Repreduce the results of PRECODE

[chenxiang3luo]

Thank you for your outstanding work. I want to recover the data under PRECODE defense with ROG. But when I change the LeNet to LeNetVB like the model in PRECODE, the code always reports bugs that RuntimeError: size mismatch, got input (16), mat (16x256), vec (512)

Traceback (most recent call last):
  File "/data2/chenxiang/svd_defense/rog/attack_fedavg.py", line 7, in <module>
    main("config_fedavg.yaml",int(sys.argv[1]))
  File "/data2/chenxiang/svd_defense/rog/main.py", line 87, in main
    recon_data = grad_inv(attacker, grad, x, onehot, model, config, logger)
  File "/data2/chenxiang/svd_defense/rog/src/attack/attacker.py", line 269, in grad_inv
    dummy_data = attacker.multistep_attack(grad, x, onehot, model, config.tau, logger)
  File "/data2/chenxiang/svd_defense/rog/src/attack/attacker.py", line 235, in multistep_attack
    dummy_data_optimizer.step(closure)
  File "/home/chenxiang/anaconda3/envs/svd/lib/python3.10/site-packages/torch/optim/lr_scheduler.py", line 75, in wrapper
    return wrapped(*args, **kwargs)
  File "/home/chenxiang/anaconda3/envs/svd/lib/python3.10/site-packages/torch/optim/optimizer.py", line 385, in wrapper
    out = func(*args, **kwargs)
  File "/home/chenxiang/anaconda3/envs/svd/lib/python3.10/site-packages/torch/optim/optimizer.py", line 76, in _use_grad
    ret = func(self, *args, **kwargs)
  File "/home/chenxiang/anaconda3/envs/svd/lib/python3.10/site-packages/torch/optim/adam.py", line 146, in step
    loss = closure()
  File "/data2/chenxiang/svd_defense/rog/src/attack/attacker.py", line 209, in closure
    pred = meta_net(F.interpolate(dummy_data, scale_factor=self.sf, mode='bicubic'))
  File "/home/chenxiang/anaconda3/envs/svd/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1511, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
  File "/home/chenxiang/anaconda3/envs/svd/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1520, in _call_impl
    return forward_call(*args, **kwargs)
  File "/data2/chenxiang/svd_defense/rog/networks/metanet.py", line 83, in forward
    output = self.net(inputs)
  File "/home/chenxiang/anaconda3/envs/svd/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1511, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
  File "/home/chenxiang/anaconda3/envs/svd/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1520, in _call_impl
    return forward_call(*args, **kwargs)
  File "/data2/chenxiang/svd_defense/rog/networks/vgg.py", line 189, in forward
    x = self.VB(x)
  File "/home/chenxiang/anaconda3/envs/svd/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1511, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
  File "/home/chenxiang/anaconda3/envs/svd/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1520, in _call_impl
    return forward_call(*args, **kwargs)
  File "/data2/chenxiang/svd_defense/rog/networks/precodenet/VariationalBlock.py", line 59, in forward
    x = self.decoder(encoding.unsqueeze(0))
  File "/home/chenxiang/anaconda3/envs/svd/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1511, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
  File "/home/chenxiang/anaconda3/envs/svd/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1520, in _call_impl
    return forward_call(*args, **kwargs)
RuntimeError: size mismatch, got input (16), mat (16x256), vec (512)

I have checked the tensor size and layer size. By the way, the vgg7_vb in your code is vgg7, so there is no example I could learn.