Add gc_efficientnetv2_rw_t weights (global context instead of SE attn). Add TF XL weights even though the fine-tuned ones don't validate that well. Change default arg for GlobalContext to use scal (mul) mode.

Author: rwightman

tests/test_models.py

@@ -25,7 +25,7 @@
     # GitHub Linux runner is slower and hits memory limits sooner than MacOS, exclude bigger models
     EXCLUDE_FILTERS = [
         '*efficientnet_l2*', '*resnext101_32x48d', '*in21k', '*152x4_bitm', '*101x3_bitm', '*50x3_bitm',
-        '*nfnet_f3*', '*nfnet_f4*', '*nfnet_f5*', '*nfnet_f6*', '*nfnet_f7*', 
+        '*nfnet_f3*', '*nfnet_f4*', '*nfnet_f5*', '*nfnet_f6*', '*nfnet_f7*', '*efficientnetv2_xl*',
         '*resnetrs350*', '*resnetrs420*', 'xcit_large_24_p8*']
 else:
     EXCLUDE_FILTERS = []

timm/models/efficientnet.py

@@ -92,11 +92,6 @@ def _cfg(url='', **kwargs):
         interpolation='bilinear'),
 
     # NOTE experimenting with alternate attention
-    'eca_efficientnet_b0': _cfg(
-        url=''),
-    'gc_efficientnet_b0': _cfg(
-        url=''),
-
     'efficientnet_b0': _cfg(
         url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/efficientnet_b0_ra-3dd342df.pth'),
     'efficientnet_b1': _cfg(
@@ -169,7 +164,7 @@ def _cfg(url='', **kwargs):
         url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/efficientnetv2_t_agc-3620981a.pth',
         input_size=(3, 224, 224), test_input_size=(3, 288, 288), pool_size=(7, 7), crop_pct=1.0),
     'gc_efficientnetv2_rw_t': _cfg(
-        url='',
+        url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/gc_efficientnetv2_rw_t_agc-927a0bde.pth',
         input_size=(3, 224, 224), test_input_size=(3, 288, 288), pool_size=(7, 7), crop_pct=1.0),
     'efficientnetv2_rw_s': _cfg(
         url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/efficientnet_v2s_ra2_288-a6477665.pth',
@@ -362,7 +357,7 @@ def _cfg(url='', **kwargs):
         mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5),
         input_size=(3, 384, 384), test_input_size=(3, 480, 480), pool_size=(12, 12), crop_pct=1.0),
     'tf_efficientnetv2_xl_in21ft1k': _cfg(
-        url='',
+        url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-effv2-weights/tf_efficientnetv2_xl_in21ft1k-06c35c48.pth',
         mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5),
         input_size=(3, 384, 384), test_input_size=(3, 512, 512), pool_size=(12, 12), crop_pct=1.0),
 
@@ -379,7 +374,7 @@ def _cfg(url='', **kwargs):
         mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), num_classes=21843,
         input_size=(3, 384, 384), test_input_size=(3, 480, 480), pool_size=(12, 12), crop_pct=1.0),
     'tf_efficientnetv2_xl_in21k': _cfg(
-        url='',
+        url='https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-effv2-weights/tf_efficientnetv2_xl_in21k-fd7e8abf.pth',
         mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), num_classes=21843,
         input_size=(3, 384, 384), test_input_size=(3, 512, 512), pool_size=(12, 12), crop_pct=1.0),
 
@@ -1276,26 +1271,6 @@ def efficientnet_b0(pretrained=False, **kwargs):
     return model
 
 
-@register_model
-def eca_efficientnet_b0(pretrained=False, **kwargs):
-    """ EfficientNet-B0 w/ ECA attn """
-    # NOTE experimental config
-    model = _gen_efficientnet(
-        'eca_efficientnet_b0', se_layer='ecam', channel_multiplier=1.0, depth_multiplier=1.0,
-        pretrained=pretrained, **kwargs)
-    return model
-
-
-@register_model
-def gc_efficientnet_b0(pretrained=False, **kwargs):
-    """ EfficientNet-B0 w/ GlobalContext """
-    # NOTE experminetal config
-    model = _gen_efficientnet(
-        'gc_efficientnet_b0', se_layer='gc', channel_multiplier=1.0, depth_multiplier=1.0,
-        pretrained=pretrained, **kwargs)
-    return model
-
-
 @register_model
 def efficientnet_b1(pretrained=False, **kwargs):
     """ EfficientNet-B1 """

timm/models/layers/global_context.py

@@ -18,7 +18,7 @@
 
 class GlobalContext(nn.Module):
 
-    def __init__(self, channels, use_attn=True, fuse_add=True, fuse_scale=False, init_last_zero=False,
+    def __init__(self, channels, use_attn=True, fuse_add=False, fuse_scale=True, init_last_zero=False,
                  rd_ratio=1./8, rd_channels=None, rd_divisor=1, act_layer=nn.ReLU, gate_layer='sigmoid'):
         super(GlobalContext, self).__init__()
         act_layer = get_act_layer(act_layer)