Upload 3594 files

visual_results/test_SSANet_color_denoising_15_visual/test_test_SSANet_color_denoising_15_visual_20260307_232946.log

@@ -0,0 +1,1549 @@
+2026-03-07 23:29:46,696 INFO: 
+                ____                _       _____  ____
+               / __ ) ____ _ _____ (_)_____/ ___/ / __ \
+              / __  |/ __ `// ___// // ___/\__ \ / /_/ /
+             / /_/ // /_/ /(__  )/ // /__ ___/ // _, _/
+            /_____/ \__,_//____//_/ \___//____//_/ |_|
+     ______                   __   __                 __      __
+    / ____/____   ____   ____/ /  / /   __  __ _____ / /__   / /
+   / / __ / __ \ / __ \ / __  /  / /   / / / // ___// //_/  / /
+  / /_/ // /_/ // /_/ // /_/ /  / /___/ /_/ // /__ / /<    /_/
+  \____/ \____/ \____/ \____/  /_____/\____/ \___//_/|_|  (_)
+    
+Version Information: 
+	BasicSR: 1.4.2
+	PyTorch: 2.1.1+cu121
+	TorchVision: 0.16.1+cu121
+2026-03-07 23:29:46,697 INFO: 
+  name: test_SSANet_color_denoising_15_visual
+  model_type: SRModel
+  scale: 1
+  num_gpu: 8
+  manual_seed: 0
+  datasets:[
+    val_0:[
+      name: Kodak24
+      type: DatasetDnCNN
+      dataroot_gt: /data2/songlianhong/Color_denoise/Kodak24
+      dataroot_lq: None
+      io_backend:[
+        type: disk
+      ]
+      sigma: 15
+      sigma_test: 15
+      phase: val
+      scale: 1
+    ]
+    val_1:[
+      name: CBSD68
+      type: DatasetDnCNN
+      dataroot_gt: /data2/songlianhong/Color_denoise/CBSD68
+      dataroot_lq: None
+      io_backend:[
+        type: disk
+      ]
+      sigma: 15
+      sigma_test: 15
+      phase: val
+      scale: 1
+    ]
+    val_2:[
+      name: McMaster
+      type: DatasetDnCNN
+      dataroot_gt: /data2/songlianhong/Color_denoise/McMaster
+      dataroot_lq: None
+      io_backend:[
+        type: disk
+      ]
+      sigma: 15
+      sigma_test: 15
+      phase: val
+      scale: 1
+    ]
+    val_3:[
+      name: Urban100
+      type: DatasetDnCNN
+      dataroot_gt: /data2/songlianhong/Color_denoise/Urban100
+      dataroot_lq: None
+      io_backend:[
+        type: disk
+      ]
+      sigma: 15
+      sigma_test: 15
+      phase: val
+      scale: 1
+    ]
+  ]
+  network_g:[
+    type: SSANet
+    dim: 160
+    num_blocks: 8
+    num_layers: 8
+    upscale: 1
+    num_heads: 1
+    ffn_expansion_factor: 2.0
+    img_range: 1.0
+    use_pixelunshuffle: True
+    tlc_flag: True
+    tlc_kernel: 72
+    activation: relu
+  ]
+  path:[
+    pretrain_network_g: experiments/denoising/ASSANet_15.pth
+    strict_load_g: True
+    resume_state: None
+    results_root: /data2/songlianhong/DLGSANet/20241107/results/test_SSANet_color_denoising_15_visual
+    log: /data2/songlianhong/DLGSANet/20241107/results/test_SSANet_color_denoising_15_visual
+    visualization: /data2/songlianhong/DLGSANet/20241107/results/test_SSANet_color_denoising_15_visual/visualization
+  ]
+  val:[
+    save_img: True
+    suffix: None
+    metrics:[
+      psnr:[
+        type: calculate_psnr
+        crop_border: 0
+        test_y_channel: False
+      ]
+    ]
+  ]
+  dist_params:[
+    backend: nccl
+    port: 22144
+  ]
+  dist: False
+  rank: 0
+  world_size: 1
+  auto_resume: False
+  is_train: False
+
+2026-03-07 23:29:46,697 INFO: Dataset [DatasetDnCNN] - Kodak24 is built.
+2026-03-07 23:29:46,698 INFO: Number of test images in Kodak24: 24
+2026-03-07 23:29:46,698 INFO: Dataset [DatasetDnCNN] - CBSD68 is built.
+2026-03-07 23:29:46,698 INFO: Number of test images in CBSD68: 68
+2026-03-07 23:29:46,699 INFO: Dataset [DatasetDnCNN] - McMaster is built.
+2026-03-07 23:29:46,699 INFO: Number of test images in McMaster: 18
+2026-03-07 23:29:46,699 INFO: Dataset [DatasetDnCNN] - Urban100 is built.
+2026-03-07 23:29:46,700 INFO: Number of test images in Urban100: 100
+2026-03-07 23:29:46,895 INFO: Network [SSANet] is created.
+2026-03-07 23:29:47,149 INFO: Network: DataParallel - SSANet, with parameters: 18,923,756
+2026-03-07 23:29:47,150 INFO: SSANet(
+  (overlap_embed): Sequential(
+    (0): OverlapPatchEmbed(
+      (proj): Sequential(
+        (0): PixelUnshuffle(downscale_factor=2)
+        (1): Conv2d(12, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
+      )
+    )
+  )
+  (deep_feature_extraction): Sequential(
+    (0): BuildBlock(
+      (body): Sequential(
+        (0): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (1): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (2): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (3): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (4): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (5): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (6): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (7): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (8): Conv2d(160, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
+      )
+    )
+    (1): BuildBlock(
+      (body): Sequential(
+        (0): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (1): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (2): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (3): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (4): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (5): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (6): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (7): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (8): Conv2d(160, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
+      )
+    )
+    (2): BuildBlock(
+      (body): Sequential(
+        (0): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (1): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (2): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (3): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (4): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (5): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (6): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (7): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (8): Conv2d(160, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
+      )
+    )
+    (3): BuildBlock(
+      (body): Sequential(
+        (0): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (1): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (2): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (3): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (4): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (5): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (6): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (7): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (8): Conv2d(160, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
+      )
+    )
+    (4): BuildBlock(
+      (body): Sequential(
+        (0): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (1): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (2): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (3): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (4): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (5): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (6): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (7): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (8): Conv2d(160, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
+      )
+    )
+    (5): BuildBlock(
+      (body): Sequential(
+        (0): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (1): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (2): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (3): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (4): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (5): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (6): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (7): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (8): Conv2d(160, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
+      )
+    )
+    (6): BuildBlock(
+      (body): Sequential(
+        (0): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (1): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (2): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (3): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (4): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (5): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (6): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (7): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (8): Conv2d(160, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
+      )
+    )
+    (7): BuildBlock(
+      (body): Sequential(
+        (0): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (1): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (2): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (3): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (4): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (5): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (6): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (7): AttBlock(
+          (norm1): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (norm2): LayerNorm(
+            (body): WithBias_LayerNorm()
+          )
+          (attn): SparseSelfAttention(
+            (project_in): Conv2d(160, 320, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dynamic_conv): IDynamicDWConv(
+              (weight): Conv2d(320, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=320, bias=False)
+            )
+            (project_out): Conv2d(160, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (act): ReLU()
+          )
+          (ffn): FeedForward(
+            (project_in): Conv2d(160, 640, kernel_size=(1, 1), stride=(1, 1), bias=False)
+            (dwconv): Conv2d(640, 640, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=640, bias=False)
+            (project_out): Conv2d(320, 160, kernel_size=(1, 1), stride=(1, 1), bias=False)
+          )
+        )
+        (8): Conv2d(160, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
+      )
+    )
+    (8): Conv2d(160, 160, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
+  )
+  (upsample): UpsampleOneStep(
+    (0): Conv2d(160, 12, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
+    (1): PixelShuffle(upscale_factor=2)
+  )
+)
+2026-03-07 23:29:47,261 INFO: Loading SSANet model from experiments/denoising/ASSANet_15.pth, with param key: [params].
+2026-03-07 23:29:47,387 INFO: Model [SRModel] is created.
+2026-03-07 23:29:47,632 INFO: Loading SSANet model from experiments/denoising/ASSANet_15.pth, with param key: [params_ema].
+2026-03-07 23:29:47,750 INFO: Testing Kodak24...
+2026-03-07 23:30:29,585 INFO: Validation Kodak24
+	 # psnr: 35.4199	Best: 35.4199 @ test_SSANet_color_denoising_15_visual iter
+
+2026-03-07 23:30:29,587 INFO: Testing CBSD68...
+2026-03-07 23:31:53,098 INFO: Validation CBSD68
+	 # psnr: 34.4658	Best: 34.4658 @ test_SSANet_color_denoising_15_visual iter
+
+2026-03-07 23:31:53,098 INFO: Testing McMaster...
+2026-03-07 23:32:39,971 INFO: Validation McMaster
+	 # psnr: 35.6979	Best: 35.6979 @ test_SSANet_color_denoising_15_visual iter
+
+2026-03-07 23:32:39,972 INFO: Testing Urban100...
+2026-03-07 23:42:44,597 INFO: Validation Urban100
+	 # psnr: 35.3656	Best: 35.3656 @ test_SSANet_color_denoising_15_visual iter
+