AttributeError: module 'config' has no attribute 'org_size'

Hi,this problem arises when I run attack_imagenet.py,how to solve it?

Hi,

Thanks for your interest in our work.
According to the code, it seems like you haven't specified the dataset = 'ImageNet'.
See these lines:

AdvFlow/config.py

Lines 45 to 59 in e679d87

    
           elif dataset == 'ImageNet': 
        
               high_res_blocks = 4     # Number of high-resolution, convolutional blocks 
        
               low_res_blocks  = 6     # Number of low-resolution, convolutional blocks 
        
               channels_hidden = 128   # Number of hidden channels for the convolutional blocks 
        
               batch_norm      = False # Batch normalization? 
        
               n_blocks        = 6     # Number of fully-connected blocks 
        
               internal_width  = 128   # Internal width of the FC blocks 
        
               fc_dropout      = 0.0   # Dropout for FC blocks 
        
               clamping        = 1.5   # Clamping parameter for avoiding exploding exponentiation 
        
               num_classes     = 1000  # Number of classes 
        
               org_size        = 299   # Image-size to re-shape the ImageNet data

	elif dataset == 'ImageNet':

	high_res_blocks = 4 # Number of high-resolution, convolutional blocks
	low_res_blocks = 6 # Number of low-resolution, convolutional blocks
	channels_hidden = 128 # Number of hidden channels for the convolutional blocks
	batch_norm = False # Batch normalization?

	n_blocks = 6 # Number of fully-connected blocks
	internal_width = 128 # Internal width of the FC blocks
	fc_dropout = 0.0 # Dropout for FC blocks

	clamping = 1.5 # Clamping parameter for avoiding exploding exponentiation

	num_classes = 1000 # Number of classes
	org_size = 299 # Image-size to re-shape the ImageNet data