diff --git a/sotabench.py b/sotabench.py
index 3a10d71d..56215cc4 100644
--- a/sotabench.py
+++ b/sotabench.py
@@ -1,3 +1,4 @@
+import torch
 from torchbench.image_classification import ImageNet
 from timm import create_model
 from timm.data import resolve_data_config, create_transform
@@ -77,7 +78,7 @@ model_list = [
     _entry('mixnet_m', 'MixNet-M', '1907.09595'),
     _entry('mixnet_s', 'MixNet-S', '1907.09595'),
     _entry('mnasnet_100', 'MnasNet-B1', '1807.11626'),
-    _entry('mobilenetv3_100', 'MobileNet V3(1.0)', '1905.02244',
+    _entry('mobilenetv3_100', 'MobileNet V3-Large 1.0', '1905.02244',
            model_desc='Trained in PyTorch with RMSProp, exponential LR decay, and hyper-params matching '
                       'paper as closely as possible.'),
     _entry('resnet18', 'ResNet-18', '1812.01187'),
@@ -216,4 +217,6 @@ for m in model_list:
         data_root=os.environ.get('IMAGENET_DIR', './imagenet')
     )
 
+    torch.cuda.empty_cache()
+
 
diff --git a/timm/models/res2net.py b/timm/models/res2net.py
index 30dffc1a..3b503e52 100644
--- a/timm/models/res2net.py
+++ b/timm/models/res2net.py
@@ -58,7 +58,7 @@ class Bottle2neck(nn.Module):
         super(Bottle2neck, self).__init__()
         assert dilation == 1 and previous_dilation == 1  # FIXME support dilation
         self.scale = scale
-        self.is_first = True if stride > 1 or downsample is not None else False
+        self.is_first = stride > 1 or downsample is not None
         self.num_scales = max(1, scale - 1)
         width = int(math.floor(planes * (base_width / 64.0))) * cardinality
         outplanes = planes * self.expansion