diff --git a/egs/librispeech/ASR/pruned_transducer_stateless7/zipformer.py b/egs/librispeech/ASR/pruned_transducer_stateless7/zipformer.py
index 1c181b1ac..2bcd7202d 100644
--- a/egs/librispeech/ASR/pruned_transducer_stateless7/zipformer.py
+++ b/egs/librispeech/ASR/pruned_transducer_stateless7/zipformer.py
@@ -853,7 +853,8 @@ class SimpleCombiner(torch.nn.Module):
                  min_weight: Tuple[float] = (0., 0.)):
         super(SimpleCombiner, self).__init__()
         assert dim2 >= dim1
-        self.weight1 = nn.Parameter(torch.ones(dim2) * min_weight[0])
+        initial_weight1 = 0.1
+        self.weight1 = nn.Parameter(torch.full((dim2,), initial_weight1))
         self.min_weight = min_weight
 
     def forward(self,
diff --git a/icefall/utils.py b/icefall/utils.py
index ad079222e..09523fdf2 100644
--- a/icefall/utils.py
+++ b/icefall/utils.py
@@ -530,9 +530,10 @@ class MetricsTracker(collections.defaultdict):
     def __add__(self, other: "MetricsTracker") -> "MetricsTracker":
         ans = MetricsTracker()
         for k, v in self.items():
-            ans[k] = v
+            ans[k] = v if v - v == 0.0 else 0.0  # discard infinities.
         for k, v in other.items():
-            ans[k] = ans[k] + v
+            if v - v == 0:  # discard infinities.
+                ans[k] = ans[k] + v
         return ans
 
     def __mul__(self, alpha: float) -> "MetricsTracker":