hvision-nku / srformer Goto Github PK

Official code for "SRFormer: Permuted Self-Attention for Single Image Super-Resolution" (ICCV 2023)

Home Page: https://openaccess.thecvf.com/content/ICCV2023/papers/Zhou_SRFormer_Permuted_Self-Attention_for_Single_Image_Super-Resolution_ICCV_2023_paper.pdf

License: Other

Python 85.10% C++ 5.61% Cuda 8.35% Shell 0.11% MATLAB 0.82%

srformer's People

Contributors

Stargazers

Watchers

Forkers

makex1n wwhio feynman1999 ip-restoration ip-superresolution wangzhi723 jinglongdu a-biao96 queahren yanlin2001 icoicqico peterzs anhuipl2010 bengissu code-wangshuyi aakashapoorv strategist922 hamzaaylak

srformer's Issues

Figure 4

In the link you provided for calculating the power spectrum, the final image is not a smooth curve, what did you do to turn it into a figure as the Figure 4.

PretrainModel folder

Sorry to bother, but I couldn't find the PretrainModel folder mentioned in the. yml file

Figure 4

Hi author, I am very interested in your visualization of Figure 4, how did you draw this figure? Thank you so much for teaching me this.

How to draw the power spectrum of the intermediate feature map？

Hello, .I would like to ask how to draw the power spectrum of the intermediate feature mapping as shown in Figure 4 in the paper.

KeyError: "No object named 'SwinIR_bw8mlp5' found in 'arch' registry!"

有遇到这个问题的吗？

Training configs for real world SR

Could you please provide the training configs for the real world sr setting? Thanks !

Error when trying to load pretrained model

Hi;

I got the fooling error when trying to load the pretrained model:
net.load_state_dict(torch.load( '/tmp/SRFormerLight_SRx2_DIV2K.pth'))

RuntimeError                              Traceback (most recent call last)
Input In [39], in <cell line: 1>()
----> 1 net.load_state_dict(torch.load( '/tmp/SRFormerLight_SRx2_DIV2K.pth'))#  '/tmp/SRFormer_SRx2_DF2K.pth'))
      2 #net= torch.load('/tmp/SRFormer_SRx2_DF2K.pth')
      4 img_size= 48

File /home/software/python/AnacondaPython3.9/lib/python3.9/site-packages/torch/nn/modules/module.py:1497, in Module.load_state_dict(self, state_dict, strict)
   1492         error_msgs.insert(
   1493             0, 'Missing key(s) in state_dict: {}. '.format(
   1494                 ', '.join('"{}"'.format(k) for k in missing_keys)))
   1496 if len(error_msgs) > 0:
-> 1497     raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
   1498                        self.__class__.__name__, "\n\t".join(error_msgs)))
   1499 return _IncompatibleKeys(missing_keys, unexpected_keys)

RuntimeError: Error(s) in loading state_dict for SRFormer:
	Missing key(s) in state_dict: "conv_first.weight", "conv_first.bias", "patch_embed.norm.weight", "patch_embed.norm.bias", "layers.0.residual_group.blocks.0.norm1.weight", "layers.0.residual_group.blocks.0.norm1.bias", "layers.0.residual_group.blocks.0.attn.relative_position_bias_table", "layers.0.residual_group.blocks.0.attn.aligned_relative_position_index", "layers.0.residual_group.blocks.0.attn.kv.weight", "layers.0.residual_group.blocks.0.attn.kv.bias", "layers.0.residual_group.blocks.0.attn.q.weight", "layers.0.residual_group.blocks.0.attn.q.bias", "layers.0.residual_group.blocks.0.attn.proj.weight", "layers.0.residual_group.blocks.0.attn.proj.bias", "layers.0.residual_group.blocks.0.norm2.weight", "layers.0.residual_group.blocks.0.norm2.bias", "layers.0.residual_group.blocks.0.mlp.fc1.weight", "layers.0.residual_group.blocks.0.mlp.fc1.bias", "layers.0.residual_group.blocks.0.mlp.dwconv.depthwise_conv.0.weight", "layers.0.residual_group.blocks.0.mlp.dwconv.depthwise_conv.0.bias", "layers.0.residual_group.blocks.0.mlp.fc2.weight", "layers.0.residual_group.blocks.0.mlp.fc2.bias", "layers.0.residual_group.blocks.1.attn_mask", "layers.0.residual_group.blocks.1.norm1.weight", "layers.0.residual_group.blocks.1.norm1.bias", "layers.0.residual_group.blocks.1.attn.relative_position_bias_table", "layers.0.residual_group.blocks.1.attn.aligned_relative_position_index", "layers.0.residual_group.blocks.1.attn.kv.weight", "layers.0.residual_group.blocks.1.attn.kv.bias", "layers.0.residual_group.blocks.1.attn.q.weight", "layers.0.residual_group.blocks.1.attn.q.bias", "layers.0.residual_group.blocks.1.attn.proj.weight", "layers.0.residual_group.blocks.1.attn.proj.bias", "layers.0.residual_group.blocks.1.norm2.weight", "layers.0.residual_group.blocks.1.norm2.bias", "layers.0.residual_group.blocks.1.mlp.fc1.weight", "layers.0.residual_group.blocks.1.mlp.fc1.bias", "layers.0.residual_group.blocks.1.mlp.dwconv.depthwise_conv.0.weight", "layers.0.residual_group.blocks.1.mlp.dwconv.depthwise_conv.0.bias", "layers.0.residual_group.blocks.1.mlp.fc2.weight", "layers.0.residual_group.blocks.1.mlp.fc2.bias", "layers.0.residual_group.blocks.2.norm1.weight", "layers.0.residual_group.blocks.2.norm1.bias", "layers.0.residual_group.blocks.2.attn.relative_position_bias_table", "layers.0.residual_group.blocks.2.attn.aligned_relative_position_index", "layers.0.residual_group.blocks.2.attn.kv.weight", "layers.0.residual_group.blocks.2.attn.kv.bias", "layers.0.residual_group.blocks.2.attn.q.weight", "layers.0.residual_group.blocks.2.attn.q.bias", "layers.0.residual_group.blocks.2.attn.proj.weight", "layers.0.residual_group.blocks.2.attn.proj.bias", "layers.0.residual_group.blocks.2.norm2.weight", "layers.0.residual_group.blocks.2.norm2.bias", "layers.0.residual_group.blocks.2.mlp.fc1.weight", "layers.0.residual_group.blocks.2.mlp.fc1.bias", "layers.0.residual_group.blocks.2.mlp.dwconv.depthwise_conv.0.weight", "layers.0.residual_group.blocks.2.mlp.dwconv.depthwise_conv.0.bias", "layers.0.residual_group.blocks.2.mlp.fc2.weight", "layers.0.residual_group.blocks.2.mlp.fc2.bias", "layers.0.residual_group.blocks.3.attn_mask", "layers.0.residual_group.blocks.3.norm1.weight", "layers.0.residual_group.blocks.3.norm1.bias", "layers.0.residual_group.blocks.3.attn.relative_position_bias_table", "layers.0.residual_group.blocks.3.attn.aligned_relative_position_index", "layers.0.residual_group.blocks.3.attn.kv.weight", "layers.0.residual_group.blocks.3.attn.kv.bias", "layers.0.residual_group.blocks.3.attn.q.weight", "layers.0.residual_group.blocks.3.attn.q.bias", "layers.0.residual_group.blocks.3.attn.proj.weight", "layers.0.residual_group.blocks.3.attn.proj.bias", "layers.0.residual_group.blocks.3.norm2.weight", "layers.0.residual_group.blocks.3.norm2.bias", "layers.0.residual_group.blocks.3.mlp.fc1.weight", "layers.0.residual_group.blocks.3.mlp.fc1.bias", "layers.0.residual_group.blocks.3.mlp.dwconv.depthwise_conv.0.weight", "layers.0.residual_group.blocks.3.mlp.dwconv.depthwise_conv.0.bias", "layers.0.residual_group.blocks.3.mlp.fc2.weight", "layers.0.residual_group.blocks.3.mlp.fc2.bias", "layers.0.residual_group.blocks.4.norm1.weight", "layers.0.residual_group.blocks.4.norm1.bias", "layers.0.residual_group.blocks.4.attn.relative_position_bias_table", "layers.0.residual_group.blocks.4.attn.aligned_relative_position_index", "layers.0.residual_group.blocks.4.attn.kv.weight", "layers.0.residual_group.blocks.4.attn.kv.bias", "layers.0.residual_group.blocks.4.attn.q.weight", "layers.0.residual_group.blocks.4.attn.q.bias", "layers.0.residual_group.blocks.4.attn.proj.weight", "layers.0.residual_group.blocks.4.attn.proj.bias", "layers.0.residual_group.blocks.4.norm2.weight", "layers.0.residual_group.blocks.4.norm2.bias", "layers.0.residual_group.blocks.4.mlp.fc1.weight", "layers.0.residual_group.blocks.4.mlp.fc1.bias", "layers.0.residual_group.blocks.4.mlp.dwconv.depthwise_conv.0.weight", "layers.0.residual_group.blocks.4.mlp.dwconv.depthwise_conv.0.bias", "layers.0.residual_group.blocks.4.mlp.fc2.weight", "layers.0.residual_group.blocks.4.mlp.fc2.bias", "layers.0.residual_group.blocks.5.attn_mask", "layers.0.residual_group.blocks.5.norm1.weight", "layers.0.residual_group.blocks.5.norm1.bias", "layers.0.residual_group.blocks.5.attn.relative_position_bias_table", "layers.0.residual_group.blocks.5.attn.aligned_relative_position_index", "layers.0.residual_group.blocks.5.attn.kv.weight", "layers.0.residual_group.blocks.5.attn.kv.bias", "layers.0.residual_group.blocks.5.attn.q.weight", "layers.0.residual_group.blocks.5.attn.q.bias", "layers.0.residual_group.blocks.5.attn.proj.weight", "layers.0.residual_group.blocks.5.attn.proj.bias", "layers.0.residual_group.blocks.5.norm2.weight", "layers.0.residual_group.blocks.5.norm2.bias", "layers.0.residual_group.blocks.5.mlp.fc1.weight", "layers.0.residual_group.blocks.5.mlp.fc1.bias", "layers.0.residual_group.blocks.5.mlp.dwconv.depthwise_conv.0.weight", "layers.0.residual_group.blocks.5.mlp.dwconv.depthwise_conv.0.bias", "layers.0.residual_group.blocks.5.mlp.fc2.weight", "layers.0.residual_group.blocks.5.mlp.fc2.bias", "layers.0.conv.weight", "layers.0.conv.bias", "layers.1.residual_group.blocks.0.norm1.weight", "layers.1.residual_group.blocks.0.norm1.bias", "layers.1.residual_group.blocks.0.attn.relative_position_bias_table", "layers.1.residual_group.blocks.0.attn.aligned_relative_position_index", "layers.1.residual_group.blocks.0.attn.kv.weight", "layers.1.residual_group.blocks.0.attn.kv.bias", "layers.1.residual_group.blocks.0.attn.q.weight", "layers.1.residual_group.blocks.0.attn.q.bias", "layers.1.residual_group.blocks.0.attn.proj.weight", "layers.1.residual_group.blocks.0.attn.proj.bias", "layers.1.residual_group.blocks.0.norm2.weight", "layers.1.residual_group.blocks.0.norm2.bias", "layers.1.residual_group.blocks.0.mlp.fc1.weight", "layers.1.residual_group.blocks.0.mlp.fc1.bias", "layers.1.residual_group.blocks.0.mlp.dwconv.depthwise_conv.0.weight", "layers.1.residual_group.blocks.0.mlp.dwconv.depthwise_conv.0.bias", "layers.1.residual_group.blocks.0.mlp.fc2.weight", "layers.1.residual_group.blocks.0.mlp.fc2.bias", "layers.1.residual_group.blocks.1.attn_mask", "layers.1.residual_group.blocks.1.norm1.weight", "layers.1.residual_group.blocks.1.norm1.bias", "layers.1.residual_group.blocks.1.attn.relative_position_bias_table", "layers.1.residual_group.blocks.1.attn.aligned_relative_position_index", "layers.1.residual_group.blocks.1.attn.kv.weight", "layers.1.residual_group.blocks.1.attn.kv.bias", "layers.1.residual_group.blocks.1.attn.q.weight", "layers.1.residual_group.blocks.1.attn.q.bias", "layers.1.residual_group.blocks.1.attn.proj.weight", "layers.1.residual_group.blocks.1.attn.proj.bias", "layers.1.residual_group.blocks.1.norm2.weight", "layers.1.residual_group.blocks.1.norm2.bias", "layers.1.residual_group.blocks.1.mlp.fc1.weight", "layers.1.residual_group.blocks.1.mlp.fc1.bias", "layers.1.residual_group.blocks.1.mlp.dwconv.depthwise_conv.0.weight", "layers.1.residual_group.blocks.1.mlp.dwconv.depthwise_conv.0.bias", "layers.1.residual_group.blocks.1.mlp.fc2.weight", "layers.1.residual_group.blocks.1.mlp.fc2.bias", "layers.1.residual_group.blocks.2.norm1.weight", "layers.1.residual_group.blocks.2.norm1.bias", "layers.1.residual_group.blocks.2.attn.relative_position_bias_table", "layers.1.residual_group.blocks.2.attn.aligned_relative_position_index", "layers.1.residual_group.blocks.2.attn.kv.weight", "layers.1.residual_group.blocks.2.attn.kv.bias", "layers.1.residual_group.blocks.2.attn.q.weight", "layers.1.residual_group.blocks.2.attn.q.bias", "layers.1.residual_group.blocks.2.attn.proj.weight", "layers.1.residual_group.blocks.2.attn.proj.bias", "layers.1.residual_group.blocks.2.norm2.weight", "layers.1.residual_group.blocks.2.norm2.bias", "layers.1.residual_group.blocks.2.mlp.fc1.weight", "layers.1.residual_group.blocks.2.mlp.fc1.bias", "layers.1.residual_group.blocks.2.mlp.dwconv.depthwise_conv.0.weight", "layers.1.residual_group.blocks.2.mlp.dwconv.depthwise_conv.0.bias", "layers.1.residual_group.blocks.2.mlp.fc2.weight", "layers.1.residual_group.blocks.2.mlp.fc2.bias", "layers.1.residual_group.blocks.3.attn_mask", "layers.1.residual_group.blocks.3.norm1.weight", "layers.1.residual_group.blocks.3.norm1.bias", "layers.1.residual_group.blocks.3.attn.relative_position_bias_table", "layers.1.residual_group.blocks.3.attn.aligned_relative_position_index", "layers.1.residual_group.blocks.3.attn.kv.weight", "layers.1.residual_group.blocks.3.attn.kv.bias", "layers.1.residual_group.blocks.3.attn.q.weight", "layers.1.residual_group.blocks.3.attn.q.bias", "layers.1.residual_group.blocks.3.attn.proj.weight", "layers.1.residual_group.blocks.3.attn.proj.bias", "layers.1.residual_group.blocks.3.norm2.weight", "layers.1.residual_group.blocks.3.norm2.bias", "layers.1.residual_group.blocks.3.mlp.fc1.weight", "layers.1.residual_group.blocks.3.mlp.fc1.bias", "layers.1.residual_group.blocks.3.mlp.dwconv.depthwise_conv.0.weight", "layers.1.residual_group.blocks.3.mlp.dwconv.depthwise_conv.0.bias", "layers.1.residual_group.blocks.3.mlp.fc2.weight", "layers.1.residual_group.blocks.3.mlp.fc2.bias", "layers.1.residual_group.blocks.4.norm1.weight", "layers.1.residual_group.blocks.4.norm1.bias", "layers.1.residual_group.blocks.4.attn.relative_position_bias_table", "layers.1.residual_group.blocks.4.attn.aligned_relative_position_index", "layers.1.residual_group.blocks.4.attn.kv.weight", "layers.1.residual_group.blocks.4.attn.kv.bias", "layers.1.residual_group.blocks.4.attn.q.weight", "layers.1.residual_group.blocks.4.attn.q.bias", "layers.1.residual_group.blocks.4.attn.proj.weight", "layers.1.residual_group.blocks.4.attn.proj.bias", "layers.1.residual_group.blocks.4.norm2.weight", "layers.1.residual_group.blocks.4.norm2.bias", "layers.1.residual_group.blocks.4.mlp.fc1.weight", "layers.1.residual_group.blocks.4.mlp.fc1.bias", "layers.1.residual_group.blocks.4.mlp.dwconv.depthwise_conv.0.weight", "layers.1.residual_group.blocks.4.mlp.dwconv.depthwise_conv.0.bias", "layers.1.residual_group.blocks.4.mlp.fc2.weight", "layers.1.residual_group.blocks.4.mlp.fc2.bias", "layers.1.residual_group.blocks.5.attn_mask", "layers.1.residual_group.blocks.5.norm1.weight", "layers.1.residual_group.blocks.5.norm1.bias", "layers.1.residual_group.blocks.5.attn.relative_position_bias_table", "layers.1.residual_group.blocks.5.attn.aligned_relative_position_index", "layers.1.residual_group.blocks.5.attn.kv.weight", "layers.1.residual_group.blocks.5.attn.kv.bias", "layers.1.residual_group.blocks.5.attn.q.weight", "layers.1.residual_group.blocks.5.attn.q.bias", "layers.1.residual_group.blocks.5.attn.proj.weight", "layers.1.residual_group.blocks.5.attn.proj.bias", "layers.1.residual_group.blocks.5.norm2.weight", "layers.1.residual_group.blocks.5.norm2.bias", "layers.1.residual_group.blocks.5.mlp.fc1.weight", "layers.1.residual_group.blocks.5.mlp.fc1.bias", "layers.1.residual_group.blocks.5.mlp.dwconv.depthwise_conv.0.weight", "layers.1.residual_group.blocks.5.mlp.dwconv.depthwise_conv.0.bias", "layers.1.residual_group.blocks.5.mlp.fc2.weight", "layers.1.residual_group.blocks.5.mlp.fc2.bias", "layers.1.conv.weight", "layers.1.conv.bias", "layers.2.residual_group.blocks.0.norm1.weight", "layers.2.residual_group.blocks.0.norm1.bias", "layers.2.residual_group.blocks.0.attn.relative_position_bias_table", "layers.2.residual_group.blocks.0.attn.aligned_relative_position_index", "layers.2.residual_group.blocks.0.attn.kv.weight", "layers.2.residual_group.blocks.0.attn.kv.bias", "layers.2.residual_group.blocks.0.attn.q.weight", "layers.2.residual_group.blocks.0.attn.q.bias", "layers.2.residual_group.blocks.0.attn.proj.weight", "layers.2.residual_group.blocks.0.attn.proj.bias", "layers.2.residual_group.blocks.0.norm2.weight", "layers.2.residual_group.blocks.0.norm2.bias", "layers.2.residual_group.blocks.0.mlp.fc1.weight", "layers.2.residual_group.blocks.0.mlp.fc1.bias", "layers.2.residual_group.blocks.0.mlp.dwconv.depthwise_conv.0.weight", "layers.2.residual_group.blocks.0.mlp.dwconv.depthwise_conv.0.bias", "layers.2.residual_group.blocks.0.mlp.fc2.weight", "layers.2.residual_group.blocks.0.mlp.fc2.bias", "layers.2.residual_group.blocks.1.attn_mask", "layers.2.residual_group.blocks.1.norm1.weight", "layers.2.residual_group.blocks.1.norm1.bias", "layers.2.residual_group.blocks.1.attn.relative_position_bias_table", "layers.2.residual_group.blocks.1.attn.aligned_relative_position_index", "layers.2.residual_group.blocks.1.attn.kv.weight", "layers.2.residual_group.blocks.1.attn.kv.bias", "layers.2.residual_group.blocks.1.attn.q.weight", "layers.2.residual_group.blocks.1.attn.q.bias", "layers.2.residual_group.blocks.1.attn.proj.weight", "layers.2.residual_group.blocks.1.attn.proj.bias", "layers.2.residual_group.blocks.1.norm2.weight", "layers.2.residual_group.blocks.1.norm2.bias", "layers.2.residual_group.blocks.1.mlp.fc1.weight", "layers.2.residual_group.blocks.1.mlp.fc1.bias", "layers.2.residual_group.blocks.1.mlp.dwconv.depthwise_conv.0.weight", "layers.2.residual_group.blocks.1.mlp.dwconv.depthwise_conv.0.bias", "layers.2.residual_group.blocks.1.mlp.fc2.weight", "layers.2.residual_group.blocks.1.mlp.fc2.bias", "layers.2.residual_group.blocks.2.norm1.weight", "layers.2.residual_group.blocks.2.norm1.bias", "layers.2.residual_group.blocks.2.attn.relative_position_bias_table", "layers.2.residual_group.blocks.2.attn.aligned_relative_position_index", "layers.2.residual_group.blocks.2.attn.kv.weight", "layers.2.residual_group.blocks.2.attn.kv.bias", "layers.2.residual_group.blocks.2.attn.q.weight", "layers.2.residual_group.blocks.2.attn.q.bias", "layers.2.residual_group.blocks.2.attn.proj.weight", "layers.2.residual_group.blocks.2.attn.proj.bias", "layers.2.residual_group.blocks.2.norm2.weight", "layers.2.residual_group.blocks.2.norm2.bias", "layers.2.residual_group.blocks.2.mlp.fc1.weight", "layers.2.residual_group.blocks.2.mlp.fc1.bias", "layers.2.residual_group.blocks.2.mlp.dwconv.depthwise_conv.0.weight", "layers.2.residual_group.blocks.2.mlp.dwconv.depthwise_conv.0.bias", "layers.2.residual_group.blocks.2.mlp.fc2.weight", "layers.2.residual_group.blocks.2.mlp.fc2.bias", "layers.2.residual_group.blocks.3.attn_mask", "layers.2.residual_group.blocks.3.norm1.weight", "layers.2.residual_group.blocks.3.norm1.bias", "layers.2.residual_group.blocks.3.attn.relative_position_bias_table", "layers.2.residual_group.blocks.3.attn.aligned_relative_position_index", "layers.2.residual_group.blocks.3.attn.kv.weight", "layers.2.residual_group.blocks.3.attn.kv.bias", "layers.2.residual_group.blocks.3.attn.q.weight", "layers.2.residual_group.blocks.3.attn.q.bias", "layers.2.residual_group.blocks.3.attn.proj.weight", "layers.2.residual_group.blocks.3.attn.proj.bias", "layers.2.residual_group.blocks.3.norm2.weight", "layers.2.residual_group.blocks.3.norm2.bias", "layers.2.residual_group.blocks.3.mlp.fc1.weight", "layers.2.residual_group.blocks.3.mlp.fc1.bias", "layers.2.residual_group.blocks.3.mlp.dwconv.depthwise_conv.0.weight", "layers.2.residual_group.blocks.3.mlp.dwconv.depthwise_conv.0.bias", "layers.2.residual_group.blocks.3.mlp.fc2.weight", "layers.2.residual_group.blocks.3.mlp.fc2.bias", "layers.2.residual_group.blocks.4.norm1.weight", "layers.2.residual_group.blocks.4.norm1.bias", "layers.2.residual_group.blocks.4.attn.relative_position_bias_table", "layers.2.residual_group.blocks.4.attn.aligned_relative_position_index", "layers.2.residual_group.blocks.4.attn.kv.weight", "layers.2.residual_group.blocks.4.attn.kv.bias", "layers.2.residual_group.blocks.4.attn.q.weight", "layers.2.residual_group.blocks.4.attn.q.bias", "layers.2.residual_group.blocks.4.attn.proj.weight", "layers.2.residual_group.blocks.4.attn.proj.bias", "layers.2.residual_group.blocks.4.norm2.weight", "layers.2.residual_group.blocks.4.norm2.bias", "layers.2.residual_group.blocks.4.mlp.fc1.weight", "layers.2.residual_group.blocks.4.mlp.fc1.bias", "layers.2.residual_group.blocks.4.mlp.dwconv.depthwise_conv.0.weight", "layers.2.residual_group.blocks.4.mlp.dwconv.depthwise_conv.0.bias", "layers.2.residual_group.blocks.4.mlp.fc2.weight", "layers.2.residual_group.blocks.4.mlp.fc2.bias", "layers.2.residual_group.blocks.5.attn_mask", "layers.2.residual_group.blocks.5.norm1.weight", "layers.2.residual_group.blocks.5.norm1.bias", "layers.2.residual_group.blocks.5.attn.relative_position_bias_table", "layers.2.residual_group.blocks.5.attn.aligned_relative_position_index", "layers.2.residual_group.blocks.5.attn.kv.weight", "layers.2.residual_group.blocks.5.attn.kv.bias", "layers.2.residual_group.blocks.5.attn.q.weight", "layers.2.residual_group.blocks.5.attn.q.bias", "layers.2.residual_group.blocks.5.attn.proj.weight", "layers.2.residual_group.blocks.5.attn.proj.bias", "layers.2.residual_group.blocks.5.norm2.weight", "layers.2.residual_group.blocks.5.norm2.bias", "layers.2.residual_group.blocks.5.mlp.fc1.weight", "layers.2.residual_group.blocks.5.mlp.fc1.bias", "layers.2.residual_group.blocks.5.mlp.dwconv.depthwise_conv.0.weight", "layers.2.residual_group.blocks.5.mlp.dwconv.depthwise_conv.0.bias", "layers.2.residual_group.blocks.5.mlp.fc2.weight", "layers.2.residual_group.blocks.5.mlp.fc2.bias", "layers.2.conv.weight", "layers.2.conv.bias", "layers.3.residual_group.blocks.0.norm1.weight", "layers.3.residual_group.blocks.0.norm1.bias", "layers.3.residual_group.blocks.0.attn.relative_position_bias_table", "layers.3.residual_group.blocks.0.attn.aligned_relative_position_index", "layers.3.residual_group.blocks.0.attn.kv.weight", "layers.3.residual_group.blocks.0.attn.kv.bias", "layers.3.residual_group.blocks.0.attn.q.weight", "layers.3.residual_group.blocks.0.attn.q.bias", "layers.3.residual_group.blocks.0.attn.proj.weight", "layers.3.residual_group.blocks.0.attn.proj.bias", "layers.3.residual_group.blocks.0.norm2.weight", "layers.3.residual_group.blocks.0.norm2.bias", "layers.3.residual_group.blocks.0.mlp.fc1.weight", "layers.3.residual_group.blocks.0.mlp.fc1.bias", "layers.3.residual_group.blocks.0.mlp.dwconv.depthwise_conv.0.weight", "layers.3.residual_group.blocks.0.mlp.dwconv.depthwise_conv.0.bias", "layers.3.residual_group.blocks.0.mlp.fc2.weight", "layers.3.residual_group.blocks.0.mlp.fc2.bias", "layers.3.residual_group.blocks.1.attn_mask", "layers.3.residual_group.blocks.1.norm1.weight", "layers.3.residual_group.blocks.1.norm1.bias", "layers.3.residual_group.blocks.1.attn.relative_position_bias_table", "layers.3.residual_group.blocks.1.attn.aligned_relative_position_index", "layers.3.residual_group.blocks.1.attn.kv.weight", "layers.3.residual_group.blocks.1.attn.kv.bias", "layers.3.residual_group.blocks.1.attn.q.weight", "layers.3.residual_group.blocks.1.attn.q.bias", "layers.3.residual_group.blocks.1.attn.proj.weight", "layers.3.residual_group.blocks.1.attn.proj.bias", "layers.3.residual_group.blocks.1.norm2.weight", "layers.3.residual_group.blocks.1.norm2.bias", "layers.3.residual_group.blocks.1.mlp.fc1.weight", "layers.3.residual_group.blocks.1.mlp.fc1.bias", "layers.3.residual_group.blocks.1.mlp.dwconv.depthwise_conv.0.weight", "layers.3.residual_group.blocks.1.mlp.dwconv.depthwise_conv.0.bias", "layers.3.residual_group.blocks.1.mlp.fc2.weight", "layers.3.residual_group.blocks.1.mlp.fc2.bias", "layers.3.residual_group.blocks.2.norm1.weight", "layers.3.residual_group.blocks.2.norm1.bias", "layers.3.residual_group.blocks.2.attn.relative_position_bias_table", "layers.3.residual_group.blocks.2.attn.aligned_relative_position_index", "layers.3.residual_group.blocks.2.attn.kv.weight", "layers.3.residual_group.blocks.2.attn.kv.bias", "layers.3.residual_group.blocks.2.attn.q.weight", "layers.3.residual_group.blocks.2.attn.q.bias", "layers.3.residual_group.blocks.2.attn.proj.weight", "layers.3.residual_group.blocks.2.attn.proj.bias", "layers.3.residual_group.blocks.2.norm2.weight", "layers.3.residual_group.blocks.2.norm2.bias", "layers.3.residual_group.blocks.2.mlp.fc1.weight", "layers.3.residual_group.blocks.2.mlp.fc1.bias", "layers.3.residual_group.blocks.2.mlp.dwconv.depthwise_conv.0.weight", "layers.3.residual_group.blocks.2.mlp.dwconv.depthwise_conv.0.bias", "layers.3.residual_group.blocks.2.mlp.fc2.weight", "layers.3.residual_group.blocks.2.mlp.fc2.bias", "layers.3.residual_group.blocks.3.attn_mask", "layers.3.residual_group.blocks.3.norm1.weight", "layers.3.residual_group.blocks.3.norm1.bias", "layers.3.residual_group.blocks.3.attn.relative_position_bias_table", "layers.3.residual_group.blocks.3.attn.aligned_relative_position_index", "layers.3.residual_group.blocks.3.attn.kv.weight", "layers.3.residual_group.blocks.3.attn.kv.bias", "layers.3.residual_group.blocks.3.attn.q.weight", "layers.3.residual_group.blocks.3.attn.q.bias", "layers.3.residual_group.blocks.3.attn.proj.weight", "layers.3.residual_group.blocks.3.attn.proj.bias", "layers.3.residual_group.blocks.3.norm2.weight", "layers.3.residual_group.blocks.3.norm2.bias", "layers.3.residual_group.blocks.3.mlp.fc1.weight", "layers.3.residual_group.blocks.3.mlp.fc1.bias", "layers.3.residual_group.blocks.3.mlp.dwconv.depthwise_conv.0.weight", "layers.3.residual_group.blocks.3.mlp.dwconv.depthwise_conv.0.bias", "layers.3.residual_group.blocks.3.mlp.fc2.weight", "layers.3.residual_group.blocks.3.mlp.fc2.bias", "layers.3.residual_group.blocks.4.norm1.weight", "layers.3.residual_group.blocks.4.norm1.bias", "layers.3.residual_group.blocks.4.attn.relative_position_bias_table", "layers.3.residual_group.blocks.4.attn.aligned_relative_position_index", "layers.3.residual_group.blocks.4.attn.kv.weight", "layers.3.residual_group.blocks.4.attn.kv.bias", "layers.3.residual_group.blocks.4.attn.q.weight", "layers.3.residual_group.blocks.4.attn.q.bias", "layers.3.residual_group.blocks.4.attn.proj.weight", "layers.3.residual_group.blocks.4.attn.proj.bias", "layers.3.residual_group.blocks.4.norm2.weight", "layers.3.residual_group.blocks.4.norm2.bias", "layers.3.residual_group.blocks.4.mlp.fc1.weight", "layers.3.residual_group.blocks.4.mlp.fc1.bias", "layers.3.residual_group.blocks.4.mlp.dwconv.depthwise_conv.0.weight", "layers.3.residual_group.blocks.4.mlp.dwconv.depthwise_conv.0.bias", "layers.3.residual_group.blocks.4.mlp.fc2.weight", "layers.3.residual_group.blocks.4.mlp.fc2.bias", "layers.3.residual_group.blocks.5.attn_mask", "layers.3.residual_group.blocks.5.norm1.weight", "layers.3.residual_group.blocks.5.norm1.bias", "layers.3.residual_group.blocks.5.attn.relative_position_bias_table", "layers.3.residual_group.blocks.5.attn.aligned_relative_position_index", "layers.3.residual_group.blocks.5.attn.kv.weight", "layers.3.residual_group.blocks.5.attn.kv.bias", "layers.3.residual_group.blocks.5.attn.q.weight", "layers.3.residual_group.blocks.5.attn.q.bias", "layers.3.residual_group.blocks.5.attn.proj.weight", "layers.3.residual_group.blocks.5.attn.proj.bias", "layers.3.residual_group.blocks.5.norm2.weight", "layers.3.residual_group.blocks.5.norm2.bias", "layers.3.residual_group.blocks.5.mlp.fc1.weight", "layers.3.residual_group.blocks.5.mlp.fc1.bias", "layers.3.residual_group.blocks.5.mlp.dwconv.depthwise_conv.0.weight", "layers.3.residual_group.blocks.5.mlp.dwconv.depthwise_conv.0.bias", "layers.3.residual_group.blocks.5.mlp.fc2.weight", "layers.3.residual_group.blocks.5.mlp.fc2.bias", "layers.3.conv.weight", "layers.3.conv.bias", "layers.4.residual_group.blocks.0.norm1.weight", "layers.4.residual_group.blocks.0.norm1.bias", "layers.4.residual_group.blocks.0.attn.relative_position_bias_table", "layers.4.residual_group.blocks.0.attn.aligned_relative_position_index", "layers.4.residual_group.blocks.0.attn.kv.weight", "layers.4.residual_group.blocks.0.attn.kv.bias", "layers.4.residual_group.blocks.0.attn.q.weight", "layers.4.residual_group.blocks.0.attn.q.bias", "layers.4.residual_group.blocks.0.attn.proj.weight", "layers.4.residual_group.blocks.0.attn.proj.bias", "layers.4.residual_group.blocks.0.norm2.weight", "layers.4.residual_group.blocks.0.norm2.bias", "layers.4.residual_group.blocks.0.mlp.fc1.weight", "layers.4.residual_group.blocks.0.mlp.fc1.bias", "layers.4.residual_group.blocks.0.mlp.dwconv.depthwise_conv.0.weight", "layers.4.residual_group.blocks.0.mlp.dwconv.depthwise_conv.0.bias", "layers.4.residual_group.blocks.0.mlp.fc2.weight", "layers.4.residual_group.blocks.0.mlp.fc2.bias", "layers.4.residual_group.blocks.1.attn_mask", "layers.4.residual_group.blocks.1.norm1.weight", "layers.4.residual_group.blocks.1.norm1.bias", "layers.4.residual_group.blocks.1.attn.relative_position_bias_table", "layers.4.residual_group.blocks.1.attn.aligned_relative_position_index", "layers.4.residual_group.blocks.1.attn.kv.weight", "layers.4.residual_group.blocks.1.attn.kv.bias", "layers.4.residual_group.blocks.1.attn.q.weight", "layers.4.residual_group.blocks.1.attn.q.bias", "layers.4.residual_group.blocks.1.attn.proj.weight", "layers.4.residual_group.blocks.1.attn.proj.bias", "layers.4.residual_group.blocks.1.norm2.weight", "layers.4.residual_group.blocks.1.norm2.bias", "layers.4.residual_group.blocks.1.mlp.fc1.weight", "layers.4.residual_group.blocks.1.mlp.fc1.bias", "layers.4.residual_group.blocks.1.mlp.dwconv.depthwise_conv.0.weight", "layers.4.residual_group.blocks.1.mlp.dwconv.depthwise_conv.0.bias", "layers.4.residual_group.blocks.1.mlp.fc2.weight", "layers.4.residual_group.blocks.1.mlp.fc2.bias", "layers.4.residual_group.blocks.2.norm1.weight", "layers.4.residual_group.blocks.2.norm1.bias", "layers.4.residual_group.blocks.2.attn.relative_position_bias_table", "layers.4.residual_group.blocks.2.attn.aligned_relative_position_index", "layers.4.residual_group.blocks.2.attn.kv.weight", "layers.4.residual_group.blocks.2.attn.kv.bias", "layers.4.residual_group.blocks.2.attn.q.weight", "layers.4.residual_group.blocks.2.attn.q.bias", "layers.4.residual_group.blocks.2.attn.proj.weight", "layers.4.residual_group.blocks.2.attn.proj.bias", "layers.4.residual_group.blocks.2.norm2.weight", "layers.4.residual_group.blocks.2.norm2.bias", "layers.4.residual_group.blocks.2.mlp.fc1.weight", "layers.4.residual_group.blocks.2.mlp.fc1.bias", "layers.4.residual_group.blocks.2.mlp.dwconv.depthwise_conv.0.weight", "layers.4.residual_group.blocks.2.mlp.dwconv.depthwise_conv.0.bias", "layers.4.residual_group.blocks.2.mlp.fc2.weight", "layers.4.residual_group.blocks.2.mlp.fc2.bias", "layers.4.residual_group.blocks.3.attn_mask", "layers.4.residual_group.blocks.3.norm1.weight", "layers.4.residual_group.blocks.3.norm1.bias", "layers.4.residual_group.blocks.3.attn.relative_position_bias_table", "layers.4.residual_group.blocks.3.attn.aligned_relative_position_index", "layers.4.residual_group.blocks.3.attn.kv.weight", "layers.4.residual_group.blocks.3.attn.kv.bias", "layers.4.residual_group.blocks.3.attn.q.weight", "layers.4.residual_group.blocks.3.attn.q.bias", "layers.4.residual_group.blocks.3.attn.proj.weight", "layers.4.residual_group.blocks.3.attn.proj.bias", "layers.4.residual_group.blocks.3.norm2.weight", "layers.4.residual_group.blocks.3.norm2.bias", "layers.4.residual_group.blocks.3.mlp.fc1.weight", "layers.4.residual_group.blocks.3.mlp.fc1.bias", "layers.4.residual_group.blocks.3.mlp.dwconv.depthwise_conv.0.weight", "layers.4.residual_group.blocks.3.mlp.dwconv.depthwise_conv.0.bias", "layers.4.residual_group.blocks.3.mlp.fc2.weight", "layers.4.residual_group.blocks.3.mlp.fc2.bias", "layers.4.residual_group.blocks.4.norm1.weight", "layers.4.residual_group.blocks.4.norm1.bias", "layers.4.residual_group.blocks.4.attn.relative_position_bias_table", "layers.4.residual_group.blocks.4.attn.aligned_relative_position_index", "layers.4.residual_group.blocks.4.attn.kv.weight", "layers.4.residual_group.blocks.4.attn.kv.bias", "layers.4.residual_group.blocks.4.attn.q.weight", "layers.4.residual_group.blocks.4.attn.q.bias", "layers.4.residual_group.blocks.4.attn.proj.weight", "layers.4.residual_group.blocks.4.attn.proj.bias", "layers.4.residual_group.blocks.4.norm2.weight", "layers.4.residual_group.blocks.4.norm2.bias", "layers.4.residual_group.blocks.4.mlp.fc1.weight", "layers.4.residual_group.blocks.4.mlp.fc1.bias", "layers.4.residual_group.blocks.4.mlp.dwconv.depthwise_conv.0.weight", "layers.4.residual_group.blocks.4.mlp.dwconv.depthwise_conv.0.bias", "layers.4.residual_group.blocks.4.mlp.fc2.weight", "layers.4.residual_group.blocks.4.mlp.fc2.bias", "layers.4.residual_group.blocks.5.attn_mask", "layers.4.residual_group.blocks.5.norm1.weight", "layers.4.residual_group.blocks.5.norm1.bias", "layers.4.residual_group.blocks.5.attn.relative_position_bias_table", "layers.4.residual_group.blocks.5.attn.aligned_relative_position_index", "layers.4.residual_group.blocks.5.attn.kv.weight", "layers.4.residual_group.blocks.5.attn.kv.bias", "layers.4.residual_group.blocks.5.attn.q.weight", "layers.4.residual_group.blocks.5.attn.q.bias", "layers.4.residual_group.blocks.5.attn.proj.weight", "layers.4.residual_group.blocks.5.attn.proj.bias", "layers.4.residual_group.blocks.5.norm2.weight", "layers.4.residual_group.blocks.5.norm2.bias", "layers.4.residual_group.blocks.5.mlp.fc1.weight", "layers.4.residual_group.blocks.5.mlp.fc1.bias", "layers.4.residual_group.blocks.5.mlp.dwconv.depthwise_conv.0.weight", "layers.4.residual_group.blocks.5.mlp.dwconv.depthwise_conv.0.bias", "layers.4.residual_group.blocks.5.mlp.fc2.weight", "layers.4.residual_group.blocks.5.mlp.fc2.bias", "layers.4.conv.weight", "layers.4.conv.bias", "layers.5.residual_group.blocks.0.norm1.weight", "layers.5.residual_group.blocks.0.norm1.bias", "layers.5.residual_group.blocks.0.attn.relative_position_bias_table", "layers.5.residual_group.blocks.0.attn.aligned_relative_position_index", "layers.5.residual_group.blocks.0.attn.kv.weight", "layers.5.residual_group.blocks.0.attn.kv.bias", "layers.5.residual_group.blocks.0.attn.q.weight", "layers.5.residual_group.blocks.0.attn.q.bias", "layers.5.residual_group.blocks.0.attn.proj.weight", "layers.5.residual_group.blocks.0.attn.proj.bias", "layers.5.residual_group.blocks.0.norm2.weight", "layers.5.residual_group.blocks.0.norm2.bias", "layers.5.residual_group.blocks.0.mlp.fc1.weight", "layers.5.residual_group.blocks.0.mlp.fc1.bias", "layers.5.residual_group.blocks.0.mlp.dwconv.depthwise_conv.0.weight", "layers.5.residual_group.blocks.0.mlp.dwconv.depthwise_conv.0.bias", "layers.5.residual_group.blocks.0.mlp.fc2.weight", "layers.5.residual_group.blocks.0.mlp.fc2.bias", "layers.5.residual_group.blocks.1.attn_mask", "layers.5.residual_group.blocks.1.norm1.weight", "layers.5.residual_group.blocks.1.norm1.bias", "layers.5.residual_group.blocks.1.attn.relative_position_bias_table", "layers.5.residual_group.blocks.1.attn.aligned_relative_position_index", "layers.5.residual_group.blocks.1.attn.kv.weight", "layers.5.residual_group.blocks.1.attn.kv.bias", "layers.5.residual_group.blocks.1.attn.q.weight", "layers.5.residual_group.blocks.1.attn.q.bias", "layers.5.residual_group.blocks.1.attn.proj.weight", "layers.5.residual_group.blocks.1.attn.proj.bias", "layers.5.residual_group.blocks.1.norm2.weight", "layers.5.residual_group.blocks.1.norm2.bias", "layers.5.residual_group.blocks.1.mlp.fc1.weight", "layers.5.residual_group.blocks.1.mlp.fc1.bias", "layers.5.residual_group.blocks.1.mlp.dwconv.depthwise_conv.0.weight", "layers.5.residual_group.blocks.1.mlp.dwconv.depthwise_conv.0.bias", "layers.5.residual_group.blocks.1.mlp.fc2.weight", "layers.5.residual_group.blocks.1.mlp.fc2.bias", "layers.5.residual_group.blocks.2.norm1.weight", "layers.5.residual_group.blocks.2.norm1.bias", "layers.5.residual_group.blocks.2.attn.relative_position_bias_table", "layers.5.residual_group.blocks.2.attn.aligned_relative_position_index", "layers.5.residual_group.blocks.2.attn.kv.weight", "layers.5.residual_group.blocks.2.attn.kv.bias", "layers.5.residual_group.blocks.2.attn.q.weight", "layers.5.residual_group.blocks.2.attn.q.bias", "layers.5.residual_group.blocks.2.attn.proj.weight", "layers.5.residual_group.blocks.2.attn.proj.bias", "layers.5.residual_group.blocks.2.norm2.weight", "layers.5.residual_group.blocks.2.norm2.bias", "layers.5.residual_group.blocks.2.mlp.fc1.weight", "layers.5.residual_group.blocks.2.mlp.fc1.bias", "layers.5.residual_group.blocks.2.mlp.dwconv.depthwise_conv.0.weight", "layers.5.residual_group.blocks.2.mlp.dwconv.depthwise_conv.0.bias", "layers.5.residual_group.blocks.2.mlp.fc2.weight", "layers.5.residual_group.blocks.2.mlp.fc2.bias", "layers.5.residual_group.blocks.3.attn_mask", "layers.5.residual_group.blocks.3.norm1.weight", "layers.5.residual_group.blocks.3.norm1.bias", "layers.5.residual_group.blocks.3.attn.relative_position_bias_table", "layers.5.residual_group.blocks.3.attn.aligned_relative_position_index", "layers.5.residual_group.blocks.3.attn.kv.weight", "layers.5.residual_group.blocks.3.attn.kv.bias", "layers.5.residual_group.blocks.3.attn.q.weight", "layers.5.residual_group.blocks.3.attn.q.bias", "layers.5.residual_group.blocks.3.attn.proj.weight", "layers.5.residual_group.blocks.3.attn.proj.bias", "layers.5.residual_group.blocks.3.norm2.weight", "layers.5.residual_group.blocks.3.norm2.bias", "layers.5.residual_group.blocks.3.mlp.fc1.weight", "layers.5.residual_group.blocks.3.mlp.fc1.bias", "layers.5.residual_group.blocks.3.mlp.dwconv.depthwise_conv.0.weight", "layers.5.residual_group.blocks.3.mlp.dwconv.depthwise_conv.0.bias", "layers.5.residual_group.blocks.3.mlp.fc2.weight", "layers.5.residual_group.blocks.3.mlp.fc2.bias", "layers.5.residual_group.blocks.4.norm1.weight", "layers.5.residual_group.blocks.4.norm1.bias", "layers.5.residual_group.blocks.4.attn.relative_position_bias_table", "layers.5.residual_group.blocks.4.attn.aligned_relative_position_index", "layers.5.residual_group.blocks.4.attn.kv.weight", "layers.5.residual_group.blocks.4.attn.kv.bias", "layers.5.residual_group.blocks.4.attn.q.weight", "layers.5.residual_group.blocks.4.attn.q.bias", "layers.5.residual_group.blocks.4.attn.proj.weight", "layers.5.residual_group.blocks.4.attn.proj.bias", "layers.5.residual_group.blocks.4.norm2.weight", "layers.5.residual_group.blocks.4.norm2.bias", "layers.5.residual_group.blocks.4.mlp.fc1.weight", "layers.5.residual_group.blocks.4.mlp.fc1.bias", "layers.5.residual_group.blocks.4.mlp.dwconv.depthwise_conv.0.weight", "layers.5.residual_group.blocks.4.mlp.dwconv.depthwise_conv.0.bias", "layers.5.residual_group.blocks.4.mlp.fc2.weight", "layers.5.residual_group.blocks.4.mlp.fc2.bias", "layers.5.residual_group.blocks.5.attn_mask", "layers.5.residual_group.blocks.5.norm1.weight", "layers.5.residual_group.blocks.5.norm1.bias", "layers.5.residual_group.blocks.5.attn.relative_position_bias_table", "layers.5.residual_group.blocks.5.attn.aligned_relative_position_index", "layers.5.residual_group.blocks.5.attn.kv.weight", "layers.5.residual_group.blocks.5.attn.kv.bias", "layers.5.residual_group.blocks.5.attn.q.weight", "layers.5.residual_group.blocks.5.attn.q.bias", "layers.5.residual_group.blocks.5.attn.proj.weight", "layers.5.residual_group.blocks.5.attn.proj.bias", "layers.5.residual_group.blocks.5.norm2.weight", "layers.5.residual_group.blocks.5.norm2.bias", "layers.5.residual_group.blocks.5.mlp.fc1.weight", "layers.5.residual_group.blocks.5.mlp.fc1.bias", "layers.5.residual_group.blocks.5.mlp.dwconv.depthwise_conv.0.weight", "layers.5.residual_group.blocks.5.mlp.dwconv.depthwise_conv.0.bias", "layers.5.residual_group.blocks.5.mlp.fc2.weight", "layers.5.residual_group.blocks.5.mlp.fc2.bias", "layers.5.conv.weight", "layers.5.conv.bias", "norm.weight", "norm.bias", "conv_after_body.weight", "conv_after_body.bias", "conv_before_upsample.0.weight", "conv_before_upsample.0.bias", "upsample.0.weight", "upsample.0.bias", "conv_last.weight", "conv_last.bias". 
	Unexpected key(s) in state_dict: "params".

couldnt find Set5/x2HR in the dataset

Hi;

I got error it didn't find set5 in the dataset. from where could download those files.

Thank you

torch.distributed.elastic.multiprocessing.errors.ChildFailedError:

Speed up the block by computing the mask in GPU device

It seems the mask is calculated in the CPU device, which slows the training processing.
When moving it into the GPU device, the speed boosts.

not work with CUDA_VISIBLE_DEVICES

Have you tested SwinIR with ConvFFN and SRFormer with Window size 32 * 32?

Also any tested U-net structure like scunet?

111

运行您提供的SRFormerLight_SRx4_DIV2K.pth模型，在set5数据集上结果如下：

can you share the training log?

In the process of reproducing, taking the light weight SR task as an example, it can reach around 38.0db on set5, which is still 0.2db away from the paper?If there is a log, it will help me

Possible to release Realworld image SR model?

Results in paper look very good, would be nice to try it.

Can we use your code on Single Image dataset?

please?

Real world model weight load problem about SRFormer_S_RealSR.pth

Thank you for your open source code, but when I load the pre-trained weights, some weights are lost.

pretrain_network_g: /home/tzz/Work_file/CODE/SRFormer-main/SRFormer_S_RealSR.pth
options: test_SRFormer-S_x4_real.yml

2024-04-25 21:51:18,810 INFO: Loading SRFormer model from /home/tzz/Work_file/CODE/SRFormer-main/SRFormer_S_RealSR.pth, with param key: [params_ema].
2024-04-25 21:51:18,834 WARNING: Current net - loaded net:
2024-04-25 21:51:18,834 WARNING: layers.0.residual_group.blocks.1.attn_mask
2024-04-25 21:51:18,834 WARNING: layers.0.residual_group.blocks.3.attn_mask
2024-04-25 21:51:18,834 WARNING: layers.0.residual_group.blocks.5.attn_mask
2024-04-25 21:51:18,834 WARNING: layers.1.residual_group.blocks.1.attn_mask
2024-04-25 21:51:18,834 WARNING: layers.1.residual_group.blocks.3.attn_mask
2024-04-25 21:51:18,834 WARNING: layers.1.residual_group.blocks.5.attn_mask
2024-04-25 21:51:18,834 WARNING: layers.2.residual_group.blocks.1.attn_mask
2024-04-25 21:51:18,834 WARNING: layers.2.residual_group.blocks.3.attn_mask
2024-04-25 21:51:18,834 WARNING: layers.2.residual_group.blocks.5.attn_mask
2024-04-25 21:51:18,834 WARNING: layers.3.residual_group.blocks.1.attn_mask
2024-04-25 21:51:18,834 WARNING: layers.3.residual_group.blocks.3.attn_mask
2024-04-25 21:51:18,834 WARNING: layers.3.residual_group.blocks.5.attn_mask
2024-04-25 21:51:18,834 WARNING: layers.4.residual_group.blocks.1.attn_mask
2024-04-25 21:51:18,834 WARNING: layers.4.residual_group.blocks.3.attn_mask
2024-04-25 21:51:18,834 WARNING: layers.4.residual_group.blocks.5.attn_mask
2024-04-25 21:51:18,834 WARNING: layers.5.residual_group.blocks.1.attn_mask
2024-04-25 21:51:18,834 WARNING: layers.5.residual_group.blocks.3.attn_mask
2024-04-25 21:51:18,834 WARNING: layers.5.residual_group.blocks.5.attn_mask
2024-04-25 21:51:18,834 WARNING: Loaded net - current net:
2024-04-25 21:51:18,859 INFO: Model [SRModel] is created.

Output is not sharp

Hi, very interesting work, I wonder if a slightly modified architecture can be used for image deblurring also. I tried the following modified model for deblurring 128x128 images

network_g:
  type: SRFormer
  upscale: 1
  in_chans: 3
  img_size: 128
  window_size: 16
  img_range: 1.
  depths: [6, 6, 6, 6]
  embed_dim: 60
  num_heads: [6, 6, 6, 6]
  mlp_ratio: 2
  # upsampler: None
  resi_connection: '1conv'

with the simple loss function

l1_loss = torch.nn.L1Loss(reduction='mean')

The outputs which I am getting are decent, but it is not able to produce sharp edges and corners. It seems to be a smooth output, closer to some oil painting. Can you please tell me if there is something wrong with this method?

I had a couple of other doubts about the model architecture:

I didn't understand the img_size parameter. I am able to pass any sized image as input and the model gives the output without any error.
Are you not using cross-window attention as SwinIR does through shifted windows?

Thanks a lot!!!

what is the batch size in DF2K experiments?

Thanks for your perfect work, and I have two questions:

what is the batch size and input resolusion in the DF2K experiments train in your paper?
what is the iteration number in DF2K train process?
thanks a lot!

KeyError: "No object named 'SRFormer' found in 'arch' registry!"

Hi, when I tried to reproduce the project by the tips in the readme, I occured to the question that
raise KeyError(f"No object named '{name}' found in '{self._name}' registry!")
KeyError: "No object named 'SRFormer' found in 'arch' registry!",how can I solve the problem since I was the first to do the image super resolution.

Pretrain base models shape(/parameters) missmatch with training config

When using the "SRFormer_SRx2_DF2K.pth" model as a pretrain for training another model, I ran into shape missmatch errors.
So I checked and recognized that the parameters in the current training config are different (=incompatible) with the ones that had been used for those pretrained models.

The SRFormer/options/train/SRFormer/train_SRFormer_SRx2_scratch.yml file uses an img_size: 48 and window_size: 24 while the pretrain had been trained with (as seen in SRFormer/options/test/SRFormer/test_SRFormer_DF2Ksrx2.yml) with img_size: 64 and window_size: 22.

This is meant as an information for someone wanting to train a base model (or do inference with one of the base pretrains), if you want to use the released pretrain models as a pretrain, you need to adapt the parameters in the train config (simply copy the network settings from the test config).

The lightweight model does not suffer from this inconsistency, the test and training configs hold the same network parameters.

how to inference my own image?

I think that test code is evaluation for training.

I use pretrained model for inference.

Can you guide me?

Thank you.

Params and FLOPs

I have a question about Tbale 5. Why are the parameters in it different from FLOPs and other papers, such as ELAN?

123

您好，自己训练模型后，测试时模型无法加载，您知道是为什么吗？

Usage of the real-world pretrained models

real-world image SR pretrained model?

请问在Real-world SR预训练模型是否忘记上传呢？

SRformer based on Real-ESRGAN?

#19 (comment)

Is this still coming at some point?

Request to add a huggingface

Or add a method to inference with local images

How long will it take for the real-world pretrained models to be released?

Community updates ;)

Wanted to share some community-based updates concerning SRFormer:

the image processing GUI chaiNNer just added SRFormer support in its latest release Alpha v0.20.0
additional community trained SRFormer models can be found on openmodeldb when filtering for the SRFormer network (there are at least two more community trained SRFormer models I know of, which are not yet on openmodeldb)
the SRFormer network is supported by the neosr framework for training real-world single-image super-resolution networks, which had been released just two weeks ago

Question about training machine.

Thanks for your excellent work! I would like to ask how many GPUs are used for training SRFormer and SwinIR? What type of GPUs are they? Looking forward to your reply.

Problems encountered during tensor reshaping operations in PyTorch

Hello, I encountered the following problem while reproducing the code. Can you please answer it? Thank you very much
x = x.view(b, h // window_size, window_size, w // window_size, window_size, c)
RuntimeError: shape '[1, 11, 3, 11, 3, 1]' is invalid for input of size 1225

Thank you :) Wanted to show you my SRFormer_light model :)

Thank you for this network :)
I just wanted to show real quick that I have trained a (4x) SRFormer_light model for anime super-resolution with real degradation (compression, noise, blur), and created visual outputs which can be compared to a SwinIR_light (=small) model I have trained on the same dataset with the same config (so same losses, gt_size, batch size, both from scratch meaning no pretrain). Just one visual comparison (more outputs in the link below) of Input, SwinIR_light and SRFormer_light:

More visual outputs can be found here and the trained SRFormer_light model file can be found here

Recommend Projects

React

A declarative, efficient, and flexible JavaScript library for building user interfaces.
Vue.js

🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript

TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
TensorFlow

An Open Source Machine Learning Framework for Everyone
Django

The Web framework for perfectionists with deadlines.
Laravel

A PHP framework for web artisans
D3

Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

javascript

JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
web

Some thing interesting about web. New door for the world.
server

A server is a program made to process requests and deliver data to clients.
Machine learning

Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Visualization

Some thing interesting about visualization, use data art
Game

Some thing interesting about game, make everyone happy.

Recommend Org

Facebook

We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft

Open source projects and samples from Microsoft.
Google

Google ❤️ Open Source for everyone.
Alibaba

Alibaba Open Source for everyone
D3

Data-Driven Documents codes.
Tencent

China tencent open source team.