reviese transformer_dla hparameter set: relu_dropout=0.1

tensor2tensor/models/transformer_dla.py

@@ -442,7 +442,7 @@ def transformer_dla_base():
   hparams.decoder_layers = 6
   hparams.decoder_layers = 6
   hparams.normalize_before = True
   hparams.normalize_before = True
   hparams.attention_dropout = 0.1
   hparams.attention_dropout = 0.1
-  hparams.residual_dropout = 0.1
+  hparams.relu_dropout = 0.1
   hparams.learning_rate = 0.4
   hparams.learning_rate = 0.4
   hparams.learning_rate_warmup_steps = 8000
   hparams.learning_rate_warmup_steps = 8000
   hparams.batch_size = 2048
   hparams.batch_size = 2048