fix

Image/run_all_models.py

@@ -1,3 +1,4 @@
+#bug
 import os
 import subprocess
 from pathlib import Path

Image/utils/dataset_utils.py

@@ -3,7 +3,7 @@ import torchvision
 import torchvision.transforms as transforms
 import os
 
-def get_cifar10_dataloaders(batch_size=128, num_workers=2, local_dataset_path=None):
+def get_cifar10_dataloaders(batch_size=128, num_workers=2, local_dataset_path=None,shuffle=True):
     """获取CIFAR10数据集的数据加载器
     
     Args:
@@ -45,16 +45,16 @@ def get_cifar10_dataloaders(batch_size=128, num_workers=2, local_dataset_path=No
     trainset = torchvision.datasets.CIFAR10(
         root=dataset_path, train=True, download=download, transform=transform_train)
     trainloader = torch.utils.data.DataLoader(
-        trainset, batch_size=batch_size, shuffle=True, num_workers=num_workers)
+        trainset, batch_size=batch_size, shuffle=shuffle, num_workers=num_workers)
 
     testset = torchvision.datasets.CIFAR10(
         root=dataset_path, train=False, download=download, transform=transform_test)
     testloader = torch.utils.data.DataLoader(
-        testset, batch_size=100, shuffle=False, num_workers=num_workers)
+        testset, batch_size=batch_size, shuffle=shuffle, num_workers=num_workers)
 
     return trainloader, testloader
 
-def get_mnist_dataloaders(batch_size=128, num_workers=2, local_dataset_path=None):
+def get_mnist_dataloaders(batch_size=128, num_workers=2, local_dataset_path=None,shuffle=True):
     """获取MNIST数据集的数据加载器
     
     Args:
@@ -100,11 +100,11 @@ def get_mnist_dataloaders(batch_size=128, num_workers=2, local_dataset_path=None
     trainset = torchvision.datasets.MNIST(
         root=dataset_path, train=True, download=download, transform=transform_train)
     trainloader = torch.utils.data.DataLoader(
-        trainset, batch_size=batch_size, shuffle=True, num_workers=num_workers)
+        trainset, batch_size=batch_size, shuffle=shuffle, num_workers=num_workers)
 
     testset = torchvision.datasets.MNIST(
         root=dataset_path, train=False, download=download, transform=transform_test)
     testloader = torch.utils.data.DataLoader(
-        testset, batch_size=100, shuffle=False, num_workers=num_workers)
+        testset, batch_size=batch_size, shuffle=shuffle, num_workers=num_workers)
 
     return trainloader, testloader

Image/utils/train_utils.py

@@ -92,9 +92,10 @@ def collect_embeddings(model, dataloader, device):
         inputs = inputs.to(device)
         _ = model(inputs)
         
-        # 找到维度最大的层
-        max_dim = 0
-        max_layer_name = None
+        # 找到维度在512-1024范围内的层
+        target_dim_range = (512, 1024)
+        suitable_layer_name = None
+        suitable_dim = None
         
         # 分析所有层的输出维度
         for name, feat in activation.items():
@@ -102,9 +103,15 @@ def collect_embeddings(model, dataloader, device):
                 continue
             # 计算展平后的维度
             flat_dim = feat.numel() // feat.shape[0]  # 每个样本的特征维度
-            if flat_dim > max_dim:
-                max_dim = flat_dim
-                max_layer_name = name
+            if target_dim_range[0] <= flat_dim <= target_dim_range[1]:
+                suitable_layer_name = name
+                suitable_dim = flat_dim
+                break  # 找到第一个符合条件的层就停止
+        
+        if suitable_layer_name is None:
+            raise ValueError(f"没有找到维度在{target_dim_range[0]}-{target_dim_range[1]}范围内的层")
+            
+        print(f"选择的特征层: {suitable_layer_name}, 特征维度: {suitable_dim}")
         
         # 清除第一次运行的激活值
         activation.clear()
@@ -115,7 +122,7 @@ def collect_embeddings(model, dataloader, device):
             _ = model(inputs)
             
             # 获取并处理特征
-            features = activation[max_layer_name]
+            features = activation[suitable_layer_name]
             flat_features = torch.flatten(features, start_dim=1)
             embeddings.append(flat_features.cpu().numpy())
             indices.extend(range(batch_idx * dataloader.batch_size,
@@ -271,8 +278,17 @@ def train_model(model, trainloader, testloader, epochs=200, lr=0.1, device='cuda
             model_path = os.path.join(epoch_dir, 'subject_model.pth')
             torch.save(model.state_dict(), model_path)
             
-            # 收集并保存嵌入向量
-            embeddings, indices = collect_embeddings(model, trainloader, device)
+            # 创建一个专门用于收集embedding的顺序dataloader
+            ordered_loader = torch.utils.data.DataLoader(
+                trainloader.dataset,  # 使用相同的数据集
+                batch_size=trainloader.batch_size,
+                shuffle=False,  # 确保顺序加载
+                num_workers=trainloader.num_workers
+            )
+            
+            # 收集并保存嵌入向量，使用顺序加载的dataloader
+            embeddings, indices = collect_embeddings(model, ordered_loader, device)
+            
             # 保存嵌入向量
             np.save(os.path.join(epoch_dir, 'train_data.npy'), embeddings)