完成实验2

Lab2/code/1.1.py

@@ -125,7 +125,7 @@ if __name__ == "__main__":
         avg_epoch_acc = total_epoch_acc / len(test_regression_dataset)
         print(
             f"Epoch [{epoch + 1}/{num_epochs}],",
-            f"Train Loss: {total_epoch_loss},",
+            f"Train Loss: {total_epoch_loss:.10f},",
             f"Used Time: {train_time * 1000:.3f}ms,",
             f"Test Acc: {avg_epoch_acc * 100:.3f}%,",
             f"Used Time: {test_time * 1000:.3f}ms",

Lab2/code/1.2.py

@@ -130,7 +130,7 @@ if __name__ == "__main__":
         avg_epoch_acc = total_epoch_acc / len(test_binarycls_dataset)
         print(
             f"Epoch [{epoch + 1}/{num_epochs}],",
-            f"Train Loss: {total_epoch_loss},",
+            f"Train Loss: {total_epoch_loss:.10f},",
             f"Used Time: {train_time * 1000:.3f}ms,",
             f"Test Acc: {avg_epoch_acc * 100:.3f}%,",
             f"Used Time: {test_time * 1000:.3f}ms",

Lab2/code/1.3.py

@@ -118,7 +118,7 @@ if __name__ == "__main__":
         avg_epoch_acc = total_epoch_acc / len(test_mnist_dataset)
         print(
             f"Epoch [{epoch + 1}/{num_epochs}],",
-            f"Train Loss: {total_epoch_loss},",
+            f"Train Loss: {total_epoch_loss:.10f},",
             f"Used Time: {train_time * 1000:.3f}ms,",
             f"Test Acc: {avg_epoch_acc * 100:.3f}%,",
             f"Used Time: {test_time * 1000:.3f}ms",

Lab2/code/2.1.py

@@ -107,7 +107,7 @@ if __name__ == "__main__":
         avg_epoch_acc = total_epoch_acc / len(test_regression_dataset)
         print(
             f"Epoch [{epoch + 1}/{num_epochs}],",
-            f"Train Loss: {total_epoch_loss},",
+            f"Train Loss: {total_epoch_loss:.10f},",
             f"Used Time: {train_time * 1000:.3f}ms,",
             f"Test Acc: {avg_epoch_acc * 100:.3f}%,",
             f"Used Time: {test_time * 1000:.3f}ms",

Lab2/code/2.2.py

@@ -111,7 +111,7 @@ if __name__ == "__main__":
         avg_epoch_acc = total_epoch_acc / len(test_binarycls_dataset)
         print(
             f"Epoch [{epoch + 1}/{num_epochs}],",
-            f"Train Loss: {total_epoch_loss},",
+            f"Train Loss: {total_epoch_loss:.10f},",
             f"Used Time: {train_time * 1000:.3f}ms,",
             f"Test Acc: {avg_epoch_acc * 100:.3f}%,",
             f"Used Time: {test_time * 1000:.3f}ms",

Lab2/code/2.3.py

@@ -99,7 +99,7 @@ if __name__ == "__main__":
         avg_epoch_acc = total_epoch_acc / len(test_mnist_dataset)
         print(
             f"Epoch [{epoch + 1}/{num_epochs}],",
-            f"Train Loss: {total_epoch_loss},",
+            f"Train Loss: {total_epoch_loss:.10f},",
             f"Used Time: {train_time * 1000:.3f}ms,",
             f"Test Acc: {avg_epoch_acc * 100:.3f}%,",
             f"Used Time: {test_time * 1000:.3f}ms",

Lab2/code/3.py

@@ -74,9 +74,37 @@ class Model_3_3(nn.Module):
         x = self.fc3(x)
         x = self.activate_fn(x)
         return x
+    
+
+class Model_3_4(nn.Module):
+    def __init__(self, num_classes):
+        super().__init__()
+        self.flatten = nn.Flatten()
+        self.fc1 = nn.Linear(in_features=28 * 28, out_features=1024)
+        self.fc2 = nn.Linear(in_features=1024, out_features=256)
+        self.fc3 = nn.Linear(in_features=256, out_features=num_classes)
+        self.activate_fn = leaky_relu
+
+    def forward(self, x: torch.Tensor):
+        x = self.flatten(x)
+        x = self.fc1(x)
+        x = self.activate_fn(x)
+
+        x = self.fc2(x)
+        x = self.activate_fn(x)
+
+        x = self.fc3(x)
+        x = self.activate_fn(x)
+        return x
 
 
 if __name__ == "__main__":
-    train_MNIST_CLS(Model=Model_3_1)
+    print("模型1开始训练，激活函数为relu：")
-    train_MNIST_CLS(Model=Model_3_2)
+    train_loss_3_1, test_acc_3_1 = train_MNIST_CLS(Model=Model_3_1) # 激活函数为relu
-    train_MNIST_CLS(Model=Model_3_3)
+    print("模型2开始训练，激活函数为sigmoid：")
+    train_loss_3_2, test_acc_3_2 = train_MNIST_CLS(Model=Model_3_2) # 激活函数为sigmoid
+    print("模型3开始训练，激活函数为tanh：")
+    train_loss_3_3, test_acc_3_3 = train_MNIST_CLS(Model=Model_3_3) # 激活函数为tanh
+    print("模型4开始训练，激活函数为leaky_relu：")
+    train_loss_3_4, test_acc_3_4 = train_MNIST_CLS(Model=Model_3_4) # 激活函数为leaky_relu
+

Lab2/code/4.py

@@ -0,0 +1,102 @@
+import time
+import numpy as np
+import torch
+from torch.nn.functional import *
+from torch.utils.data import Dataset, DataLoader
+from torch import nn
+from torchvision import datasets, transforms
+from tqdm import tqdm
+from utils import *
+
+import ipdb
+
+class Model_4_1(nn.Module):
+    def __init__(self, num_classes):
+        super().__init__()
+        self.flatten = nn.Flatten()
+        self.fc1 = nn.Linear(in_features=28 * 28, out_features=512)
+        self.fc2 = nn.Linear(in_features=512, out_features=num_classes)
+        self.activate_fn = leaky_relu
+
+    def forward(self, x: torch.Tensor):
+        x = self.flatten(x)
+        x = self.fc1(x)
+        x = self.activate_fn(x)
+
+        x = self.fc2(x)
+        x = self.activate_fn(x)
+        return x
+    
+
+class Model_4_2(nn.Module):
+    def __init__(self, num_classes):
+        super().__init__()
+        self.flatten = nn.Flatten()
+        self.fc1 = nn.Linear(in_features=28 * 28, out_features=1024)
+        self.fc2 = nn.Linear(in_features=1024, out_features=num_classes)
+        self.activate_fn = leaky_relu
+
+    def forward(self, x: torch.Tensor):
+        x = self.flatten(x)
+        x = self.fc1(x)
+        x = self.activate_fn(x)
+
+        x = self.fc2(x)
+        x = self.activate_fn(x)
+        return x
+    
+
+class Model_4_3(nn.Module):
+    def __init__(self, num_classes):
+        super().__init__()
+        self.flatten = nn.Flatten()
+        self.fc1 = nn.Linear(in_features=28 * 28, out_features=512)
+        self.fc2 = nn.Linear(in_features=512, out_features=512)
+        self.fc3 = nn.Linear(in_features=512, out_features=num_classes)
+        self.activate_fn = leaky_relu
+
+    def forward(self, x: torch.Tensor):
+        x = self.flatten(x)
+        x = self.fc1(x)
+        x = self.activate_fn(x)
+
+        x = self.fc2(x)
+        x = self.activate_fn(x)
+
+        x = self.fc3(x)
+        x = self.activate_fn(x)
+        return x
+    
+
+class Model_4_4(nn.Module):
+    def __init__(self, num_classes):
+        super().__init__()
+        self.flatten = nn.Flatten()
+        self.fc1 = nn.Linear(in_features=28 * 28, out_features=1024)
+        self.fc2 = nn.Linear(in_features=1024, out_features=1024)
+        self.fc3 = nn.Linear(in_features=1024, out_features=num_classes)
+        self.activate_fn = leaky_relu
+
+    def forward(self, x: torch.Tensor):
+        x = self.flatten(x)
+        x = self.fc1(x)
+        x = self.activate_fn(x)
+
+        x = self.fc2(x)
+        x = self.activate_fn(x)
+
+        x = self.fc3(x)
+        x = self.activate_fn(x)
+        return x
+
+
+if __name__ == "__main__":
+    print("模型1开始训练，hidden_size=512，hidden_layer=1 ：")
+    train_loss_4_1, test_acc_4_1 = train_MNIST_CLS(Model=Model_4_1) # hidden_size=512, hidden_layer=1
+    print("模型2开始训练，hidden_size=1024，hidden_layer=1 ：")
+    train_loss_4_2, test_acc_4_2 = train_MNIST_CLS(Model=Model_4_2) # hidden_size=1024, hidden_layer=1
+    print("模型3开始训练，hidden_size=512，hidden_layer=2 ：")
+    train_loss_4_3, test_acc_4_3 = train_MNIST_CLS(Model=Model_4_3) # hidden_size=512, hidden_layer=2
+    print("模型4开始训练，hidden_size=1024，hidden_layer=2 ：")
+    train_loss_4_4, test_acc_4_4 = train_MNIST_CLS(Model=Model_4_4) # hidden_size=1024, hidden_layer=2
+

Lab2/code/utils.py

@@ -106,7 +106,7 @@ class My_optimizer:
 
 
 def train_MNIST_CLS(Model:nn.Module):
-    learning_rate = 5e-2
+    learning_rate = 8e-2
     num_epochs = 10
     batch_size = 512
     num_classes = 10
@@ -139,6 +139,8 @@ def train_MNIST_CLS(Model:nn.Module):
     criterion = nn.CrossEntropyLoss()
     optimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)
     
+    train_loss = list()
+    test_acc = list()
     for epoch in range(num_epochs):
         model.train()
         total_epoch_loss = 0
@@ -182,8 +184,11 @@ def train_MNIST_CLS(Model:nn.Module):
         avg_epoch_acc = total_epoch_acc / len(test_mnist_dataset)
         print(
             f"Epoch [{epoch + 1}/{num_epochs}],",
-            f"Train Loss: {total_epoch_loss},",
+            f"Train Loss: {total_epoch_loss:.10f},",
             f"Used Time: {train_time * 1000:.3f}ms,",
             f"Test Acc: {avg_epoch_acc * 100:.3f}%,",
             f"Used Time: {test_time * 1000:.3f}ms",
-        )
+        )
+        train_loss.append(total_epoch_loss)
+        test_acc.append(avg_epoch_acc * 100)
+    return train_loss, test_acc