update train script

evaluation.py

@@ -0,0 +1,41 @@
+import tqdm
+import torch
+
+
+def performance(loader, net, device):
+    net.eval()
+    correct, total = 0, 0
+    with torch.no_grad():
+        for batch_idx, (inputs, targets) in enumerate(tqdm.tqdm(loader)):
+            inputs, targets  = inputs.to(device), targets.to(device)
+            outputs = net(inputs)
+            _, predicted = torch.max(outputs.data, 1)
+            total += targets.size(0)
+            correct += predicted.eq(targets.data).cpu().sum().item()
+        acc = correct/total
+    return acc
+
+
+if __name__ == "__main__":
+    import argparse
+    from ward import WARD
+    import torch.utils.data as Data
+
+    parser = argparse.ArgumentParser(description='Feature Graph Networks')
+    parser.add_argument("--load", type=str, required=True, help="load pretrained model file")
+    parser.add_argument("--device", type=str, default='cuda:0', help="cuda or cpu")
+    parser.add_argument("--data-root", type=str, default='/data/datasets', help="dataset location to be download")
+    parser.add_argument("--duration", type=int, default=50, help="duration")
+    parser.add_argument("--batch-size", type=int, default=100, help="minibatch size")
+    args = parser.parse_args(); print(args)
+
+    test_data = WARD(root=args.data_root, duration=args.duration, train=False)
+    test_loader = Data.DataLoader(dataset=test_data, batch_size=args.batch_size, shuffle=True, drop_last=True)
+    train_data = WARD(root=args.data_root, duration=args.duration, train=True)
+    train_loader = Data.DataLoader(dataset=train_data, batch_size=args.batch_size, shuffle=True, drop_last=True)
+
+    net = torch.load(args.load)
+    train_acc = performance(train_loader, net, args.device)
+    test_acc = performance(test_loader, net, args.device)
+    print("Evaluating model: ", args.load)
+    print("Train Acc: %f; Test Acc: %f"%(train_acc, test_acc))

lifelong.py

@@ -14,7 +14,7 @@
 from gat import GAT
 from fgn import FGN
 from ward import WARD
-from nonlifelong import performance
+from evaluation import performance
 from torch_util import count_parameters
 
 
@@ -97,8 +97,7 @@ def sample(self, inputs, targets):
     parser.add_argument("--device", type=str, default='cuda:0', help="cuda or cpu")
     parser.add_argument("--data-root", type=str, default='/data/datasets', help="dataset location to be download")
     parser.add_argument("--model", type=str, default='FGN', help="FGN or GAT")
-    parser.add_argument("--load", type=str, default=None, help="load pretrained model file")
-    parser.add_argument("--save", type=str, default='saves/test', help="model file to save")
+    parser.add_argument("--save", type=str, default='saves', help="location to save model")
     parser.add_argument("--optim", type=str, default='SGD', help="SGD or Adam")
     parser.add_argument("--lr", type=float, default=1e-3, help="learning rate")
     parser.add_argument("--duration", type=int, default=50, help="duration")
@@ -109,7 +108,7 @@ def sample(self, inputs, targets):
     parser.add_argument("--seed", type=int, default=0, help='Random seed.')
     args = parser.parse_args(); print(args)
     os.makedirs(args.data_root, exist_ok=True)
-    os.makedirs('saves', exist_ok=True)
+    os.makedirs(args.save, exist_ok=True)
     torch.manual_seed(args.seed)
     Nets = {'fgn':FGN, 'gat':GAT}
     Net = Nets[args.model.lower()]
@@ -129,5 +128,10 @@ def sample(self, inputs, targets):
             test_acc = performance(test_loader, lgl.net, args.device)
             print('Test Acc: %f'%(test_acc))
             if args.save is not None:
-                print('Saving model to', args.save+'-%d.model'%(batch_idx))
-                torch.save(lgl.net, args.save+'-%d.model'%(batch_idx))
+                filename = args.save+'/lifelong-%s-s%d-it%d.model'%(args.model, args.seed, batch_idx+1)
+                print('Saving model to', filename)
+                torch.save(lgl.net, filename)
+
+    test_acc = performance(test_loader, lgl.net, args.device)
+    print('Final Test Acc: %f'%(test_acc))
+    torch.save(lgl.net, args.save+'/lifelong-%s-s%d.model'%(args.model, args.seed))

nonlifelong.py

@@ -13,24 +13,11 @@
 from gat import GAT
 from fgn import FGN
 from ward import WARD
+from evaluation import performance
 from torch_util import count_parameters
 from torch_util import EarlyStopScheduler
 
 
-def performance(loader, net, device):
-    net.eval()
-    correct, total = 0, 0
-    with torch.no_grad():
-        for batch_idx, (inputs, targets) in enumerate(tqdm.tqdm(loader)):
-            inputs, targets  = inputs.to(device), targets.to(device)
-            outputs = net(inputs)
-            _, predicted = torch.max(outputs.data, 1)
-            total += targets.size(0)
-            correct += predicted.eq(targets.data).cpu().sum().item()
-        acc = correct/total
-    return acc
-
-
 def train(loader, net, device):
     net.train()
     correct, total = 0, 0
@@ -54,8 +41,7 @@ def train(loader, net, device):
     parser.add_argument("--device", type=str, default='cuda:0', help="cuda or cpu")
     parser.add_argument("--data-root", type=str, default='/data/datasets', help="dataset location to be download")
     parser.add_argument("--model", type=str, default='FGN', help="FGN or GAT")
-    parser.add_argument("--load", type=str, default=None, help="load pretrained model file")
-    parser.add_argument("--save", type=str, default='saves/test', help="model file to save")
+    parser.add_argument("--save", type=str, default='saves', help="location to save model")
     parser.add_argument("--optim", type=str, default='SGD', help="SGD or Adam")
     parser.add_argument("--lr", type=float, default=1e-3, help="learning rate")
     parser.add_argument("--epoch", type=int, default=50, help="epoch")
@@ -64,7 +50,7 @@ def train(loader, net, device):
     parser.add_argument("--seed", type=int, default=0, help='Random seed.')
     args = parser.parse_args(); print(args)
     os.makedirs(args.data_root, exist_ok=True)
-    os.makedirs('saves', exist_ok=True)
+    os.makedirs(args.save, exist_ok=True)
     torch.manual_seed(args.seed)
     Nets = {'fgn':FGN, 'gat':GAT}
     Net = Nets[args.model.lower()]
@@ -88,8 +74,9 @@ def train(loader, net, device):
 
         if best_acc < test_acc and args.save is not None:
             best_acc = test_acc
-            print('Saving new best model to', args.save+'.model')
-            torch.save(net, args.save+'.model')
+            filename = args.save+'/nonlifelong-%s-s%d.model'%(args.model, args.seed)
+            print('Saving new best model to', filename)
+            torch.save(net, filename)
 
         if scheduler.step(1-test_acc):
             print('Early Stoping..')