feat: 谱归一化

cycle.sh

@@ -1,7 +1,7 @@
-i = $1
+i=$1
 while true
 do
     sh gan.sh "$i"
-    ((i++))
+    i=$((i+1))
     echo "第 $i 次循环完成"
 done

gan.sh

@@ -1,5 +1,5 @@
 # 训练部分
-python3 -m minamo.train --epochs 30 --resume true
+python3 -m minamo.train --epochs 10 --resume true
 python3 -m minamo.train --epochs 10 --resume true --train "datasets/minamo-dataset-1.json" --validate "datasets/minamo-eval-1.json"
 python3 -m minamo.train --epochs 10 --resume true
 python3 -m ginka.train --epochs 10 --resume true
@@ -8,8 +8,8 @@ python3 -m ginka.validate
 mv "minamo-dataset.json" "datasets/minamo-dataset-$1.json"
 mv "minamo-eval.json" "datasets/minamo-eval-$1.json"
 cd data
-pnpm minamo "../minamo-dataset.json" "../result/ginka_val.json" "../../Apeiria/project" assigned
-pnpm minamo "../minamo-eval.json" "../result/ginka_val.json" "../../Apeiria-eval/project" assigned:100:10
+pnpm minamo "../minamo-dataset.json" "../result/ginka_val.json" "../../Apeiria/project" assigned:100:2
+pnpm minamo "../minamo-eval.json" "../result/ginka_val.json" "../../Apeiria-eval/project" assigned:100:2
 pnpm merge "../datasets/minamo-dataset-merged.json" "../datasets/minamo-dataset-merged.json" "../datasets/minamo-dataset-$1.json"
 pnpm merge "../datasets/minamo-eval-merged.json" "../datasets/minamo-eval-merged.json" "../datasets/minamo-eval-$1.json"
 cd ..

ginka/model/model.py

@@ -5,7 +5,7 @@ from .unet import GinkaUNet
 from .sample import MapDownSample
 
 class GinkaModel(nn.Module):
-    def __init__(self, feat_dim=256, base_ch=32, num_classes=32):
+    def __init__(self, feat_dim=256, base_ch=64, num_classes=32):
         """Ginka Model 模型定义部分
         """
         super().__init__()
@@ -27,6 +27,6 @@ class GinkaModel(nn.Module):
         x = self.fc(feat)
         x = x.view(-1, self.base_ch, 32, 32)
         x = self.unet(x)
-        x = self.pool(x)
+        x = F.interpolate(x, (13, 13), mode='bilinear')
         return x, F.softmax(x, dim=1)
     

ginka/train.py

@@ -48,7 +48,7 @@ def train():
     )
     
     # 设定优化器与调度器
-    optimizer = optim.AdamW(model.parameters(), lr=1e-3)
+    optimizer = optim.AdamW(model.parameters(), lr=5e-3)
     scheduler = optim.lr_scheduler.CosineAnnealingWarmRestarts(optimizer, T_0=10, T_mult=2, eta_min=1e-6)
     criterion = GinkaLoss(minamo)
     

ginka/validate.py

@@ -66,7 +66,7 @@ def matrix_to_image_cv(map_matrix, tile_set, tile_size=32):
 def validate():
     print(f"Using {'cuda' if torch.cuda.is_available() else 'cpu'} to validate model.")
     model = GinkaModel()
-    state = torch.load("result/ginka_checkpoint/10.pth", map_location=device)["model_state"]
+    state = torch.load("result/ginka.pth", map_location=device)["model_state"]
     model.load_state_dict(state)
     model.to(device)
     

minamo/dataset.py

@@ -4,7 +4,7 @@ import torch.nn.functional as F
 from torch.utils.data import Dataset
 from shared.graph import convert_soft_map_to_graph
 
-def random_smooth_onehot(onehot_map, min_main=0.8, max_main=1.0, epsilon=0.8):
+def random_smooth_onehot(onehot_map, min_main=0.65, max_main=1.0, epsilon=0.35):
     """
     生成随机平滑的 one-hot 编码，使主类别概率不再固定，而是随机波动
     """

minamo/model/topo.py

@@ -1,6 +1,7 @@
 import torch
 import torch.nn as nn
 import torch.nn.functional as F
+from torch.nn.utils import spectral_norm
 from torch_geometric.nn import global_mean_pool, TopKPooling, GATConv
 from torch_geometric.data import Data
 
@@ -18,6 +19,12 @@ class MinamoTopoModel(nn.Module):
         self.conv_ins1 = GATConv(hidden_dim*16, hidden_dim*8, heads=2)
         self.conv3 = GATConv(hidden_dim*16, out_dim, concat=False)
         
+        self.conv1.lin = spectral_norm(self.conv1.lin)
+        self.conv2.lin = spectral_norm(self.conv2.lin)
+        self.conv_ins2.lin = spectral_norm(self.conv_ins2.lin)
+        self.conv_ins1.lin = spectral_norm(self.conv_ins1.lin)
+        self.conv3.lin = spectral_norm(self.conv3.lin)
+        
         # 正则化
         self.norm1 = nn.LayerNorm(hidden_dim*16)
         self.norm2 = nn.LayerNorm(hidden_dim*16)

minamo/model/vision.py

@@ -1,6 +1,7 @@
 import torch
 import torch.nn as nn
 import torch.nn.functional as F
+from torch.nn.utils import spectral_norm
 from shared.attention import CBAM
 
 class MinamoVisionModel(nn.Module):
@@ -11,26 +12,26 @@ class MinamoVisionModel(nn.Module):
         
         # 卷积部分
         self.vision_conv = nn.Sequential(
-            nn.Conv2d(conv_ch, conv_ch*2, 3, padding=1),
+            spectral_norm(nn.Conv2d(conv_ch, conv_ch*2, 3, padding=1)),
             nn.BatchNorm2d(conv_ch*2),
             CBAM(conv_ch*2),
             nn.GELU(),
             nn.MaxPool2d(2),
             nn.Dropout2d(0.4),
             
-            nn.Conv2d(conv_ch*2, conv_ch*4, 3, padding=1),
+            spectral_norm(nn.Conv2d(conv_ch*2, conv_ch*4, 3, padding=1)),
             nn.BatchNorm2d(conv_ch*4),
             CBAM(conv_ch*4),
             nn.GELU(),
             nn.MaxPool2d(2),
             nn.Dropout2d(0.4),
             
-            nn.Conv2d(conv_ch*4, conv_ch*8, 3, padding=1),
+            spectral_norm(nn.Conv2d(conv_ch*4, conv_ch*8, 3, padding=1)),
             nn.BatchNorm2d(conv_ch*8),
             CBAM(conv_ch*8),
             nn.GELU(),
             
-            nn.Conv2d(conv_ch*8, conv_ch*8, 3, padding=1),
+            spectral_norm(nn.Conv2d(conv_ch*8, conv_ch*8, 3, padding=1)),
             nn.BatchNorm2d(conv_ch*8),
             CBAM(conv_ch*8),
             nn.GELU(),

minamo/train.py

@@ -56,7 +56,7 @@ def train():
     
     # 设定优化器与调度器
     optimizer = optim.AdamW(model.parameters(), lr=1e-3, weight_decay=1e-2)
-    scheduler = optim.lr_scheduler.CosineAnnealingWarmRestarts(optimizer, T_0=10, T_mult=2, eta_min=1e-6)
+    scheduler = optim.lr_scheduler.CosineAnnealingWarmRestarts(optimizer, T_0=3, T_mult=2, eta_min=1e-6)
     criterion = MinamoLoss()
     
     if args.resume:
@@ -126,7 +126,7 @@ def train():
         scheduler.step()
         
         # 每十轮推理一次验证集
-        if (epoch + 1) % 5 == 0:
+        if (epoch + 1) % 1 == 0:
             model.eval()
             val_loss = 0
             with torch.no_grad():