Python根据给定模型计算权值

pip install tensorflow

import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
import numpy as np
 
# 定义一个简单的顺序模型
model = Sequential([
    Dense(64, activation='relu', input_shape=(784,)),  # 假设输入是784维的（例如，28x28的图像展平）
    Dense(10, activation='softmax')  # 假设有10个输出类别（例如，MNIST数据集）
])
 
# 编译模型（虽然在这个例子中我们不会训练它）
model.compile(optimizer='adam',
              loss='sparse_categorical_crossentropy',
              metrics=['accuracy'])
 
# 假设我们有一些训练数据（这里我们不会真正使用它们进行训练）
# X_train = np.random.rand(60000, 784)  # 60000个样本，每个样本784维
# y_train = np.random.randint(10, size=(60000,))  # 60000个标签，每个标签是0到9之间的整数
 
# 初始化模型权值（在实际应用中，我们会通过训练来更新这些权值）
model.build((None, 784))  # 这将基于input_shape创建模型的权重
 
# 提取并打印模型的权值
for layer in model.layers:
    # 获取层的权值
    weights, biases = layer.get_weights()
    
    # 打印权值的形状和值（这里我们只打印形状和权值的前几个元素以避免输出过长）
    print(f"Layer: {layer.name}")
    print(f"  Weights shape: {weights.shape}")
    print(f"  Weights (first 5 elements): {weights[:5]}")  # 只打印前5个元素作为示例
    print(f"  Biases shape: {biases.shape}")
    print(f"  Biases (first 5 elements): {biases[:5]}")  # 只打印前5个元素作为示例
    print("\n")
 
# 注意：在实际应用中，我们会通过调用model.fit()来训练模型，训练后权值会被更新。
# 例如：model.fit(X_train, y_train, epochs=5)
 
# 由于我们没有真正的训练数据，也没有进行训练，所以上面的权值是随机初始化的。

pip install torch torchvision

import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import DataLoader, TensorDataset
 
# 定义一个简单的神经网络
class SimpleNN(nn.Module):
    def __init__(self, input_size, hidden_size, output_size):
        super(SimpleNN, self).__init__()
        self.fc1 = nn.Linear(input_size, hidden_size)
        self.relu = nn.ReLU()
        self.fc2 = nn.Linear(hidden_size, output_size)
 
    def forward(self, x):
        out = self.fc1(x)
        out = self.relu(out)
        out = self.fc2(out)
        return out
 
# 生成一些示例数据
input_size = 10
hidden_size = 5
output_size = 1
num_samples = 100
 
X = torch.randn(num_samples, input_size)
y = torch.randn(num_samples, output_size)
 
# 创建数据加载器
dataset = TensorDataset(X, y)
dataloader = DataLoader(dataset, batch_size=10, shuffle=True)
 
# 初始化模型、损失函数和优化器
model = SimpleNN(input_size, hidden_size, output_size)
criterion = nn.MSELoss()
optimizer = optim.SGD(model.parameters(), lr=0.01)
 
# 训练模型
num_epochs = 10
for epoch in range(num_epochs):
    for inputs, targets in dataloader:
        optimizer.zero_grad()
        outputs = model(inputs)
        loss = criterion(outputs, targets)
        loss.backward()
        optimizer.step()
    print(f'Epoch [{epoch+1}/{num_epochs}], Loss: {loss.item():.4f}')
 
# 保存模型（可选）
torch.save(model.state_dict(), 'simple_nn_model.pth')

# 加载模型（如果保存了）
# model = SimpleNN(input_size, hidden_size, output_size)
# model.load_state_dict(torch.load('simple_nn_model.pth'))
 
# 提取权值
for name, param in model.named_parameters():
    if param.requires_grad:
        print(f"Parameter name: {name}")
        print(f"Shape: {param.shape}")
        print(f"Values: {param.data.numpy()}\n")

import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import DataLoader, TensorDataset
 
# 定义一个简单的神经网络
class SimpleNN(nn.Module):
    def __init__(self, input_size, hidden_size, output_size):
        super(SimpleNN, self).__init__()
        self.fc1 = nn.Linear(input_size, hidden_size)
        self.relu = nn.ReLU()
        self.fc2 = nn.Linear(hidden_size, output_size)
 
    def forward(self, x):
        out = self.fc1(x)
        out = self.relu(out)
        out = self.fc2(out)
        return out
 
# 生成一些示例数据
input_size = 10
hidden_size = 5
output_size = 1
num_samples = 100
 
X = torch.randn(num_samples, input_size)
y = torch.randn(num_samples, output_size)
 
# 创建数据加载器
dataset = TensorDataset(X, y)
dataloader = DataLoader(dataset, batch_size=10, shuffle=True)
 
# 初始化模型、损失函数和优化器
model = SimpleNN(input_size, hidden_size, output_size)
criterion = nn.MSELoss()
optimizer = optim.SGD(model.parameters(), lr=0.01)
 
# 训练模型
num_epochs = 10
for epoch in range(num_epochs):
    for inputs, targets in dataloader:
        optimizer.zero_grad()
        outputs = model(inputs)
        loss = criterion(outputs, targets)
        loss.backward()
        optimizer.step()
    print(f'Epoch [{epoch+1}/{num_epochs}], Loss: {loss.item():.4f}')
 
# 保存模型（可选）
# torch.save(model.state_dict(), 'simple_nn_model.pth')
 
# 提取权值
for name, param in model.named_parameters():
    if param.requires_grad:
        print(f"Parameter name: {name}")
        print(f"Shape: {param.shape}")
        print(f"Values: {param.data.numpy()}\n")

import torch
import torch.nn as nn
 
# 假设我们有一个已定义的模型架构，这里我们再次定义它以确保一致性
class MyModel(nn.Module):
    def __init__(self):
        super(MyModel, self).__init__()
        self.layer1 = nn.Linear(10, 50)  # 假设输入特征为10，隐藏层单元为50
        self.layer2 = nn.Linear(50, 1)   # 假设输出特征为1
 
    def forward(self, x):
        x = torch.relu(self.layer1(x))
        x = self.layer2(x)
        return x
 
# 实例化模型
model = MyModel()
 
# 加载已保存的状态字典（假设模型保存在'model.pth'文件中）
model_path = 'model.pth'
model.load_state_dict(torch.load(model_path))
 
# 将模型设置为评估模式（对于推理是必需的，但对于提取权值不是必需的）
model.eval()
 
# 提取权值
for name, param in model.named_parameters():
    print(f"Parameter name: {name}")
    print(f"Shape: {param.shape}")
    print(f"Values: {param.data.numpy()}\n")
 
# 注意：如果我们只想保存权值而不是整个模型，我们可以在训练完成后只保存状态字典
# torch.save(model.state_dict(), 'model_weights.pth')
# 然后在需要时加载它们
# model = MyModel()
# model.load_state_dict(torch.load('model_weights.pth'))