feat: 为 transformer 加入起始与结束 token

data/shell.sh

@@ -0,0 +1,4 @@
+# 自动处理塔信息
+pnpm auto "result.json" "F:/mota-ai/total data/towerinfo.json" "F:\mota-ai\total data/games"
+# 将数据按比例区分为训练集和数据集
+pnpm eval "ginka-dataset.json" "ginka-eval.json" "result.json" 0.02

data/src/auto.ts

@@ -240,17 +240,17 @@ const labelConfig: IAutoLabelConfig = {
     classes: {
         empty: 0,
         wall: 1,
-        decoration: 2,
-        commonDoors: [3],
-        specialDoors: [6, 6],
-        keys: [7],
-        redGems: [10],
-        blueGems: [13],
-        greenGems: [16],
-        potions: [19],
-        items: [23],
-        enemies: [26],
-        entry: 29
+        decoration: 16,
+        commonDoors: [2],
+        specialDoors: [2, 2],
+        keys: [3],
+        redGems: [4],
+        blueGems: [5],
+        greenGems: [6],
+        potions: [7],
+        items: [8],
+        enemies: [9],
+        entry: 10
     },
     allowedSize: [[13, 13]],
     allowUselessBranch: true,

data/src/shared.ts

@@ -1,17 +1,17 @@
 // 基本图块定义
 export const emptyTiles = new Set([0]);
 export const wallTiles = new Set([1]);
-export const decorationTiles = new Set([2]);
-export const commonDoorTiles = new Set([3]);
-export const specialDoorTiles = new Set([6]);
-export const keyTiles = new Set([7]);
-export const redGemTiles = new Set([10]);
-export const blueGemTiles = new Set([13]);
-export const greenGemTiles = new Set([16]);
-export const potionTiles = new Set([19]);
-export const itemTiles = new Set([23]);
-export const enemyTiles = new Set([26]);
-export const entryTiles = new Set([29]);
+export const decorationTiles = new Set([16]);
+export const commonDoorTiles = new Set([2]);
+export const specialDoorTiles = new Set([2]);
+export const keyTiles = new Set([3]);
+export const redGemTiles = new Set([4]);
+export const blueGemTiles = new Set([5]);
+export const greenGemTiles = new Set([6]);
+export const potionTiles = new Set([7]);
+export const itemTiles = new Set([8]);
+export const enemyTiles = new Set([9]);
+export const entryTiles = new Set([10]);
 
 // 组合图块定义
 export const doorTiles = commonDoorTiles.union(specialDoorTiles);

ginka/train_transformer_vae.py

@@ -42,16 +42,8 @@ from shared.image import matrix_to_image_cv
 # 12. 咸鱼门数量（多层咸鱼门只算一个）
 
 # 图块定义：
-# 0. 空地, 1. 墙壁, 2. 装饰（用于野外装饰，视为空地）, 
-# 3. 黄门, 4. 蓝门, 5. 红门, 6. 机关门, 其余种类的门如绿门都视为红门
-# 7-9. 黄蓝红门钥匙，机关门不使用钥匙开启
-# 10-12. 三种等级的红宝石
-# 13-15. 三种等级的蓝宝石
-# 16-18. 三种等级的绿宝石
-# 19-22. 四种等级的血瓶
-# 23-25. 三种等级的道具
-# 26-28. 三种等级的怪物
-# 29. 入口，不区分楼梯和箭头
+# 0. 空地, 1. 墙壁, 2. 门, 3. 钥匙, 4. 红宝石, 5. 蓝宝石, 6. 绿宝石, 7. 血瓶
+# 8. 道具, 9. 怪物, 10. 入口, 14. 起始 token, 15. 终止 token
 
 BATCH_SIZE = 8
 LATENT_DIM = 48
@@ -60,6 +52,7 @@ SELF_GATE = 0.5
 GATE_EPOCH = 5
 VAL_BATCH_DIVIDER = 8
 PROB_STEP = 0.05
+NUM_CLASSES = 16
 
 device = torch.device(
     "cuda:1" if torch.cuda.is_available()
@@ -89,7 +82,7 @@ def train():
     
     args = parse_arguments()
     
-    vae = GinkaTransformerVAE(latent_dim=LATENT_DIM).to(device)
+    vae = GinkaTransformerVAE(num_classes=NUM_CLASSES, latent_dim=LATENT_DIM).to(device)
     
     dataset = GinkaRNNDataset(args.train, device)
     dataset_val = GinkaRNNDataset(args.validate, device)

ginka/transformer_vae/decoder.py

@@ -10,7 +10,7 @@ class GinkaTransformerDecoder(nn.Module):
         self.dim_ff = dim_ff
         self.map_size = map_size
         self.embedding = nn.Embedding(num_classes, dim_ff)
-        self.pos_embedding = nn.Embedding(map_size, dim_ff)
+        self.pos_embedding = nn.Embedding(map_size + 1, dim_ff)
         self.encoder = nn.TransformerEncoder(
             nn.TransformerEncoderLayer(d_model=dim_ff, dim_feedforward=dim_ff, nhead=nhead, batch_first=True),
             num_layers=max(num_layers // 2, 1)
@@ -31,12 +31,14 @@ class GinkaTransformerDecoder(nn.Module):
         B, L = target_map.shape
         
         memory = self.encoder(z.unsqueeze(1)) # [B, 1, dim_ff]
-        mask = torch.triu(torch.ones(L, L, dtype=torch.bool)).to(z.device) # [B, H * W, H * W]
+        mask = torch.triu(torch.ones(L + 1, L + 1, dtype=torch.bool)).to(z.device) # [B, H * W, H * W]
         
         # when training, use teacher forcing
         if not self.autoregressive:
-            map = self.embedding(target_map)
-            pos_embed = self.pos_embedding(torch.arange(L, dtype=torch.long).to(z.device))
+            first_token = torch.tensor([31], dtype=torch.long).to(z.device).repeat(B, 1)
+            with_first = torch.cat([first_token, target_map], dim=1)
+            map = self.embedding(with_first)
+            pos_embed = self.pos_embedding(torch.arange(L + 1, dtype=torch.long).to(z.device))
             map = map + pos_embed # [B, H * W, dim_ff]
             decoded = self.decoder(map, memory, tgt_mask=mask) # [B, H * W, dim_ff]
             output = self.fc(decoded)
@@ -44,7 +46,7 @@ class GinkaTransformerDecoder(nn.Module):
         
         # when evaling, use autoregressive generation
         else:
-            output = torch.zeros([B, L], dtype=torch.int).to(z.device)
+            output = torch.zeros([B, L + 1], dtype=torch.int).to(z.device)
             for idx in range(0, self.map_size):
                 embed = self.embedding(output)
                 pos_embed = self.pos_embedding(torch.IntTensor([idx]).repeat(B, 1).to(z.device))
@@ -77,7 +79,7 @@ class GinkaTransformerVAEDecoder(nn.Module):
     def forward(self, z: torch.Tensor, map: torch.Tensor):
         hidden = self.input(z)
         output = self.decoder(hidden, map)
-        return output[:, 0:self.map_size]
+        return output
 
 if __name__ == "__main__":
     device = torch.device("cpu")
@@ -87,7 +89,6 @@ if __name__ == "__main__":
     
     # 初始化模型
     model = GinkaTransformerVAEDecoder().to(device)
-    model.eval()
     
     print_memory("初始化后")
     

ginka/vae_rnn/loss.py

@@ -7,6 +7,8 @@ class VAELoss:
     
     def vae_loss(self, logits, target, mu, logvar, beta=0.1):
         # target: [B, 169]
+        end_token = torch.tensor([15], dtype=torch.long).to(logits.device)
+        target = torch.cat([target, end_token], dim=1)
         target = F.one_hot(target, num_classes=self.num_classes).float()
         recon_loss = F.cross_entropy(logits, target)