# 實(shí)現(xiàn)一個(gè)線性回歸
# 所有的層結(jié)構(gòu)和損失函數(shù)都來(lái)自于 torch.nn
# torch.optim 是一個(gè)實(shí)現(xiàn)各種優(yōu)化算法的包，調(diào)用的時(shí)候必須是需要優(yōu)化的參數(shù)傳入，這些參數(shù)都必須是Variable
 
x_train = np.array([[3.3],[4.4],[5.5],[6.71],[6.93],[4.168],[9.779],[6.182],[7.59],[2.167],[7.042],[10.791],[5.313],[7.997],[3.1]],dtype=np.float32)
y_train = np.array([[1.7],[2.76],[2.09],[3.19],[1.694],[1.573],[3.366],[2.596],[2.53],[1.221],[2.827],[3.465],[1.65],[2.904],[1.3]],dtype=np.float32)
 
# 首先我們需要將array轉(zhuǎn)化成tensor，因?yàn)閜ytorch處理的單元是Tensor
 
x_train = torch.from_numpy(x_train)
y_train = torch.from_numpy(y_train)
 
 
# def a simple network
 
class LinearRegression(nn.Module):
    def __init__(self):
        super(LinearRegression,self).__init__()
        self.linear = nn.Linear(1, 1)  # input and output is 2_dimension
    def forward(self, x):
        out = self.linear(x)
        return out
 
 
if torch.cuda.is_available():
    model = LinearRegression().cuda()
    #model = model.cuda()
else:
    model = LinearRegression()
    #model = model.cuda()
 
# 定義loss function 和 optimize func
criterion = nn.MSELoss()   # 均方誤差作為優(yōu)化函數(shù)
optimizer = torch.optim.SGD(model.parameters(),lr=1e-3)
num_epochs = 30000
for epoch in range(num_epochs):
    if torch.cuda.is_available():
        inputs = Variable(x_train).cuda()
        outputs = Variable(y_train).cuda()
    else:
        inputs = Variable(x_train)
        outputs = Variable(y_train)
 
    # forward
    out = model(inputs)
    loss = criterion(out,outputs)
 
    # backword
    optimizer.zero_grad()  # 每次做反向傳播之前都要進(jìn)行歸零梯度。不然梯度會(huì)累加在一起，造成不收斂的結(jié)果
    loss.backward()
    optimizer.step()
 
    if (epoch +1)%20==0:
        print('Epoch[{}/{}], loss: {:.6f}'.format(epoch+1,num_epochs,loss.data))
 
 
model.eval()  # 將模型變成測(cè)試模式
predict = model(Variable(x_train).cuda())
predict = predict.data.cpu().numpy()
plt.plot(x_train.numpy(),y_train.numpy(),'ro',label = 'original data')
plt.plot(x_train.numpy(),predict,label = 'Fitting line')
plt.show()

# _*_encoding=utf-8_*_
# pytorch 里面最基本的操作對(duì)象是Tensor,pytorch 的tensor可以和numpy的ndarray相互轉(zhuǎn)化。
# 實(shí)現(xiàn)一個(gè)線性回歸
# 所有的層結(jié)構(gòu)和損失函數(shù)都來(lái)自于 torch.nn
# torch.optim 是一個(gè)實(shí)現(xiàn)各種優(yōu)化算法的包，調(diào)用的時(shí)候必須是需要優(yōu)化的參數(shù)傳入，這些參數(shù)都必須是Variable
 
 
# 實(shí)現(xiàn) y = b + w1 *x + w2 *x**2 +w3*x**3
import os
os.environ['CUDA_DEVICE_ORDER']="PCI_BUS_ID"
os.environ['CUDA_VISIBLE_DEVICES']='0'
import torch
import numpy as np
from torch.autograd import Variable
import matplotlib.pyplot as plt
from torch import nn
 
 
# pre_processing
def make_feature(x):
    x = x.unsqueeze(1)   # unsquenze 是為了添加維度1的，0表示第一維度，1表示第二維度，將tensor大小由3變?yōu)椋?,1）
    return torch.cat([x ** i for i in range(1, 4)], 1)
 
# 定義好真實(shí)的數(shù)據(jù)
 
 
def f(x):
    W_output = torch.Tensor([0.5, 3, 2.4]).unsqueeze(1)
    b_output = torch.Tensor([0.9])
    return x.mm(W_output)+b_output[0]  # 外積，矩陣乘法
 
 
# 批量處理數(shù)據(jù)
def get_batch(batch_size =32):
 
    random = torch.randn(batch_size)
    x = make_feature(random)
    y = f(x)
    if torch.cuda.is_available():
 
        return Variable(x).cuda(),Variable(y).cuda()
    else:
        return Variable(x),Variable(y)
 
 
 
# def model
class poly_model(nn.Module):
    def __init__(self):
        super(poly_model,self).__init__()
        self.poly = nn.Linear(3,1)
    def forward(self,input):
        output = self.poly(input)
        return output
 
if torch.cuda.is_available():
    print("sdf")
    model = poly_model().cuda()
else:
    model = poly_model()
 
 
# 定義損失函數(shù)和優(yōu)化器
criterion = nn.MSELoss()
optimizer = torch.optim.SGD(model.parameters(), lr=1e-3)
 
epoch = 0
while True:
    batch_x, batch_y = get_batch()
    #print(batch_x)
    output = model(batch_x)
    loss = criterion(output,batch_y)
    print_loss = loss.data
    optimizer.zero_grad()
    loss.backward()
    optimizer.step()
    epoch = epoch +1
    if print_loss < 1e-3:
        print(print_loss)
        break
 
model.eval()
print("Epoch = {}".format(epoch))
 
batch_x, batch_y = get_batch()
predict = model(batch_x)
a = predict - batch_y
y = torch.sum(a)
print('y = ',y)
predict = predict.data.cpu().numpy()
plt.plot(batch_x.cpu().numpy(),batch_y.cpu().numpy(),'ro',label = 'Original data')
plt.plot(batch_x.cpu().numpy(),predict,'b', ls='--',label = 'Fitting line')
plt.show()