Fix some checkpoint / model str regressions

Author: rwightman

tests/test_models.py

@@ -86,6 +86,15 @@ def test_model_default_cfgs(model_name, batch_size):
     assert any([k.startswith(first_conv) for k in state_dict.keys()]), f'{first_conv} not in model params'
 
 
+if 'GITHUB_ACTIONS' not in os.environ:
+    @pytest.mark.timeout(120)
+    @pytest.mark.parametrize('model_name', list_models())
+    @pytest.mark.parametrize('batch_size', [1])
+    def test_model_load_pretrained(model_name, batch_size):
+        """Run a single forward pass with each model"""
+        create_model(model_name, pretrained=True)
+
+
 EXCLUDE_JIT_FILTERS = [
     '*iabn*', 'tresnet*',  # models using inplace abn unlikely to ever be scriptable
     'dla*', 'hrnet*',  # hopefully fix at some point

timm/models/cspnet.py

@@ -433,7 +433,7 @@ def cspresnext50(pretrained=False, **kwargs):
 @register_model
 def cspresnext50_iabn(pretrained=False, **kwargs):
     norm_layer = get_norm_act_layer('iabn')
-    return _create_cspnet('cspresnext50', pretrained=pretrained, norm_layer=norm_layer, **kwargs)
+    return _create_cspnet('cspresnext50_iabn', pretrained=pretrained, norm_layer=norm_layer, **kwargs)
 
 
 @register_model
@@ -444,7 +444,7 @@ def cspdarknet53(pretrained=False, **kwargs):
 @register_model
 def cspdarknet53_iabn(pretrained=False, **kwargs):
     norm_layer = get_norm_act_layer('iabn')
-    return _create_cspnet('cspdarknet53', pretrained=pretrained, block_fn=DarkBlock, norm_layer=norm_layer, **kwargs)
+    return _create_cspnet('cspdarknet53_iabn', pretrained=pretrained, block_fn=DarkBlock, norm_layer=norm_layer, **kwargs)
 
 
 @register_model

timm/models/res2net.py

@@ -189,7 +189,7 @@ def res2net50_48w_2s(pretrained=False, **kwargs):
     """
     model_args = dict(
         block=Bottle2neck, layers=[3, 4, 6, 3], base_width=48, block_args=dict(scale=2), **kwargs)
-    return _create_res2net('res2net50_26w_8s', pretrained, **model_args)
+    return _create_res2net('res2net50_48w_2s', pretrained, **model_args)
 
 
 @register_model
@@ -200,7 +200,7 @@ def res2net50_14w_8s(pretrained=False, **kwargs):
     """
     model_args = dict(
         block=Bottle2neck, layers=[3, 4, 6, 3], base_width=14, block_args=dict(scale=8), **kwargs)
-    return _create_res2net('res2net50_26w_8s', pretrained, **model_args)
+    return _create_res2net('res2net50_14w_8s', pretrained, **model_args)
 
 
 @register_model

timm/models/resnet.py

@@ -624,7 +624,7 @@ def resnet26d(pretrained=False, **kwargs):
     """Constructs a ResNet-26 v1d model.
     This is technically a 28 layer ResNet, sticking with 'd' modifier from Gluon for now.
     """
-    model_args = dict(block=Bottleneck, layers=[2, 2, 2, 2], stem_type='deep', avg_down=True, **kwargs)
+    model_args = dict(block=Bottleneck, layers=[2, 2, 2, 2], stem_width=32, stem_type='deep', avg_down=True, **kwargs)
     return _create_resnet('resnet26d', pretrained, **model_args)
 
 
@@ -1129,9 +1129,3 @@ def senet154(pretrained=False, **kwargs):
         block=Bottleneck, layers=[3, 8, 36, 3], cardinality=64, base_width=4, stem_type='deep',
         down_kernel_size=3, block_reduce_first=2, block_args=dict(attn_layer='se'), **kwargs)
     return _create_resnet('senet154', pretrained, **model_args)
-
-
-@register_model
-def eseresnet50(pretrained=False, **kwargs):
-    model_args = dict(block=Bottleneck, layers=[3, 4, 6, 3], block_args=dict(attn_layer='ese'), **kwargs)
-    return _create_resnet('seresnet50', pretrained, **model_args)