Fix #1078, DarkNet has 6 feature maps. Make vgg and darknet out_indices handling/comments equivalent

timm/models/cspnet.py

@@ -408,10 +408,14 @@ class CspNet(nn.Module):
 
 def _create_cspnet(variant, pretrained=False, **kwargs):
     cfg_variant = variant.split('_')[0]
+    if 'darknet' in variant:
+        # NOTE: DarkNet is one of few models with stride==1 features w/ 6 out_indices [0..5]
+        kwargs.setdefault('out_indices', (0, 1, 2, 3, 4, 5))
     return build_model_with_cfg(
         CspNet, variant, pretrained,
         default_cfg=default_cfgs[variant],
-        feature_cfg=dict(flatten_sequential=True), model_cfg=model_cfgs[cfg_variant],
+        model_cfg=model_cfgs[cfg_variant],
+        feature_cfg=dict(flatten_sequential=True),
         **kwargs)
 
 

timm/models/vgg.py

@@ -179,13 +179,13 @@ def _filter_fn(state_dict):
 
 def _create_vgg(variant: str, pretrained: bool, **kwargs: Any) -> VGG:
     cfg = variant.split('_')[0]
-    # NOTE: VGG is one of the only models with stride==1 features, so indices are offset from other models
-    out_indices = kwargs.get('out_indices', (0, 1, 2, 3, 4, 5))
+    # NOTE: VGG is one of few models with stride==1 features w/ 6 out_indices [0..5]
+    kwargs.setdefault('out_indices', (0, 1, 2, 3, 4, 5))
     model = build_model_with_cfg(
         VGG, variant, pretrained,
         default_cfg=default_cfgs[variant],
         model_cfg=cfgs[cfg],
-        feature_cfg=dict(flatten_sequential=True, out_indices=out_indices),
+        feature_cfg=dict(flatten_sequential=True),
         pretrained_filter_fn=_filter_fn,
         **kwargs)
     return model