Avoid error if svd fails

egs/librispeech/ASR/pruned_transducer_stateless7/optim.py

@@ -913,13 +913,14 @@ class NeutralGradient(Optimizer):
         P = torch.matmul(Y, Y.t())
 
         if random.random() < 1.0: #0.025:
-
-            # TEMP:
-            _,s,_ = P.svd()
-            print(f"Min,max eig of P: {s.min()},{s.max()}")
-
             # TODO: remove this testing code.
             assert (P - P.t()).abs().mean() < 0.01  # make sure symmetric.
+            try:
+                P = 0.5 * (P + P.t())
+                _,s,_ = P.svd()
+                print(f"Min,max eig of P: {s.min()},{s.max()}")
+            except:
+                pass
             # testing...  note, this is only true modulo "eps"
             C_check = torch.matmul(torch.matmul(P, G), P)
             C_smoothed = torch.matmul(Q.t(), Q)