Deprecated modules

[nshervt]

I tried using kymatio with this code. If you were interested in updating it, I switched the scattering method to kymatio but there are more things to be fixed, for instance, in STL, it seems that torch.legacy that you are using in utils has been deprecated and needs to be changed. In imagenet test, line torch.randn(8).cuda() again uses cuda, and when I comment .cuda() out, I fall into a rabbit hole of errors. the latest being

Traceback (most recent call last):
  File "/Users/Nshervt/Desktop/scalingscattering-master/imagenet/main_test.py", line 186, in <module>
    main()
  File "/Users/Nshervt/Desktop/scalingscattering-master/imagenet/main_test.py", line 65, in main
    model, params, stats = models.__dict__[opt.model](N=opt.N,J=opt.scat)
  File "/Users/Nshervt/Desktop/scalingscattering-master/imagenet/models/scatter_resnet.py", line 151, in scatresnet6_2
    model = ScatResNet(J,N)
  File "/Users/Nshervt/Desktop/scalingscattering-master/imagenet/models/scatter_resnet.py", line 61, in __init__
    self.bn0 = nn.BatchNorm2d(3*self.nfscat,eps=1e-5, momentum=0.9, affine=False)
  File "/Users/Nshervt/anaconda3/lib/python3.7/site-packages/torch/nn/modules/batchnorm.py", line 34, in __init__
    self.register_buffer('running_mean', torch.zeros(num_features))
TypeError: zeros(): argument 'size' (position 1) must be tuple of ints, not float

[edouardoyallon]

Here, the issue is easy, simply replace 3*self.nfscat by 3*int(self.nfscat)

Furthermore, the STL script is broken because we used old-school scripts (engine, opencv..)... I'm not really happy that this work isn't reproducible with the new version of torch, thus I'd be happy to help to reimplment it!

[nshervt]

That would be great, Thanks! There are still errors in imagenet, I'll do a PR when I got it to work, or let you know if there were issues that I couldn't solve.

[eugenium]

btw for imagenet (esp if one is interested to train the models) this repo might also be useful and was done with newer versions of pytorch: https://github.com/edouardoyallon/pyscatlight. Its using only 1st order though but that can be easily adapted

[nshervt]

Thanks @eugenium , I just tried running this, but this is also written for pytorch0.4 and doesn't work with newer versions. I opened an issue with specific problem if you were interested.