How can I customize the model arch crnn.cfg about crnn-audio-classification HOT 1 CLOSED

@takeofuture so some of the details of how the cfg file works will be in https://github.com/ksanjeevan/torchparse. Basically, you can do something like:

  [conv2d]
      out_channels=16
      kernel_size=3
      stride=1
      padding=valid
  [batchnorm2d]
  [elu]
  [maxpool2d]
      kernel_size=3
      stride=3
  [dropout]
      p=0.1
[conv2d]
      out_channels=16
      kernel_size=3
      stride=1
      padding=valid
  [batchnorm2d]
  [elu]
  [maxpool2d]
      kernel_size=3
      stride=3
  [dropout]
      p=0.1
...

And keep adding convolutions. At no point do you have to pre-define an input size as depending on the model you've defined, multiple input dimensions can be valid.

If using torchparse is confusing, or you'd rather define your model some other way, you can just write your own in net/, or modify

from crnn-audio-classification.