Question about the use of pruningNet model

[liwenwei123]

Hello liuzechun, Thanks for your brilliant work.But i have some problem about how to use the pruningNet model you provide. I download the Model-MetaP-Mbv1-0.75 you provide in readme and want to evaluate the Flops and Acc1 in chart. I use the evaluate.py.But when i load the checkpoint.pth.tar i download,it says there is error in loading state_dict for DataParallel.like... ....size mismatch for module.feature.0.conv1.weight: copying a param with shape torch.Size([25, 3, 3, 3]) from checkpoint, the shape in current model is torch.Size([31, 3, 3, 3]). size mismatch for module.feature.0.bn1.weight: copying a param with shape torch.Size([25]) from checkpoint, the shape in current model is torch.Size([31])...... The code i load the stat_dict is : model = MobileNetV1() logging.info(model) model = nn.DataParallel(model).cuda() checkpoint = torch.load(checkpoint_tar) model.load_state_dict(checkpoint['state_dict']) I print the model you define in MobileNetV1.py,it's different from the structure i download from readme.Is the way i use it incorrect? Thanks again and looking forward to your reply.

Anna

[liuzechun]

Dear Anna, Thank you for your nice word, and for your interest in my work. I will check the model for you soon!

[diligent321]

hello, are you clear about how to use the pruningNet model to evaluate the Acc1. About more discription, please refer to another similar issue https://github.com/liuzechun/MetaPruning/issues/33,

appreciated for your kindly reply. @liwenwei123