Autoencoder Error- Help Needed

1

I build an autoencoder similar to VGG 19. But my decoder is throwing some error. Using it on CIFAR10 dataset

mode WRN(
  (encoder): Sequential(
    (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU(inplace=True)
    (3): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (4): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (5): ReLU(inplace=True)
    (6): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (7): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (8): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (9): ReLU(inplace=True)
    (10): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (11): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (12): ReLU(inplace=True)
    (13): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (14): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (15): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (16): ReLU(inplace=True)
    (17): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (18): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (19): ReLU(inplace=True)
    (20): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (21): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (22): ReLU(inplace=True)
    (23): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (24): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (25): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (26): ReLU(inplace=True)
    (27): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (28): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (29): ReLU(inplace=True)
    (30): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (31): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (32): ReLU(inplace=True)
    (33): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (34): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (35): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (36): ReLU(inplace=True)
    (37): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (38): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (39): ReLU(inplace=True)
    (40): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (41): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (42): ReLU(inplace=True)
    (43): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (latent_mu): Linear(in_features=512, out_features=60, bias=False)
  (latent_std): Linear(in_features=512, out_features=60, bias=False)
  (classifier): Sequential(
    (0): Linear(in_features=60, out_features=10, bias=False)
  )
  (latent_decoder): Linear(in_features=60, out_features=512, bias=False)
  (decoder): Sequential(
    (0): ConvTranspose2d(512, 320, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
    (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (2): ReLU(inplace=True)
    (3): ConvTranspose2d(320, 320, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
    (4): BatchNorm2d(320, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (5): ReLU(inplace=True)
    (6): ConvTranspose2d(320, 320, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
    (7): BatchNorm2d(320, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (8): ReLU(inplace=True)
    (9): ConvTranspose2d(320, 320, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
    (10): BatchNorm2d(320, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (11): ReLU(inplace=True)
    (12): ConvTranspose2d(320, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
    (13): BatchNorm2d(320, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (14): ReLU(inplace=True)
    (15): ConvTranspose2d(160, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
    (16): BatchNorm2d(160, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (17): ReLU(inplace=True)
    (18): ConvTranspose2d(160, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
    (19): BatchNorm2d(160, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (20): ReLU(inplace=True)
    (21): ConvTranspose2d(160, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
    (22): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (23): ReLU(inplace=True)
    (24): ConvTranspose2d(16, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
    (25): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (26): ReLU(inplace=True)
    (27): ConvTranspose2d(16, 16, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
    (28): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (29): ReLU(inplace=True)
    (30): ConvTranspose2d(16, 3, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
  )
)
Traceback (most recent call last):
  File "newmain.py", line 286, in <module>
    main(args)
  File "newmain.py", line 247, in main
    iterations)
  File "/content/drive/My Drive/Andromeda/thirdparty/openset/ActiveLearning/lib/Training/train.py", line 312, in train_var_joint
    class_samples, recon_samples, mu, std = model(inp)
  File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 550, in __call__
    result = self.forward(*input, **kwargs)
  File "/content/drive/My Drive/Andromeda/thirdparty/openset/ActiveLearning/model.py", line 797, in forward
    output_samples[i] = self.decode(z)
  File "/content/drive/My Drive/Andromeda/thirdparty/openset/ActiveLearning/model.py", line 761, in decode
    x = self.decoder(z)
  File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 550, in __call__
    result = self.forward(*input, **kwargs)
  File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/container.py", line 100, in forward
    input = module(input)
  File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 550, in __call__
    result = self.forward(*input, **kwargs)
  File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/batchnorm.py", line 106, in forward
    exponential_average_factor, self.eps)
  File "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py", line 1923, in batch_norm
    training, momentum, eps, torch.backends.cudnn.enabled
RuntimeError: running_mean should contain 320 elements not 512
2

It looks like this part of the model is wrong:

  (decoder): Sequential(
    (0): ConvTranspose2d(512, 320, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
    (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)

Since the transposed convolution will output 320 channels, while the following batchnorm layer expects 512 input channels.

3

Thank you. I realized it afterwards.

Have one more question for you.(your opinion is good enough). What does AvergaePool does according to it and it is really useful? I am getting better results of my VGG CIFAR 10 classifier without that. Is that okay?

 self.avgpool = nn.AdaptiveAvgPool2d((8, 8))
4

Adaptive pooling layers take a specific output shape and make sure to pool the input to this shape.
This is useful, if your input tensors have a variable spatial shape.
Average pooling layers work similar to max pooling, but calculate the average of the window instead of the max value.

If it’s not helping, you could of course remove it. Since the output shape is defined in this layer, you might need to use F.interpolate or another method to create the same output shape.

5

Sure… thank you for your input