Ken1256

I find out if `torch.nn.L1Loss(reduction='mean')` AdaBound work fine, But `torch.nn.L1Loss(reduction='sum')` Nan loss. (Sorry, after double check the code I change reduction='mean' to reduction='sum', But Adam both work fine. Normally 'mean'...

I try `optimizer = adabound.AdaBound(model.parameters(), lr=1e-4, final_lr=1e-4)` still Nan loss.

Why enc_blk_nums[3] "receptive field" better than middle_blk_num? How does the deblur + denoise task allocate 36 blocks?

I don't know how to get inputShape value, just guess...

Also https://github.com/developer0hye/Simple-PyTorch-Deformable-Convolution-v2/blob/main/experiment.ipynb

``` from vapoursynth import core import vapoursynth as vs pp = VpsFilterSource matrix = "709" # pp = core.std.SetFrameProp(pp, prop="_Matrix", intval=2) # pp = core.std.SetFrameProp(pp, prop="_Transfer", intval=2) # pp =...

get_flat_mask l_img = (0.2989 * r + 0.587 * g + 0.114 * b).unsqueeze(dim=1) 0.2989 0.587 0.114 where the RGB number come from?