Use torch.logsumexp in advanced_tutorial.py

`torch.logsumexp` is numerically stabilized: https://pytorch.org/docs/stable/generated/torch.logsumexp.html

Found with TorchFix https://github.com/pytorch-labs/torchfix/

Author: kit1980

beginner_source/nlp/advanced_tutorial.py

@@ -142,8 +142,7 @@ def prepare_sequence(seq, to_ix):
 def log_sum_exp(vec):
     max_score = vec[0, argmax(vec)]
     max_score_broadcast = max_score.view(1, -1).expand(1, vec.size()[1])
-    return max_score + \
-        torch.log(torch.sum(torch.exp(vec - max_score_broadcast)))
+    return max_score + torch.logsumexp(vec - max_score_broadcast)
 
 #####################################################################
 # Create model