[Hotfix] Fix GAT example and clarify the usage of early stop (#1065)

* upd

* rm redundancy:

* upd

examples/mxnet/gat/README.md

@@ -19,5 +19,7 @@ pip install requests
 
 ### Usage (make sure that DGLBACKEND is changed into mxnet)
 ```bash
-DGLBACKEND=mxnet python3 train.py --dataset cora --gpu 0 --num-heads 8
+DGLBACKEND=mxnet python3 train.py --dataset cora --gpu 0
+DGLBACKEND=mxnet python3 train.py --dataset citeseer --gpu 0 --early-stop
+DGLBACKEND=mxnet python3 train.py --dataset pubmed --gpu 0 --early-stop
 ```

examples/mxnet/gat/train.py

 """
 Graph Attention Networks in DGL using SPMV optimization.
 Multiple heads are also batched together for faster training.
-Compared with the original paper, this code does not implement
-early stopping.
 References
 ----------
 Paper: https://arxiv.org/abs/1710.10903
@@ -76,6 +74,7 @@ def main(args):
                 args.alpha,
                 args.residual)
 
+    if args.early_stop:
         stopper = EarlyStopping(patience=100)
     model.initialize(ctx=ctx)
 
@@ -99,8 +98,12 @@ def main(args):
             epoch, loss.asnumpy()[0], np.mean(dur), n_edges / np.mean(dur) / 1000))
         val_accuracy = evaluate(model, features, labels, val_mask)
         print("Validation Accuracy {:.4f}".format(val_accuracy))
+        if args.early_stop:
             if stopper.step(val_accuracy, model): 
                 break
+    print()
+
+    if args.early_stop:
         model.load_parameters('model.param')
     test_accuracy = evaluate(model, features, labels, test_mask)
     print("Test Accuracy {:.4f}".format(test_accuracy))
@@ -134,6 +137,8 @@ if __name__ == '__main__':
                         help="weight decay")
     parser.add_argument('--alpha', type=float, default=0.2,
                         help="the negative slop of leaky relu")
+    parser.add_argument('--early-stop', action='store_true', default=False,
+                        help="indicates whether to use early stop or not")
     args = parser.parse_args()
     print(args)
 

examples/pytorch/gat/README.md

@@ -27,11 +27,11 @@ python3 train.py --dataset=cora --gpu=0
 ```
 
 ```bash
-python3 train.py --dataset=citeseer --gpu=0
+python3 train.py --dataset=citeseer --gpu=0 --early-stop
 ```
 
 ```bash
-python3 train.py --dataset=pubmed --gpu=0 --num-out-heads=8 --weight-decay=0.001
+python3 train.py --dataset=pubmed --gpu=0 --num-out-heads=8 --weight-decay=0.001 --early-stop
 ```
 
 ```bash
@@ -43,9 +43,9 @@ Results
 
 | Dataset | Test Accuracy | Time(s) | Baseline#1 times(s) | Baseline#2 times(s) |
 | ------- | ------------- | ------- | ------------------- | ------------------- |
-| Cora | 84.0% | 0.0113 | 0.0982 (**8.7x**) | 0.0424 (**3.8x**) |
-| Citeseer | 70.7% | 0.0111 | n/a | n/a |
-| Pubmed | 78.0% | 0.0115 | n/a | n/a |
+| Cora | 84.02(0.40) | 0.0113 | 0.0982 (**8.7x**) | 0.0424 (**3.8x**) |
+| Citeseer | 70.91(0.79) | 0.0111 | n/a | n/a |
+| Pubmed | 78.57(0.75) | 0.0115 | n/a | n/a |
 
 * All the accuracy numbers are obtained after 300 epochs.
 * The time measures how long it takes to train one epoch.

examples/pytorch/gat/train.py

 """
 Graph Attention Networks in DGL using SPMV optimization.
 Multiple heads are also batched together for faster training.
-Compared with the original paper, this code does not implement
-early stopping.
 References
 ----------
 Paper: https://arxiv.org/abs/1710.10903
@@ -95,6 +93,7 @@ def main(args):
                 args.negative_slope,
                 args.residual)
     print(model)
+    if args.early_stop:
         stopper = EarlyStopping(patience=100)
     if cuda:
         model.cuda()
@@ -127,6 +126,7 @@ def main(args):
             val_acc = accuracy(logits[val_mask], labels[val_mask])
         else:
             val_acc = evaluate(model, features, labels, val_mask)
+            if args.early_stop:
                 if stopper.step(val_acc, model):   
                     break
 
@@ -136,6 +136,7 @@ def main(args):
                      val_acc, n_edges / np.mean(dur) / 1000))
 
     print()
+    if args.early_stop:
         model.load_state_dict(torch.load('es_checkpoint.pt'))
     acc = evaluate(model, features, labels, test_mask)
     print("Test Accuracy {:.4f}".format(acc))
@@ -169,6 +170,8 @@ if __name__ == '__main__':
                         help="weight decay")
     parser.add_argument('--negative-slope', type=float, default=0.2,
                         help="the negative slope of leaky relu")
+    parser.add_argument('--early-stop', action='store_true', default=False,
+                        help="indicates whether to use early stop or not")
     parser.add_argument('--fastmode', action="store_true", default=False,
                         help="skip re-evaluate the validation set")
     args = parser.parse_args()