assert x.size()[1] == self.in_feats AssertionError

[zabir-nabil]

Hi, thanks for sharing this amazing repository.

When using aamsoftmax, I am getting this error for all of the models.

Save path: exps/nist_old_bang_youtube_aamsoft_rawnet
Embedding size is 256, encoder SAP.
Initialised AAMSoftmax margin 0.100 scale 30.000
Initialised Adam optimizer
Initialised step LR scheduler
torch.Size([100, 2, 256]) 256
----------------------
Traceback (most recent call last):
  File "./trainSpeakerNet.py", line 304, in <module>
    main()
  File "./trainSpeakerNet.py", line 300, in main
    main_worker(0, None, args)
  File "./trainSpeakerNet.py", line 238, in main_worker
    loss, traineer = trainer.train_network(train_loader, verbose=(args.gpu == 0));
  File "/triplet_sre/voxceleb_trainer/SpeakerNet.py", line 111, in train_network
    nloss, prec1 = self.__model__(data, label)
  File "/opt/conda/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
    return forward_call(*input, **kwargs)
  File "/triplet_sre/voxceleb_trainer/SpeakerNet.py", line 23, in forward
    return self.module(x, label)
  File "/opt/conda/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
    return forward_call(*input, **kwargs)
  File "/triplet_sre/voxceleb_trainer/SpeakerNet.py", line 51, in forward
    nloss, prec1 = self.__L__.forward(outp,label)
  File "/triplet_sre/voxceleb_trainer/loss/aamsoftmax.py", line 39, in forward
    assert x.size()[1] == self.in_feats
AssertionError

It seems, x has this shape torch.Size([100, 2, 256]), now x.size()[1] is 2 and not 256, so it throws this error. How can I fix this error?

[Jungjee]

Hi @zabir-nabil.
May I first ask for the configuration you used?

• I'm suspecting that you may have set nPerSpeaker to 2 whereas you should set to 1 for aamsoftmax