Hands on deep learning (III)

 

 

 

 

  Manual implementation

%matplotlib inline
import random
import torch
from d2l import torch as d2l
def synthetic_data(w, b, num_examples):  #@save
    """ Generate y=Xw+b+ noise """
    X = torch.normal(0, 1, (num_examples, len(w)))
    y = torch.matmul(X, w) + b
    y += torch.normal(0, 0.01, y.shape)
    return X, y.reshape((-1, 1))
def data_iter(batch_size, features, labels):
    num_examples = len(features)
    indices = list(range(num_examples))
    #  These samples are read at random , There is no specific order 
    random.shuffle(indices)
    for i in range(0, num_examples, batch_size):
        batch_indices = torch.tensor(
            indices[i: min(i + batch_size, num_examples)])
        yield features[batch_indices], labels[batch_indices]
def linreg(X, w, b):  #@save
    """ linear regression model """
    return torch.matmul(X, w) + b
def squared_loss(y_hat, y):  #@save
    """ Mean square loss """
    return (y_hat - y.reshape(y_hat.shape)) ** 2 / 2
def sgd(params, lr, batch_size):  #@save
    """ Small batch random gradient drop """
    with torch.no_grad():
        for param in params:
            param -= lr * param.grad / batch_size
            param.grad.zero_()

true_w = torch.tensor([2, -3.4])
true_b = 4.2
features, labels = synthetic_data(true_w, true_b, 1000)
batch_size = 10

for X, y in data_iter(batch_size, features, labels):
    print(X, '\n', y)
    break
w = torch.normal(0, 0.01, size=(2,1), requires_grad=True)
b = torch.zeros(1, requires_grad=True)
lr = 0.03
num_epochs = 3
net = linreg
loss = squared_loss

for epoch in range(num_epochs):
    for X, y in data_iter(batch_size, features, labels):
        l = loss(net(X, w, b), y)  # X and y Small batch loss 
        #  because l The shape is (batch_size,1), Instead of a scalar .l All elements in are added together ,
        #  And calculate about [w,b] Gradient of 
        l.sum().backward()
        sgd([w, b], lr, batch_size)  #  Update the parameter with the gradient of the parameter 
    with torch.no_grad():
        train_l = loss(net(features, w, b), labels)
        print(f'epoch {epoch + 1}, loss {float(train_l.mean()):f}')
print(f'w The estimation error of : {true_w - w.reshape(true_w.shape)}')
print(f'b The estimation error of : {true_b - b}')

Using frames

import numpy as np
import torch
from torch.utils import data
from d2l import torch as d2l
from torch import nn
true_w = torch.tensor([2, -3.4])
true_b = 4.2
features, labels = d2l.synthetic_data(true_w, true_b, 1000)
def load_array(data_arrays, batch_size, is_train=True):  #@save
    """ Construct a PyTorch Data iterators """
    dataset = data.TensorDataset(*data_arrays)
    return data.DataLoader(dataset, batch_size, shuffle=is_train)

batch_size = 10
data_iter = load_array((features, labels), batch_size)
net = nn.Sequential(nn.Linear(2, 1))
net[0].weight.data.normal_(0, 0.01)
net[0].bias.data.fill_(0)
loss = nn.MSELoss()
trainer = torch.optim.SGD(net.parameters(), lr=0.03)
num_epochs = 3
for epoch in range(num_epochs):
    for X, y in data_iter:
        l = loss(net(X) ,y)
        trainer.zero_grad()
        l.backward()
        trainer.step()
    l = loss(net(features), labels)
    print(f'epoch {epoch + 1}, loss {l:f}')