diff --git a/README.md b/README.md
index 6ded296a7..34539494a 100644
--- a/README.md
+++ b/README.md
@@ -6,6 +6,15 @@ I'm playing with [PyTorch](http://pytorch.org/) on the CIFAR10 dataset.
 - Python 3.6+
 - PyTorch 1.0+
 
+## Training
+```
+# Start training with: 
+CUDA_VISIBLE_DEVICES=0 python main.py
+
+# You can manually resume the training with: 
+CUDA_VISIBLE_DEVICES=0 python main.py --resume --lr=0.01
+```
+
 ## Accuracy
 | Model             | Acc.        |
 | ----------------- | ----------- |
@@ -22,11 +31,5 @@ I'm playing with [PyTorch](http://pytorch.org/) on the CIFAR10 dataset.
 | [DenseNet121](https://arxiv.org/abs/1608.06993)       | 95.04%      |
 | [PreActResNet18](https://arxiv.org/abs/1603.05027)    | 95.11%      |
 | [DPN92](https://arxiv.org/abs/1707.01629)             | 95.16%      |
+| [DLA](https://arxiv.org/abs/1707.064)                 | 95.47%      |
 
-## Learning rate adjustment
-I manually change the `lr` during training:
-- `0.1` for epoch `[0,150)`
-- `0.01` for epoch `[150,250)`
-- `0.001` for epoch `[250,350)`
-
-Resume the training with `python main.py --resume --lr=0.01`
diff --git a/main.py b/main.py
index 3491a845c..05ca1eb90 100644
--- a/main.py
+++ b/main.py
@@ -86,7 +86,7 @@
 criterion = nn.CrossEntropyLoss()
 optimizer = optim.SGD(net.parameters(), lr=args.lr,
                       momentum=0.9, weight_decay=5e-4)
-scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer, T_max=100)
+scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer, T_max=200)
 
 
 # Training
@@ -148,7 +148,7 @@ def test(epoch):
         best_acc = acc
 
 
-for epoch in range(start_epoch, start_epoch+100):
+for epoch in range(start_epoch, start_epoch+200):
     train(epoch)
     test(epoch)
     scheduler.step()