Update SK network configs, add weights for skresnet8 and skresnext50

timm/models/sknet.py

@@ -13,7 +13,7 @@ def _cfg(url='', **kwargs):
     return {
         'url': url,
         'num_classes': 1000, 'input_size': (3, 224, 224), 'pool_size': (7, 7),
-        'crop_pct': 0.875, 'interpolation': 'bilinear',
+        'crop_pct': 0.875, 'interpolation': 'bicubic',
         'mean': IMAGENET_DEFAULT_MEAN, 'std': IMAGENET_DEFAULT_STD,
         'first_conv': 'conv1', 'classifier': 'fc',
         **kwargs
@@ -21,11 +21,13 @@ def _cfg(url='', **kwargs):
 
 
 default_cfgs = {
-    'skresnet18': _cfg(url=''),
-    'skresnet26d': _cfg(),
+    'skresnet18': _cfg(
+        url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/skresnet18_ra-4eec2804.pth'),
+    'skresnet34': _cfg(url=''),
     'skresnet50': _cfg(),
     'skresnet50d': _cfg(),
-    'skresnext50_32x4d': _cfg(),
+    'skresnext50_32x4d': _cfg(
+        url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/skresnext50_ra-f40e40bf.pth'),
 }
 
 
@@ -134,24 +136,10 @@ class SelectiveKernelBottleneck(nn.Module):
 
 @register_model
 def skresnet18(pretrained=False, num_classes=1000, in_chans=3, **kwargs):
-    """Constructs a ResNet-18 model.
-    """
-    default_cfg = default_cfgs['skresnet18']
-    sk_kwargs = dict(
-        min_attn_channels=16,
-    )
-    model = ResNet(
-        SelectiveKernelBasic, [2, 2, 2, 2], num_classes=num_classes, in_chans=in_chans,
-        block_args=dict(sk_kwargs=sk_kwargs), **kwargs)
-    model.default_cfg = default_cfg
-    if pretrained:
-        load_pretrained(model, default_cfg, num_classes, in_chans)
-    return model
-
+    """Constructs a Selective Kernel ResNet-18 model.
 
-@register_model
-def sksresnet18(pretrained=False, num_classes=1000, in_chans=3, **kwargs):
-    """Constructs a ResNet-18 model.
+    Different from configs in Select Kernel paper or "Compounding the Performance Improvements..." this
+    variation splits the input channels to the selective convolutions to keep param count down.
     """
     default_cfg = default_cfgs['skresnet18']
     sk_kwargs = dict(
@@ -169,17 +157,21 @@ def sksresnet18(pretrained=False, num_classes=1000, in_chans=3, **kwargs):
 
 
 @register_model
-def skresnet26d(pretrained=False, num_classes=1000, in_chans=3, **kwargs):
-    """Constructs a ResNet-26 model.
+def skresnet34(pretrained=False, num_classes=1000, in_chans=3, **kwargs):
+    """Constructs a Selective Kernel ResNet-34 model.
+
+    Different from configs in Select Kernel paper or "Compounding the Performance Improvements..." this
+    variation splits the input channels to the selective convolutions to keep param count down.
     """
-    default_cfg = default_cfgs['skresnet26d']
+    default_cfg = default_cfgs['skresnet34']
     sk_kwargs = dict(
-        keep_3x3=False,
+        min_attn_channels=16,
+        attn_reduction=8,
+        split_input=True
     )
     model = ResNet(
-        SelectiveKernelBottleneck, [2, 2, 2, 2],  stem_width=32, stem_type='deep', avg_down=True,
-        num_classes=num_classes, in_chans=in_chans, block_args=dict(sk_kwargs=sk_kwargs), zero_init_last_bn=False
-        **kwargs)
+        SelectiveKernelBasic, [3, 4, 6, 3], num_classes=num_classes, in_chans=in_chans,
+        block_args=dict(sk_kwargs=sk_kwargs), zero_init_last_bn=False, **kwargs)
     model.default_cfg = default_cfg
     if pretrained:
         load_pretrained(model, default_cfg, num_classes, in_chans)
@@ -189,11 +181,12 @@ def skresnet26d(pretrained=False, num_classes=1000, in_chans=3, **kwargs):
 @register_model
 def skresnet50(pretrained=False, num_classes=1000, in_chans=3, **kwargs):
     """Constructs a Select Kernel ResNet-50 model.
-    Based on config in "Compounding the Performance Improvements of Assembled Techniques in a
-    Convolutional Neural Network"
+
+    Different from configs in Select Kernel paper or "Compounding the Performance Improvements..." this
+    variation splits the input channels to the selective convolutions to keep param count down.
     """
     sk_kwargs = dict(
-        attn_reduction=2,
+        split_input=True,
     )
     default_cfg = default_cfgs['skresnet50']
     model = ResNet(
@@ -208,11 +201,12 @@ def skresnet50(pretrained=False, num_classes=1000, in_chans=3, **kwargs):
 @register_model
 def skresnet50d(pretrained=False, num_classes=1000, in_chans=3, **kwargs):
     """Constructs a Select Kernel ResNet-50-D model.
-    Based on config in "Compounding the Performance Improvements of Assembled Techniques in a
-    Convolutional Neural Network"
+
+    Different from configs in Select Kernel paper or "Compounding the Performance Improvements..." this
+    variation splits the input channels to the selective convolutions to keep param count down.
     """
     sk_kwargs = dict(
-        attn_reduction=2,
+        split_input=True,
     )
     default_cfg = default_cfgs['skresnet50d']
     model = ResNet(