file "model_best.pth.tar"?

Have a nice day!

Nhat

[gruossomonica]

Hi,
I get a similar error when I load the model mobilenetv2_total_with_prior_channel.tar in VideoTest.ipynb.
I changed the model_root and the model filename accordingly with my folders and files.

---------------------------------------------------------------------------
UnicodeDecodeError                        Traceback (most recent call last)
<ipython-input-13-0ef5f99e810e> in <module>
     51 bestModelFile = os.path.join(exp_args.model_root, 'mobilenetv2_total_with_prior_channel.tar')
     52 if os.path.isfile(bestModelFile):
---> 53     checkpoint_video = torch.load(bestModelFile)
     54     netmodel_video.load_state_dict(checkpoint_video['state_dict'])
     55     print ("minLoss: ", checkpoint_video['minLoss'], checkpoint_video['epoch'])

~\AppData\Local\Programs\Python\Python36\lib\site-packages\torch\serialization.py in load(f, map_location, pickle_module)
    366         f = open(f, 'rb')
    367     try:
--> 368         return _load(f, map_location, pickle_module)
    369     finally:
    370         if new_fd:

~\AppData\Local\Programs\Python\Python36\lib\site-packages\torch\serialization.py in _load(f, map_location, pickle_module)
    540     unpickler = pickle_module.Unpickler(f)
    541     unpickler.persistent_load = persistent_load
--> 542     result = unpickler.load()
    543 
    544     deserialized_storage_keys = pickle_module.load(f)

UnicodeDecodeError: 'ascii' codec can't decode byte 0xb7 in position 3: ordinal not in range(128)

Looking forward to your reply.
Best regards.

Monica Gruosso

[shivSD]

as per requirement you need to use python 2.7 not 3.x

[DAVID-Hown]

Hello, have you solved it? I found a decompression error after downloading the model file。

[vigneshchandar]

For the weight file, I'd downloaded, the encoding was in 'latin1' format.

Try passing encoding='latin1' when loading the file.
torch.load(weight_file_path, map_location=torch.device('cpu'), encoding='latin1')