Pytourch Deep Neuron Network¶
1. Fire up¶
import torch
from torch.autograd import Variable
import torch.nn as nn
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.metrics import accuracy_score
2. Data generating¶
x1_Pos = []
x2_Pos = []
y_Pos = []
for i in range(1000):
temp = 4.0 * np.random.rand() - 2.0
y_Pos.append(1)
x1_Pos.append(temp)
if i % 2 == 0:
x2_Pos.append(np.sqrt(4.0 - temp ** 2) + 0.3 * np.random.randn ( ))
elif i % 2 == 1:
x2_Pos.append(- np.sqrt(4.0 - temp ** 2) + 0.3 * np.random.randn () )
这段代码生成了一个包含1000个数据点的数据集，其中每个数据点有三个特征：x1_Pos，x2_Pos和y_Pos。
- 对于每个数据点，首先生成一个在区间[-2.0, 2.0]上均匀分布的随机数
temp。 - 对于
y_Pos特征，将其设置为常数1，表示正例。 - 对于
x1_Pos特征，将其设置为temp的值。 - 对于
x2_Pos特征，根据以下规则进行设置： - 如果当前数据点的索引i是偶数，则将
x2_Pos设置为sqrt(4.0 - temp^2) + 0.3 * 随机正态分布的随机数。 - 如果当前数据点的索引i是奇数，则将
x2_Pos设置为- sqrt(4.0 - temp^2) + 0.3 * 随机正态分布的随机数。
综合起来，这段代码生成了一个具有两个特征（x1_Pos和x2_Pos）和一个标签（y_Pos）的二分类数据集，其中x1_Pos和x2_Pos的取值与彼此相关，并且存在一定的噪声。
x1_Neg = []
x2_Neg = []
y_Neg = []
for i in range (1000):
temp = 10.0 * np.random.rand() - 5.0
y_Neg.append(0)
x1_Neg. append (temp)
if i % 2 == 0:
x2_Neg.append(np.sqrt(25.0 - temp ** 2) + 0.3 * np.random.randn ())
elif i % 2 == 1:
x2_Neg.append(- np.sqrt (25.0 - temp ** 2) + 0.3 * np.random.randn ())
plt.figure(figsize = (8, 6))
plt.scatter(x1_Pos, x2_Pos, color = "black", label = "Class 1", alpha = 0.5)
plt.scatter(x1_Neg, x2_Neg, color = "red", label = "Class 0", alpha = 0.5)
plt.xlabel("xl")
plt.ylabel("x2")
plt.legend ()
plt.grid()
plt.show()
Dict = {"x1": x1_Pos + x1_Neg, "x2": x2_Pos + x2_Neg, "y": y_Pos + y_Neg}
DataTrain = pd.DataFrame(Dict)
DataTrain
| x1 | x2 | y | |
|---|---|---|---|
| 0 | 0.117529 | 1.440490 | 1 |
| 1 | -1.255701 | -1.325910 | 1 |
| 2 | 1.950944 | 0.462942 | 1 |
| 3 | 0.744222 | -1.889267 | 1 |
| 4 | -0.637758 | 1.741075 | 1 |
| ... | ... | ... | ... |
| 1995 | -4.094008 | -3.365285 | 0 |
| 1996 | -4.273857 | 3.205604 | 0 |
| 1997 | 1.081440 | -4.366669 | 0 |
| 1998 | 3.146540 | 4.072895 | 0 |
| 1999 | 2.274089 | -4.855089 | 0 |
2000 rows × 3 columns
x1_Pos = []
x2_Pos = []
y_Pos = []
for i in range(250):
temp = 4.0 * np.random.rand() - 2.0
y_Pos.append(1)
x1_Pos.append(temp)
if i % 2 == 0:
x2_Pos.append(np.sqrt(4.0 - temp ** 2) + 0.3 * np.random.randn ( ))
elif i % 2 == 1:
x2_Pos.append(- np.sqrt(4.0 - temp ** 2) + 0.3 * np.random.randn () )
x1_Neg = []
x2_Neg = []
y_Neg = []
for i in range (250):
temp = 10.0 * np.random.rand() - 5.0
y_Neg.append(0)
x1_Neg. append (temp)
if i % 2 == 0:
x2_Neg.append(np.sqrt(25.0 - temp ** 2) + 0.3 * np.random.randn ())
elif i % 2 == 1:
x2_Neg.append(- np.sqrt (25.0 - temp ** 2) + 0.3 * np.random.randn ())
Dict = {"x1": x1_Pos + x1_Neg, "x2": x2_Pos + x2_Neg, "y": y_Pos + y_Neg}
DataTest = pd.DataFrame(Dict)
DataTest
| x1 | x2 | y | |
|---|---|---|---|
| 0 | 1.620214 | 1.175676 | 1 |
| 1 | -0.509711 | -1.657162 | 1 |
| 2 | -0.471200 | 2.130239 | 1 |
| 3 | 0.906513 | -1.790748 | 1 |
| 4 | -1.163231 | 1.673213 | 1 |
| ... | ... | ... | ... |
| 495 | 3.438152 | -3.321079 | 0 |
| 496 | -3.602392 | 3.356225 | 0 |
| 497 | -3.557373 | -3.948084 | 0 |
| 498 | -0.313884 | 4.804572 | 0 |
| 499 | 4.260027 | -2.314735 | 0 |
500 rows × 3 columns
3. Model training¶
\[\vec{x}\in\mathbb{R}^{1\times 2}\]
\[W_1\in\mathbb{R}^{2\times 10}; b_1\in\mathbb{R}^{1\times 10}\]
\[W_2\in\mathbb{R}^{10\times 2}; b_2\in\mathbb{R}^{1\times 2}\]
\[z_1 = \sigma(\vec{x}W_1+b_1)\]
\[z_2 = \sigma(z_1W_2+b_2)\]
\[\hat{y} = {\rm{softmax}}(z_2)\]
\[Loss (W_1, W_2, b_1, b_2) = -\frac{1}{N}\sum_{i=1}^{N}[y_i\log\hat{y} +（1-y_1)\log(1-\hat{y})]\]
\[k = 0,1,2,...\]
\[W_1^{(k+1)} = W_1^{(k)} - \alpha_k\frac{\partial Loss(W_1^{(k)}, W_2^{(k)}, b_1^{(k)}, b_2^{(k})}{\partial
W_1}\]
\[W_2^{(k+1)} = W_2^{(k)} - \alpha_k\frac{\partial Loss(W_1^{(k)}, W_2^{(k)}, b_1^{(k)}, b_2^{(k})}{\partial
W_2}\]
\[b_1^{(k+1)} = b_1^{(k)} - \alpha_k\frac{\partial Loss(W_1^{(k)}, W_2^{(k)}, b_1^{(k)}, b_2^{(k})}{\partial
b_1}\]
\[b_2^{(k+1)} = b_2^{(k)} - \alpha_k\frac{\partial Loss(W_1^{(k)}, W_2^{(k)}, b_1^{(k)}, b_2^{(k})}{\partial
b_2}\]
class DeepNeuralNetworkModel(nn.Module):
# Constructor of the class
def __init__(self, input_dim1, output_dim1, input_dim2, output_dim2): # output_dim1 = input_dim2
super(DeepNeuralNetworkModel, self).__init__()
# Fully Connected Layer 1
self.FC_layer1 = nn.Linear(input_dim1, output_dim1)
# nn.init.constant_(self.FC_layerl.weight, 0.1)
# nn.init.constant_(self.FC_layer1.bias, -0.1)
# Fully Connected Layer 2
self.FC_layer2 = nn.Linear(input_dim2, output_dim2)
# nn.init.constant_(self.FC_layer2.weight, 0.1)
# nn.init.constant_(self.FC_layer2.bias, -0.1)
# Activation Function Sigmoid()
self.act_sig = nn.Sigmoid()
# Forward propagation function
def forward(self, x): # dim of x: N*2
z1_ = self.FC_layer1(x)
z1 = self.act_sig(z1_)
z2_ = self.FC_layer2(z1)
z2 = self.act_sig(z2_)
return z2
这段代码定义了一个名为DeepNeuralNetworkModel的神经网络模型类。该类继承自nn.Module，是PyTorch中构建神经网络模型的基类。
在该类中，包含以下几个主要部分：
-
构造函数
__init__：初始化模型的参数和层结构。它接收输入维度和输出维度作为参数，并定义了两个全连接层（self.FC_layer1和self.FC_layer2），以及一个Sigmoid激活函数（self.act_sig）。 -
前向传播函数
forward：定义了数据在模型中前向传播的流程。输入数据x首先通过第一个全连接层（self.FC_layer1）进行线性变换，然后经过Sigmoid激活函数（self.act_sig）。接着，输出结果再通过第二个全连接层（self.FC_layer2）进行线性变换，最后再经过一次Sigmoid激活函数得到最终的输出结果。
该模型的结构是一个两层的全连接神经网络:
- 第一层输入维度为
input_dim1，第一层输出维度为output_dim1 - 第二层的输入维度为
output_dim1（即第一层的输出维度），第二层的输出维度为output_dim2。
整个模型的作用是将输入数据映射到一个输出结果，通过两个全连接层和激活函数的组合来实现非线性映射和特征提取。
X_vec = torch.tensor(DataTrain[["x1", "x2"]].values, dtype=torch.float32) # N*2
y_vec = torch.tensor(DataTrain["y"].values, dtype=torch.int64).reshape(-1, 1) # N*1
alpha = 0.2
DNN_Model = DeepNeuralNetworkModel(2, 10, 10, 2)
# 优化DNN model里的所有参数，lr:learning rate 学习率（动态变化）
optimizer = torch.optim.SGD(DNN_Model.parameters(), lr = alpha)
# 损失函数
loss_function = nn.CrossEntropyLoss ()
# Dynamically Change the learning rate
def adjust_learning_rate(optimizer, epoch): # epoch就可以理解为是一次迭代
lr = alpha / (1 + 0.00001 * epoch)
for param_group in optimizer.param_groups:
param_group['lr'] = lr
Iter_times = 200000
loss_list = []
for i in range(Iter_times):
outputs = DNN_Model.forward(X_vec) # forward propagation
loss = loss_function(outputs, torch.squeeze(y_vec)) # compute loss
loss.backward() # backward propagation
optimizer.step() # update parameters
optimizer.zero_grad() # Reset grad to 0
if (i + 1) % 500 == 0:
print(i + 1, "iterations have been completed!")
print("-> Now loss =", loss)
print("=========================================")
adjust_learning_rate(optimizer, i)
loss_list.append(loss)
# length = loss_list.__len__()
# if(torch.abs(loss_listrlength - 11 - loss_list rlength - 21)<10** 1-15) and length >=2):
# break
500 iterations have been completed!
-> Now loss = tensor(0.6814, grad_fn=<NllLossBackward0>)
=========================================
1000 iterations have been completed!
-> Now loss = tensor(0.6406, grad_fn=<NllLossBackward0>)
=========================================
1500 iterations have been completed!
-> Now loss = tensor(0.5620, grad_fn=<NllLossBackward0>)
=========================================
2000 iterations have been completed!
-> Now loss = tensor(0.4912, grad_fn=<NllLossBackward0>)
=========================================
2500 iterations have been completed!
-> Now loss = tensor(0.4461, grad_fn=<NllLossBackward0>)
=========================================
3000 iterations have been completed!
-> Now loss = tensor(0.4169, grad_fn=<NllLossBackward0>)
=========================================
3500 iterations have been completed!
-> Now loss = tensor(0.3933, grad_fn=<NllLossBackward0>)
=========================================
4000 iterations have been completed!
-> Now loss = tensor(0.3756, grad_fn=<NllLossBackward0>)
=========================================
4500 iterations have been completed!
-> Now loss = tensor(0.3634, grad_fn=<NllLossBackward0>)
=========================================
5000 iterations have been completed!
-> Now loss = tensor(0.3549, grad_fn=<NllLossBackward0>)
=========================================
5500 iterations have been completed!
-> Now loss = tensor(0.3486, grad_fn=<NllLossBackward0>)
=========================================
6000 iterations have been completed!
-> Now loss = tensor(0.3439, grad_fn=<NllLossBackward0>)
=========================================
6500 iterations have been completed!
-> Now loss = tensor(0.3403, grad_fn=<NllLossBackward0>)
=========================================
7000 iterations have been completed!
-> Now loss = tensor(0.3374, grad_fn=<NllLossBackward0>)
=========================================
7500 iterations have been completed!
-> Now loss = tensor(0.3350, grad_fn=<NllLossBackward0>)
=========================================
8000 iterations have been completed!
-> Now loss = tensor(0.3331, grad_fn=<NllLossBackward0>)
=========================================
8500 iterations have been completed!
-> Now loss = tensor(0.3314, grad_fn=<NllLossBackward0>)
=========================================
9000 iterations have been completed!
-> Now loss = tensor(0.3300, grad_fn=<NllLossBackward0>)
=========================================
9500 iterations have been completed!
-> Now loss = tensor(0.3289, grad_fn=<NllLossBackward0>)
=========================================
10000 iterations have been completed!
-> Now loss = tensor(0.3278, grad_fn=<NllLossBackward0>)
=========================================
10500 iterations have been completed!
-> Now loss = tensor(0.3269, grad_fn=<NllLossBackward0>)
=========================================
11000 iterations have been completed!
-> Now loss = tensor(0.3261, grad_fn=<NllLossBackward0>)
=========================================
11500 iterations have been completed!
-> Now loss = tensor(0.3254, grad_fn=<NllLossBackward0>)
=========================================
12000 iterations have been completed!
-> Now loss = tensor(0.3248, grad_fn=<NllLossBackward0>)
=========================================
12500 iterations have been completed!
-> Now loss = tensor(0.3242, grad_fn=<NllLossBackward0>)
=========================================
13000 iterations have been completed!
-> Now loss = tensor(0.3237, grad_fn=<NllLossBackward0>)
=========================================
13500 iterations have been completed!
-> Now loss = tensor(0.3232, grad_fn=<NllLossBackward0>)
=========================================
14000 iterations have been completed!
-> Now loss = tensor(0.3228, grad_fn=<NllLossBackward0>)
=========================================
14500 iterations have been completed!
-> Now loss = tensor(0.3224, grad_fn=<NllLossBackward0>)
=========================================
15000 iterations have been completed!
-> Now loss = tensor(0.3220, grad_fn=<NllLossBackward0>)
=========================================
15500 iterations have been completed!
-> Now loss = tensor(0.3217, grad_fn=<NllLossBackward0>)
=========================================
16000 iterations have been completed!
-> Now loss = tensor(0.3214, grad_fn=<NllLossBackward0>)
=========================================
16500 iterations have been completed!
-> Now loss = tensor(0.3211, grad_fn=<NllLossBackward0>)
=========================================
17000 iterations have been completed!
-> Now loss = tensor(0.3208, grad_fn=<NllLossBackward0>)
=========================================
17500 iterations have been completed!
-> Now loss = tensor(0.3206, grad_fn=<NllLossBackward0>)
=========================================
18000 iterations have been completed!
-> Now loss = tensor(0.3204, grad_fn=<NllLossBackward0>)
=========================================
18500 iterations have been completed!
-> Now loss = tensor(0.3202, grad_fn=<NllLossBackward0>)
=========================================
19000 iterations have been completed!
-> Now loss = tensor(0.3200, grad_fn=<NllLossBackward0>)
=========================================
19500 iterations have been completed!
-> Now loss = tensor(0.3198, grad_fn=<NllLossBackward0>)
=========================================
20000 iterations have been completed!
-> Now loss = tensor(0.3196, grad_fn=<NllLossBackward0>)
=========================================
20500 iterations have been completed!
-> Now loss = tensor(0.3194, grad_fn=<NllLossBackward0>)
=========================================
21000 iterations have been completed!
-> Now loss = tensor(0.3193, grad_fn=<NllLossBackward0>)
=========================================
21500 iterations have been completed!
-> Now loss = tensor(0.3191, grad_fn=<NllLossBackward0>)
=========================================
22000 iterations have been completed!
-> Now loss = tensor(0.3190, grad_fn=<NllLossBackward0>)
=========================================
22500 iterations have been completed!
-> Now loss = tensor(0.3188, grad_fn=<NllLossBackward0>)
=========================================
23000 iterations have been completed!
-> Now loss = tensor(0.3187, grad_fn=<NllLossBackward0>)
=========================================
23500 iterations have been completed!
-> Now loss = tensor(0.3186, grad_fn=<NllLossBackward0>)
=========================================
24000 iterations have been completed!
-> Now loss = tensor(0.3185, grad_fn=<NllLossBackward0>)
=========================================
24500 iterations have been completed!
-> Now loss = tensor(0.3184, grad_fn=<NllLossBackward0>)
=========================================
25000 iterations have been completed!
-> Now loss = tensor(0.3182, grad_fn=<NllLossBackward0>)
=========================================
25500 iterations have been completed!
-> Now loss = tensor(0.3181, grad_fn=<NllLossBackward0>)
=========================================
26000 iterations have been completed!
-> Now loss = tensor(0.3180, grad_fn=<NllLossBackward0>)
=========================================
26500 iterations have been completed!
-> Now loss = tensor(0.3180, grad_fn=<NllLossBackward0>)
=========================================
27000 iterations have been completed!
-> Now loss = tensor(0.3179, grad_fn=<NllLossBackward0>)
=========================================
27500 iterations have been completed!
-> Now loss = tensor(0.3178, grad_fn=<NllLossBackward0>)
=========================================
28000 iterations have been completed!
-> Now loss = tensor(0.3177, grad_fn=<NllLossBackward0>)
=========================================
28500 iterations have been completed!
-> Now loss = tensor(0.3176, grad_fn=<NllLossBackward0>)
=========================================
29000 iterations have been completed!
-> Now loss = tensor(0.3175, grad_fn=<NllLossBackward0>)
=========================================
29500 iterations have been completed!
-> Now loss = tensor(0.3175, grad_fn=<NllLossBackward0>)
=========================================
30000 iterations have been completed!
-> Now loss = tensor(0.3174, grad_fn=<NllLossBackward0>)
=========================================
30500 iterations have been completed!
-> Now loss = tensor(0.3173, grad_fn=<NllLossBackward0>)
=========================================
31000 iterations have been completed!
-> Now loss = tensor(0.3173, grad_fn=<NllLossBackward0>)
=========================================
31500 iterations have been completed!
-> Now loss = tensor(0.3172, grad_fn=<NllLossBackward0>)
=========================================
32000 iterations have been completed!
-> Now loss = tensor(0.3171, grad_fn=<NllLossBackward0>)
=========================================
32500 iterations have been completed!
-> Now loss = tensor(0.3171, grad_fn=<NllLossBackward0>)
=========================================
33000 iterations have been completed!
-> Now loss = tensor(0.3170, grad_fn=<NllLossBackward0>)
=========================================
33500 iterations have been completed!
-> Now loss = tensor(0.3170, grad_fn=<NllLossBackward0>)
=========================================
34000 iterations have been completed!
-> Now loss = tensor(0.3169, grad_fn=<NllLossBackward0>)
=========================================
34500 iterations have been completed!
-> Now loss = tensor(0.3169, grad_fn=<NllLossBackward0>)
=========================================
35000 iterations have been completed!
-> Now loss = tensor(0.3168, grad_fn=<NllLossBackward0>)
=========================================
35500 iterations have been completed!
-> Now loss = tensor(0.3168, grad_fn=<NllLossBackward0>)
=========================================
36000 iterations have been completed!
-> Now loss = tensor(0.3167, grad_fn=<NllLossBackward0>)
=========================================
36500 iterations have been completed!
-> Now loss = tensor(0.3167, grad_fn=<NllLossBackward0>)
=========================================
37000 iterations have been completed!
-> Now loss = tensor(0.3166, grad_fn=<NllLossBackward0>)
=========================================
37500 iterations have been completed!
-> Now loss = tensor(0.3166, grad_fn=<NllLossBackward0>)
=========================================
38000 iterations have been completed!
-> Now loss = tensor(0.3165, grad_fn=<NllLossBackward0>)
=========================================
38500 iterations have been completed!
-> Now loss = tensor(0.3165, grad_fn=<NllLossBackward0>)
=========================================
39000 iterations have been completed!
-> Now loss = tensor(0.3165, grad_fn=<NllLossBackward0>)
=========================================
39500 iterations have been completed!
-> Now loss = tensor(0.3164, grad_fn=<NllLossBackward0>)
=========================================
40000 iterations have been completed!
-> Now loss = tensor(0.3164, grad_fn=<NllLossBackward0>)
=========================================
40500 iterations have been completed!
-> Now loss = tensor(0.3164, grad_fn=<NllLossBackward0>)
=========================================
41000 iterations have been completed!
-> Now loss = tensor(0.3163, grad_fn=<NllLossBackward0>)
=========================================
41500 iterations have been completed!
-> Now loss = tensor(0.3163, grad_fn=<NllLossBackward0>)
=========================================
42000 iterations have been completed!
-> Now loss = tensor(0.3163, grad_fn=<NllLossBackward0>)
=========================================
42500 iterations have been completed!
-> Now loss = tensor(0.3162, grad_fn=<NllLossBackward0>)
=========================================
43000 iterations have been completed!
-> Now loss = tensor(0.3162, grad_fn=<NllLossBackward0>)
=========================================
43500 iterations have been completed!
-> Now loss = tensor(0.3162, grad_fn=<NllLossBackward0>)
=========================================
44000 iterations have been completed!
-> Now loss = tensor(0.3161, grad_fn=<NllLossBackward0>)
=========================================
44500 iterations have been completed!
-> Now loss = tensor(0.3161, grad_fn=<NllLossBackward0>)
=========================================
45000 iterations have been completed!
-> Now loss = tensor(0.3161, grad_fn=<NllLossBackward0>)
=========================================
45500 iterations have been completed!
-> Now loss = tensor(0.3160, grad_fn=<NllLossBackward0>)
=========================================
46000 iterations have been completed!
-> Now loss = tensor(0.3160, grad_fn=<NllLossBackward0>)
=========================================
46500 iterations have been completed!
-> Now loss = tensor(0.3160, grad_fn=<NllLossBackward0>)
=========================================
47000 iterations have been completed!
-> Now loss = tensor(0.3160, grad_fn=<NllLossBackward0>)
=========================================
47500 iterations have been completed!
-> Now loss = tensor(0.3159, grad_fn=<NllLossBackward0>)
=========================================
48000 iterations have been completed!
-> Now loss = tensor(0.3159, grad_fn=<NllLossBackward0>)
=========================================
48500 iterations have been completed!
-> Now loss = tensor(0.3159, grad_fn=<NllLossBackward0>)
=========================================
49000 iterations have been completed!
-> Now loss = tensor(0.3159, grad_fn=<NllLossBackward0>)
=========================================
49500 iterations have been completed!
-> Now loss = tensor(0.3158, grad_fn=<NllLossBackward0>)
=========================================
50000 iterations have been completed!
-> Now loss = tensor(0.3158, grad_fn=<NllLossBackward0>)
=========================================
50500 iterations have been completed!
-> Now loss = tensor(0.3158, grad_fn=<NllLossBackward0>)
=========================================
51000 iterations have been completed!
-> Now loss = tensor(0.3158, grad_fn=<NllLossBackward0>)
=========================================
51500 iterations have been completed!
-> Now loss = tensor(0.3157, grad_fn=<NllLossBackward0>)
=========================================
52000 iterations have been completed!
-> Now loss = tensor(0.3157, grad_fn=<NllLossBackward0>)
=========================================
52500 iterations have been completed!
-> Now loss = tensor(0.3157, grad_fn=<NllLossBackward0>)
=========================================
53000 iterations have been completed!
-> Now loss = tensor(0.3157, grad_fn=<NllLossBackward0>)
=========================================
53500 iterations have been completed!
-> Now loss = tensor(0.3157, grad_fn=<NllLossBackward0>)
=========================================
54000 iterations have been completed!
-> Now loss = tensor(0.3156, grad_fn=<NllLossBackward0>)
=========================================
54500 iterations have been completed!
-> Now loss = tensor(0.3156, grad_fn=<NllLossBackward0>)
=========================================
55000 iterations have been completed!
-> Now loss = tensor(0.3156, grad_fn=<NllLossBackward0>)
=========================================
55500 iterations have been completed!
-> Now loss = tensor(0.3156, grad_fn=<NllLossBackward0>)
=========================================
56000 iterations have been completed!
-> Now loss = tensor(0.3156, grad_fn=<NllLossBackward0>)
=========================================
56500 iterations have been completed!
-> Now loss = tensor(0.3156, grad_fn=<NllLossBackward0>)
=========================================
57000 iterations have been completed!
-> Now loss = tensor(0.3155, grad_fn=<NllLossBackward0>)
=========================================
57500 iterations have been completed!
-> Now loss = tensor(0.3155, grad_fn=<NllLossBackward0>)
=========================================
58000 iterations have been completed!
-> Now loss = tensor(0.3155, grad_fn=<NllLossBackward0>)
=========================================
58500 iterations have been completed!
-> Now loss = tensor(0.3155, grad_fn=<NllLossBackward0>)
=========================================
59000 iterations have been completed!
-> Now loss = tensor(0.3155, grad_fn=<NllLossBackward0>)
=========================================
59500 iterations have been completed!
-> Now loss = tensor(0.3155, grad_fn=<NllLossBackward0>)
=========================================
60000 iterations have been completed!
-> Now loss = tensor(0.3154, grad_fn=<NllLossBackward0>)
=========================================
60500 iterations have been completed!
-> Now loss = tensor(0.3154, grad_fn=<NllLossBackward0>)
=========================================
61000 iterations have been completed!
-> Now loss = tensor(0.3154, grad_fn=<NllLossBackward0>)
=========================================
61500 iterations have been completed!
-> Now loss = tensor(0.3154, grad_fn=<NllLossBackward0>)
=========================================
62000 iterations have been completed!
-> Now loss = tensor(0.3154, grad_fn=<NllLossBackward0>)
=========================================
62500 iterations have been completed!
-> Now loss = tensor(0.3154, grad_fn=<NllLossBackward0>)
=========================================
63000 iterations have been completed!
-> Now loss = tensor(0.3153, grad_fn=<NllLossBackward0>)
=========================================
63500 iterations have been completed!
-> Now loss = tensor(0.3153, grad_fn=<NllLossBackward0>)
=========================================
64000 iterations have been completed!
-> Now loss = tensor(0.3153, grad_fn=<NllLossBackward0>)
=========================================
64500 iterations have been completed!
-> Now loss = tensor(0.3153, grad_fn=<NllLossBackward0>)
=========================================
65000 iterations have been completed!
-> Now loss = tensor(0.3153, grad_fn=<NllLossBackward0>)
=========================================
65500 iterations have been completed!
-> Now loss = tensor(0.3153, grad_fn=<NllLossBackward0>)
=========================================
66000 iterations have been completed!
-> Now loss = tensor(0.3153, grad_fn=<NllLossBackward0>)
=========================================
66500 iterations have been completed!
-> Now loss = tensor(0.3153, grad_fn=<NllLossBackward0>)
=========================================
67000 iterations have been completed!
-> Now loss = tensor(0.3152, grad_fn=<NllLossBackward0>)
=========================================
67500 iterations have been completed!
-> Now loss = tensor(0.3152, grad_fn=<NllLossBackward0>)
=========================================
68000 iterations have been completed!
-> Now loss = tensor(0.3152, grad_fn=<NllLossBackward0>)
=========================================
68500 iterations have been completed!
-> Now loss = tensor(0.3152, grad_fn=<NllLossBackward0>)
=========================================
69000 iterations have been completed!
-> Now loss = tensor(0.3152, grad_fn=<NllLossBackward0>)
=========================================
69500 iterations have been completed!
-> Now loss = tensor(0.3152, grad_fn=<NllLossBackward0>)
=========================================
70000 iterations have been completed!
-> Now loss = tensor(0.3152, grad_fn=<NllLossBackward0>)
=========================================
70500 iterations have been completed!
-> Now loss = tensor(0.3152, grad_fn=<NllLossBackward0>)
=========================================
71000 iterations have been completed!
-> Now loss = tensor(0.3151, grad_fn=<NllLossBackward0>)
=========================================
71500 iterations have been completed!
-> Now loss = tensor(0.3151, grad_fn=<NllLossBackward0>)
=========================================
72000 iterations have been completed!
-> Now loss = tensor(0.3151, grad_fn=<NllLossBackward0>)
=========================================
72500 iterations have been completed!
-> Now loss = tensor(0.3151, grad_fn=<NllLossBackward0>)
=========================================
73000 iterations have been completed!
-> Now loss = tensor(0.3151, grad_fn=<NllLossBackward0>)
=========================================
73500 iterations have been completed!
-> Now loss = tensor(0.3151, grad_fn=<NllLossBackward0>)
=========================================
74000 iterations have been completed!
-> Now loss = tensor(0.3151, grad_fn=<NllLossBackward0>)
=========================================
74500 iterations have been completed!
-> Now loss = tensor(0.3151, grad_fn=<NllLossBackward0>)
=========================================
75000 iterations have been completed!
-> Now loss = tensor(0.3151, grad_fn=<NllLossBackward0>)
=========================================
75500 iterations have been completed!
-> Now loss = tensor(0.3151, grad_fn=<NllLossBackward0>)
=========================================
76000 iterations have been completed!
-> Now loss = tensor(0.3150, grad_fn=<NllLossBackward0>)
=========================================
76500 iterations have been completed!
-> Now loss = tensor(0.3150, grad_fn=<NllLossBackward0>)
=========================================
77000 iterations have been completed!
-> Now loss = tensor(0.3150, grad_fn=<NllLossBackward0>)
=========================================
77500 iterations have been completed!
-> Now loss = tensor(0.3150, grad_fn=<NllLossBackward0>)
=========================================
78000 iterations have been completed!
-> Now loss = tensor(0.3150, grad_fn=<NllLossBackward0>)
=========================================
78500 iterations have been completed!
-> Now loss = tensor(0.3150, grad_fn=<NllLossBackward0>)
=========================================
79000 iterations have been completed!
-> Now loss = tensor(0.3150, grad_fn=<NllLossBackward0>)
=========================================
79500 iterations have been completed!
-> Now loss = tensor(0.3150, grad_fn=<NllLossBackward0>)
=========================================
80000 iterations have been completed!
-> Now loss = tensor(0.3150, grad_fn=<NllLossBackward0>)
=========================================
80500 iterations have been completed!
-> Now loss = tensor(0.3150, grad_fn=<NllLossBackward0>)
=========================================
81000 iterations have been completed!
-> Now loss = tensor(0.3150, grad_fn=<NllLossBackward0>)
=========================================
81500 iterations have been completed!
-> Now loss = tensor(0.3149, grad_fn=<NllLossBackward0>)
=========================================
82000 iterations have been completed!
-> Now loss = tensor(0.3149, grad_fn=<NllLossBackward0>)
=========================================
82500 iterations have been completed!
-> Now loss = tensor(0.3149, grad_fn=<NllLossBackward0>)
=========================================
83000 iterations have been completed!
-> Now loss = tensor(0.3149, grad_fn=<NllLossBackward0>)
=========================================
83500 iterations have been completed!
-> Now loss = tensor(0.3149, grad_fn=<NllLossBackward0>)
=========================================
84000 iterations have been completed!
-> Now loss = tensor(0.3149, grad_fn=<NllLossBackward0>)
=========================================
84500 iterations have been completed!
-> Now loss = tensor(0.3149, grad_fn=<NllLossBackward0>)
=========================================
85000 iterations have been completed!
-> Now loss = tensor(0.3149, grad_fn=<NllLossBackward0>)
=========================================
85500 iterations have been completed!
-> Now loss = tensor(0.3149, grad_fn=<NllLossBackward0>)
=========================================
86000 iterations have been completed!
-> Now loss = tensor(0.3149, grad_fn=<NllLossBackward0>)
=========================================
86500 iterations have been completed!
-> Now loss = tensor(0.3149, grad_fn=<NllLossBackward0>)
=========================================
87000 iterations have been completed!
-> Now loss = tensor(0.3149, grad_fn=<NllLossBackward0>)
=========================================
87500 iterations have been completed!
-> Now loss = tensor(0.3149, grad_fn=<NllLossBackward0>)
=========================================
88000 iterations have been completed!
-> Now loss = tensor(0.3148, grad_fn=<NllLossBackward0>)
=========================================
88500 iterations have been completed!
-> Now loss = tensor(0.3148, grad_fn=<NllLossBackward0>)
=========================================
89000 iterations have been completed!
-> Now loss = tensor(0.3148, grad_fn=<NllLossBackward0>)
=========================================
89500 iterations have been completed!
-> Now loss = tensor(0.3148, grad_fn=<NllLossBackward0>)
=========================================
90000 iterations have been completed!
-> Now loss = tensor(0.3148, grad_fn=<NllLossBackward0>)
=========================================
90500 iterations have been completed!
-> Now loss = tensor(0.3148, grad_fn=<NllLossBackward0>)
=========================================
91000 iterations have been completed!
-> Now loss = tensor(0.3148, grad_fn=<NllLossBackward0>)
=========================================
91500 iterations have been completed!
-> Now loss = tensor(0.3148, grad_fn=<NllLossBackward0>)
=========================================
92000 iterations have been completed!
-> Now loss = tensor(0.3148, grad_fn=<NllLossBackward0>)
=========================================
92500 iterations have been completed!
-> Now loss = tensor(0.3148, grad_fn=<NllLossBackward0>)
=========================================
93000 iterations have been completed!
-> Now loss = tensor(0.3148, grad_fn=<NllLossBackward0>)
=========================================
93500 iterations have been completed!
-> Now loss = tensor(0.3148, grad_fn=<NllLossBackward0>)
=========================================
94000 iterations have been completed!
-> Now loss = tensor(0.3148, grad_fn=<NllLossBackward0>)
=========================================
94500 iterations have been completed!
-> Now loss = tensor(0.3148, grad_fn=<NllLossBackward0>)
=========================================
95000 iterations have been completed!
-> Now loss = tensor(0.3148, grad_fn=<NllLossBackward0>)
=========================================
95500 iterations have been completed!
-> Now loss = tensor(0.3147, grad_fn=<NllLossBackward0>)
=========================================
96000 iterations have been completed!
-> Now loss = tensor(0.3147, grad_fn=<NllLossBackward0>)
=========================================
96500 iterations have been completed!
-> Now loss = tensor(0.3147, grad_fn=<NllLossBackward0>)
=========================================
97000 iterations have been completed!
-> Now loss = tensor(0.3147, grad_fn=<NllLossBackward0>)
=========================================
97500 iterations have been completed!
-> Now loss = tensor(0.3147, grad_fn=<NllLossBackward0>)
=========================================
98000 iterations have been completed!
-> Now loss = tensor(0.3147, grad_fn=<NllLossBackward0>)
=========================================
98500 iterations have been completed!
-> Now loss = tensor(0.3147, grad_fn=<NllLossBackward0>)
=========================================
99000 iterations have been completed!
-> Now loss = tensor(0.3147, grad_fn=<NllLossBackward0>)
=========================================
99500 iterations have been completed!
-> Now loss = tensor(0.3147, grad_fn=<NllLossBackward0>)
=========================================
100000 iterations have been completed!
-> Now loss = tensor(0.3147, grad_fn=<NllLossBackward0>)
=========================================
100500 iterations have been completed!
-> Now loss = tensor(0.3147, grad_fn=<NllLossBackward0>)
=========================================
101000 iterations have been completed!
-> Now loss = tensor(0.3147, grad_fn=<NllLossBackward0>)
=========================================
101500 iterations have been completed!
-> Now loss = tensor(0.3147, grad_fn=<NllLossBackward0>)
=========================================
102000 iterations have been completed!
-> Now loss = tensor(0.3147, grad_fn=<NllLossBackward0>)
=========================================
102500 iterations have been completed!
-> Now loss = tensor(0.3147, grad_fn=<NllLossBackward0>)
=========================================
103000 iterations have been completed!
-> Now loss = tensor(0.3147, grad_fn=<NllLossBackward0>)
=========================================
103500 iterations have been completed!
-> Now loss = tensor(0.3147, grad_fn=<NllLossBackward0>)
=========================================
104000 iterations have been completed!
-> Now loss = tensor(0.3147, grad_fn=<NllLossBackward0>)
=========================================
104500 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
105000 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
105500 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
106000 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
106500 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
107000 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
107500 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
108000 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
108500 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
109000 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
109500 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
110000 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
110500 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
111000 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
111500 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
112000 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
112500 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
113000 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
113500 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
114000 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
114500 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
115000 iterations have been completed!
-> Now loss = tensor(0.3146, grad_fn=<NllLossBackward0>)
=========================================
115500 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
116000 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
116500 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
117000 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
117500 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
118000 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
118500 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
119000 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
119500 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
120000 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
120500 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
121000 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
121500 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
122000 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
122500 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
123000 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
123500 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
124000 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
124500 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
125000 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
125500 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
126000 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
126500 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
127000 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
127500 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
128000 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
128500 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
129000 iterations have been completed!
-> Now loss = tensor(0.3145, grad_fn=<NllLossBackward0>)
=========================================
129500 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
130000 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
130500 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
131000 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
131500 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
132000 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
132500 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
133000 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
133500 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
134000 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
134500 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
135000 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
135500 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
136000 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
136500 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
137000 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
137500 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
138000 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
138500 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
139000 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
139500 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
140000 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
140500 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
141000 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
141500 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
142000 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
142500 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
143000 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
143500 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
144000 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
144500 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
145000 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
145500 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
146000 iterations have been completed!
-> Now loss = tensor(0.3144, grad_fn=<NllLossBackward0>)
=========================================
146500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
147000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
147500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
148000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
148500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
149000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
149500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
150000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
150500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
151000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
151500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
152000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
152500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
153000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
153500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
154000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
154500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
155000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
155500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
156000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
156500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
157000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
157500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
158000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
158500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
159000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
159500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
160000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
160500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
161000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
161500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
162000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
162500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
163000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
163500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
164000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
164500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
165000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
165500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
166000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
166500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
167000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
167500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
168000 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
168500 iterations have been completed!
-> Now loss = tensor(0.3143, grad_fn=<NllLossBackward0>)
=========================================
169000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
169500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
170000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
170500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
171000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
171500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
172000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
172500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
173000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
173500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
174000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
174500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
175000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
175500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
176000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
176500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
177000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
177500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
178000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
178500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
179000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
179500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
180000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
180500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
181000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
181500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
182000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
182500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
183000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
183500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
184000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
184500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
185000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
185500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
186000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
186500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
187000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
187500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
188000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
188500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
189000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
189500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
190000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
190500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
191000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
191500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
192000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
192500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
193000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
193500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
194000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
194500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
195000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
195500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
196000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
196500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
197000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
197500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
198000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
198500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
199000 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
199500 iterations have been completed!
-> Now loss = tensor(0.3142, grad_fn=<NllLossBackward0>)
=========================================
200000 iterations have been completed!
-> Now loss = tensor(0.3141, grad_fn=<NllLossBackward0>)
=========================================
4. Visualization of the Cross Entropy Loss Function¶
plt.figure(figsize = (12, 6))
length = len(loss_list)
print("The length of loss_list is:", length)
plt.plot(np.arange(1, 20001, 1), [loss.detach().numpy() for loss in loss_list[0:20000]], "black")
plt.xlabel("epoch")
plt.ylabel("loss")
plt.show()
The length of loss_list is: 200000
# 开始测试
X_vec_test = torch.tensor(DataTest[["x1","x2"]].values, dtype=torch.float32) # N*2
y_vec_test = torch.tensor(DataTest["y"].values, dtype=torch.int64).reshape(-1, 1) # N*1
pred = DNN_Model.forward(X_vec_test)
pred_vec = pred [:, 1]
pred_vec[pred_vec > 0.50] = 1
pred_vec[pred_vec <= 0.50 ] = 0
pred_vec
tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.,
1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.,
1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.,
1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.,
1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.,
1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.,
1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.,
1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.,
1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.,
1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.,
1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.,
1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.,
1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.,
1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 0., 0.,
0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,
0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
grad_fn=<AsStridedBackward0>)
y_pred_np = y_vec_test.detach().numpy()
y_pred_np = np.squeeze(y_pred_np)
print("Shape of y_pred_p:", y_pred_np.shape)
pred_vec_np = pred_vec.detach().numpy()
pred_vec_np = np. squeeze (pred_vec_np)
print ("Shape of y_pred_p:", pred_vec_np.shape)
Shape of y_pred_p: (500,)
Shape of y_pred_p: (500,)
accuracy = accuracy_score (y_vec_test, y_pred_np)
print ("The accuracy score is:", accuracy)
The accuracy score is: 1.0