warmup learning rate should be float?

Current:

    parser.add_argument("--warmup-learning-rate", type=int, default=1e-6,
                        help="learning rate for warmup")

You're right.
Also warmup option has a problem that only applies the first 1 epoch.

I can't touch my pc now. I will check when I get home .（next week）

You're right. Also warmup option has a problem that only applies the first 1 epoch.

I can't touch my pc now. I will check when I get home .（next week）

I read your code of Waifu2xDataset. For the __getitem__, at the end there's a line TF.pad(y, [-self.model_offset] * 4). Could you explain a little bit? What is model_offset?

What is model_offset?

model_offset is the unpad size of the model output.

Historically, waifu2x models do not use zero padding for conv2d. This is because of the problem of visible seams in tiled rendering. ( releated to nagadomi/waifu2x#238 , The quoted comment was written by me. )

So the output size of the model will be smaller than 2x the input.
For example, UpConv7.i2i_scale = 2, UpConv7.i2i_offset = 14

nunif/waifu2x/models/upconv_7.py

Lines 7 to 11 in 2992576

    
           class UpConv7(I2IBaseModel): 
        
               name = "waifu2x.upconv_7" 
        
               def __init__(self, in_channels=3, out_channels=3): 
        
                   super(UpConv7, self).__init__(locals(), scale=2, offset=14, in_channels=in_channels)

input x = 256x256

nunif/waifu2x/models/upconv_7.py

Lines 41 to 51 in 2992576

    
           if __name__ == "__main__": 
        
               import torch 
        
               from nunif.models import create_model 
        
               device = "cuda:0" 
        
               model = create_model(UpConv7.name, in_channels=3, out_channels=3).to(device) 
        
               print(model) 
        
               x = torch.zeros((1, 3, 256, 256)).to(device) 
        
               with torch.no_grad(): 
        
                   z = model(x) 
        
                   print(z.shape) 
        
                   print(model.name, model.i2i_offset, model.i2i_scale)

torch.Size([1, 3, 484, 484])

output size is 484x484.

output_size = input_size * scale - offset * 2 ( 484 = 256 * 2 - 14 * 2)

Fixed 5b30b03

Also warmup option has a problem that only applies the first 1 epoch.

This problem has not yet fixed.

	class UpConv7(I2IBaseModel):
	name = "waifu2x.upconv_7"

	def __init__(self, in_channels=3, out_channels=3):
	super(UpConv7, self).__init__(locals(), scale=2, offset=14, in_channels=in_channels)

	if __name__ == "__main__":
	import torch
	from nunif.models import create_model
	device = "cuda:0"
	model = create_model(UpConv7.name, in_channels=3, out_channels=3).to(device)
	print(model)
	x = torch.zeros((1, 3, 256, 256)).to(device)
	with torch.no_grad():
	z = model(x)
	print(z.shape)
	print(model.name, model.i2i_offset, model.i2i_scale)