Merge branches 'master' and 'master' of github.com:unanmed/ginka-generator

ginka/model/loss.py

@@ -335,13 +335,13 @@ class GinkaLoss(nn.Module):
         
         losses = [
             minamo_loss * self.weight[0],
-            border_loss * self.weight[1] * 0.1,
+            border_loss * self.weight[1],
             entrance_loss * self.weight[2],
             count_loss * self.weight[3],
             illegal_loss * self.weight[4]
         ]
     
         # 梯度归一化
-        scaled_losses = [loss / (loss.detach() + 1e-6) for loss in losses]
-        total_loss = sum(scaled_losses)
-        return total_loss
+        # scaled_losses = [loss / (loss.detach() + 1e-6) for loss in losses]
+        total_loss = sum(losses)
+        return total_loss

ginka/train.py

@@ -8,19 +8,21 @@ from .model.model import GinkaModel
 from .model.loss import GinkaLoss
 from .dataset import GinkaDataset
 from minamo.model.model import MinamoModel
+from shared.args import parse_arguments
 
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 os.makedirs("result", exist_ok=True)
 os.makedirs("result/ginka_checkpoint", exist_ok=True)
 
-epochs = 150
-
 # 在生成器输出后添加梯度检查钩子
 def grad_hook(module, grad_input, grad_output):
     print(f"Generator output grad norm: {grad_output[0].norm().item()}")
 
 def train():
     print(f"Using {'cuda' if torch.cuda.is_available() else 'cpu'} to train model.")
+    
+    args = parse_arguments("result/ginka.pth", "ginka-dataset.json", 'ginka-eval.json')
+    
     model = GinkaModel()
     model.to(device)
     minamo = MinamoModel(32)
@@ -53,9 +55,15 @@ def train():
     # model.register_full_backward_hook(grad_hook)
     # converter.register_full_backward_hook(grad_hook)
     # criterion.register_full_backward_hook(grad_hook)
+    if args.resume:
+        data = torch.load(args.from_state, map_location=device)
+        model.load_state_dict(data["model_state"])
+        if args.load_optim:
+            optimizer.load_state_dict(data["optimizer_state"])
+        print("Train from loaded state.")
     
     # 开始训练
-    for epoch in tqdm(range(epochs)):
+    for epoch in tqdm(range(args.epochs)):
         model.train()
         total_loss = 0
         
@@ -118,7 +126,7 @@ def train():
             tqdm.write(f"[INFO {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}] Validation::loss: {avg_val_loss:.6f}")
             torch.save({
                 "model_state": model.state_dict(),
-                "optimizer_state": optimizer.state_dict(),
+                # "optimizer_state": optimizer.state_dict(),
             }, f"result/ginka_checkpoint/{epoch + 1}.pth")
                 
         
@@ -126,7 +134,7 @@ def train():
     
     torch.save({
         "model_state": model.state_dict(),
-        "optimizer_state": optimizer.state_dict(),
+        # "optimizer_state": optimizer.state_dict(),
     }, f"result/ginka.pth")
 
 if __name__ == "__main__":

minamo/train.py

@@ -8,13 +8,12 @@ from tqdm import tqdm
 from .model.model import MinamoModel
 from .model.loss import MinamoLoss
 from .dataset import MinamoDataset
+from shared.args import parse_arguments
 
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 os.makedirs("result", exist_ok=True)
 os.makedirs("result/minamo_checkpoint", exist_ok=True)
 
-epochs = 150
-
 def collate_fn(batch):
     """动态处理不同尺寸地图的批处理"""
     map1_batch = [item[0] for item in batch]
@@ -35,6 +34,9 @@ def collate_fn(batch):
 
 def train():
     print(f"Using {'cuda' if torch.cuda.is_available() else 'cpu'} to train model.")
+    
+    args = parse_arguments("result/minamo.pth", "minamo-dataset.json", 'minamo-eval.json')
+    
     model = MinamoModel(32)
     model.to(device)
 
@@ -57,8 +59,15 @@ def train():
     scheduler = optim.lr_scheduler.CosineAnnealingWarmRestarts(optimizer, T_0=10, T_mult=2, eta_min=1e-6)
     criterion = MinamoLoss()
     
+    if args.resume:
+        data = torch.load(args.from_state, map_location=device)
+        model.load_state_dict(data["model_state"])
+        if args.load_optim:
+            optimizer.load_state_dict(data["optimizer_state"])
+        print("Train from loaded state.")
+    
     # 开始训练
-    for epoch in tqdm(range(epochs)):
+    for epoch in tqdm(range(args.epochs)):
         model.train()
         total_loss = 0
         

shared/args.py

@@ -0,0 +1,12 @@
+import argparse    
+    
+def parse_arguments(from_default: str, train_default: str, val_default: str):
+    parser = argparse.ArgumentParser(description="training codes")
+    parser.add_argument("--resume", type=bool, default=False)
+    parser.add_argument("--from_state", type=str, default=from_default)
+    parser.add_argument("--load_optim", type=bool, default=False)
+    parser.add_argument("--train", type=str, default=train_default)
+    parser.add_argument("--validate", type=str, default=val_default)
+    parser.add_argument("--epochs", type=int, default=150)
+    args = parser.parse_args()
+    return args