Fix pooling in mnasnet, more sensible default for AMP opt level

models/adaptive_avgmax_pool.py

@@ -49,7 +49,7 @@ def select_adaptive_pool2d(x, pool_type='avg', output_size=1):
     return x
 
 
-class AdaptiveAvgMaxPool2d(torch.nn.Module):
+class AdaptiveAvgMaxPool2d(nn.Module):
     def __init__(self, output_size=1):
         super(AdaptiveAvgMaxPool2d, self).__init__()
         self.output_size = output_size
@@ -58,7 +58,7 @@ class AdaptiveAvgMaxPool2d(torch.nn.Module):
         return adaptive_avgmax_pool2d(x, self.output_size)
 
 
-class AdaptiveCatAvgMaxPool2d(torch.nn.Module):
+class AdaptiveCatAvgMaxPool2d(nn.Module):
     def __init__(self, output_size=1):
         super(AdaptiveCatAvgMaxPool2d, self).__init__()
         self.output_size = output_size
@@ -67,7 +67,7 @@ class AdaptiveCatAvgMaxPool2d(torch.nn.Module):
         return adaptive_catavgmax_pool2d(x, self.output_size)
 
 
-class SelectAdaptivePool2d(torch.nn.Module):
+class SelectAdaptivePool2d(nn.Module):
     """Selectable global pooling layer with dynamic input kernel size
     """
     def __init__(self, output_size=1, pool_type='avg'):

models/mnasnet.py

@@ -185,7 +185,6 @@ class MnasBlock(nn.Module):
         # Pointwise projection
         x = self.conv_project(x)
         x = self.bn2(x)
-        # Residual
         if self.has_residual:
             return x + residual
         else:
@@ -268,7 +267,7 @@ class MnasNet(nn.Module):
         x = self.bn1(x)
         x = self.act_fn(x)
         if pool:
-            x = self.avg_pool(x)
+            x = self.global_pool(x)
             x = x.view(x.size(0), -1)
         return x
 

train.py

@@ -156,6 +156,9 @@ def main():
         global_pool=args.gp,
         checkpoint_path=args.initial_checkpoint)
 
+    print('Model %s created, param count: %d' %
+          (args.model, sum([m.numel() for m in model.parameters()])))
+
     data_config = resolve_data_config(model, args, verbose=args.local_rank == 0)
 
     # optionally resume from a checkpoint
@@ -178,7 +181,7 @@ def main():
         optimizer.load_state_dict(optimizer_state)
 
     if has_apex and args.amp:
-        model, optimizer = amp.initialize(model, optimizer, opt_level='O3')
+        model, optimizer = amp.initialize(model, optimizer, opt_level='O1')
         use_amp = True
         print('AMP enabled')
     else: