Fix grad in-place operation (#779)

p.add_(-lr, p.grad) throws RuntimeError: a leaf Variable that requires grad is being used in an in-place operation, using p.data.add_(-lr, p.grad) fixes this issue.

Author: yangchris11

word_language_model/main.py

@@ -178,7 +178,7 @@ def train():
         # `clip_grad_norm` helps prevent the exploding gradient problem in RNNs / LSTMs.
         torch.nn.utils.clip_grad_norm_(model.parameters(), args.clip)
         for p in model.parameters():
-            p.add_(-lr, p.grad)
+            p.data.add_(-lr, p.grad)
 
         total_loss += loss.item()