@[Daniel](Pytorch初学者实例教程)
# Pytorch初学者实例教程
大家好，这里将通过一个具体的实战项目，帮助大家迅速了解如何使用Pytorch来构建自己的神经网络架构。本项目的数据取自"深度学习原理与Pytorch实战"第三章，代码也参考其书中的代码，但是做了大量的解释和少量的变动。同时封装了**方法类**，训练好了**模型**。不是有很多硬件资源的同学们可以**直接调用模型来预测**。
## 数据集的简介
数据集使用的是国外一家共享单车开源数据集，为了避免不必要的麻烦，我把[数据集](https://github.com/TangxzNLP/dispatch_pytorch_demo/tree/master/dispatch_pytorch_demo)放在了我的github仓库中，欢迎大家点击和star。更多关于数据集的简介放在了[dispatch_self_function.py](https://github.com/TangxzNLP/dispatch_pytorch_demo/blob/master/dispatch_pytorch_demo/dispatch_self_function.py)文件中。
## 主要代码简介
1. [dispatch_self_function.py](https://github.com/TangxzNLP/dispatch_pytorch_demo/blob/master/dispatch_pytorch_demo/dispatch_self_function.py)是训练模型的代码，github上富有非常详细的说明，这里只放一部分源码。
```python
a, 手写用Tensor运算的神经网路
已知的数据：X.shape[]=(16875, 56), Y.shape = (16875, 1)
构建网络的参数：
input_size = X.shape[1],也就是一组特征向量的纬度，这里是56.
hidden_size = 10, 设置10隐含节点
output_size = 1, 1个输出节点
batch_size = 128, 设置批处理大小为 128组向量
则(三层神经网络：输入，隐含，输出)：
weights.shape = (56, 10)
biases.shape = (,10)
weights2.shape = (10, 1)
"""
input_size = X.shape[1]
hidden_size = 10
output_size = 1
batch_size = 128
weights = torch.randn([input_size, hidden_size], dtype = torch.double, requires_grad = True)
biases = torch.randn([hidden_size], dtype = torch.double, requires_grad = True)
weights2 = torch.randn([hidden_size, output_size], dtype = torch.double, requires_grad = True)
def neu(x):
hidden = x.mm(weights) + biases.expand(x.size()[0], hidden_size)
hidden = torch.sigmoid(hidden)
output = hidden.mm(weights2)
return output
def cost(x, y):
error = torch.mean((x-y)**2)
return error
def optimizer_step(learning_rate):
weights.data.add_(-learning_rate * weights.grad.data)
biases.data.add_(-learning_rate * biases.grad.data)
weights2.data.add_(-learning_rate * weights2.grad.data)
def zero_grad():
if weights.grad is not None and biases.grad is not None and weights2.grad is not None:
weights.grad.data.zero_()
biases.grad.data.zero_()
weights2.grad.data.zero_()
# 神经网络的训练
losses = []
```
2. 模型评估[dispatch_evaluate.py](https://github.com/TangxzNLP/dispatch_pytorch_demo/blob/master/dispatch_pytorch_demo/dispatch_evaluate.py)
```python
# 用训练好的神经网络在测试集上进行预测
targets = test_targets['cnt']
targets = targets.values.reshape([len(targets), 1])
targets = targets.astype('float')
x = torch.tensor(test_features.values, dtype = torch.double, requires_grad = True)
y = torch.tensor(targets, dtype = torch.double, requires_grad = True)
# 用神经网络进行预测
predict = neu(x)
predict = predict.data.numpy()
print(predict)
```
3. 模型[model.pkl](https://github.com/TangxzNLP/dispatch_pytorch_demo/blob/master/dispatch_pytorch_demo/model.pkl)
4. 代码方法类[model_function.py](https://github.com/TangxzNLP/dispatch_pytorch_demo/blob/master/dispatch_pytorch_demo/model_function.py)
## 代码网址
[https://github.com/TangxzNLP/dispatch_pytorch_demo/](https://github.com/TangxzNLP/dispatch_pytorch_demo/)帮助到您，可以帮我点个star.欢迎留言交流讨论。
网友评论