diff --git a/config.json b/config.json
index b9214573..6887d973 100644
--- a/config.json
+++ b/config.json
@@ -27,11 +27,15 @@
     "embedding_size": 256,    
     "text_cleaner": "english_cleaners",
     "epochs": 1000,
-    "lr": 0.0015,
+    
+    // "lr": 0.0015,
+    "lr": 0.0001,
+    "lr_decay": false,
     "warmup_steps": 4000,
-    "batch_size":32,
+
+    "batch_size": 32,
     "eval_batch_size":32,
-    "r": 1,
+    "r": 5,
     "wd": 0.000001,
     "checkpoint": true,
     "save_step": 5000,
diff --git a/train.py b/train.py
index bbe55515..9426aa3c 100644
--- a/train.py
+++ b/train.py
@@ -58,7 +58,8 @@ def train(model, criterion, criterion_st, data_loader, optimizer, optimizer_st,
             epoch * len(data_loader) + 1
 
         # setup lr
-        scheduler.step()
+        if c.lr_decay:
+            scheduler.step()
         optimizer.zero_grad()
         optimizer_st.zero_grad()
 
@@ -92,6 +93,7 @@ def train(model, criterion, criterion_st, data_loader, optimizer, optimizer_st,
 
         # backpass and check the grad norm for spec losses
         loss.backward(retain_graph=True)
+        # custom weight decay
         for group in optimizer.param_groups:
             for param in group['params']:
                 param.data = param.data.add(-c.wd * group['lr'], param.data)