Pytorch极简入门教程（六）——多层感知机

# -*- coding: utf-8 -*-
import torch
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from torch import nn
data = pd.read_csv("dataset/HR.csv")
print("data.head():\t", data.head())
data.info()
print("data.info():\t", data.info())
data.part.unique() # 查看part里面的东西
print("data.part.unique()", data.part.unique())
data.salary.unique() # 查看salary里面存在的东西
print("data.salary.unique()", data.salary.unique())
# 分组查看
data.groupby(["salary", "part"]).size()
print("data.groupby(['salary', 'part']).size()",data.groupby(['salary', 'part']).size())
# 将特征转换成数值化（独热编码）
pd.get_dummies(data.salary)
print("pd.get_dummies(data.salary):\t", pd.get_dummies(data.salary))

# 将salary转换成独热编码添加至原有数据
data=data.join(pd.get_dummies(data.salary))
print("data.head()):\t", data.head())
# 将part转换成独热编码添加至原有数据
data=data.join(pd.get_dummies(data.part)) # 10个职业 独热编码扩展10位
# 删除"salary"的特征
del data["salary"]
del data["part"]
data.head()
print("data.head():\t", data.head())

# 查看是否会离职数据
data.left.value_counts()  # value_counts() 值计数的方式
print("data.left.value_counts():\n", data.left.value_counts())

Y_data = data.left.values.reshape(-1, 1)
print("Y_data.shape:\t", Y_data.shape)

# Tensor 查看行传可以用.size()或者.shape 但是
Y =torch.from_numpy(Y_data).type(torch.float32)
print("Y.shape", Y.shape) #  == print("Y.size()", Y.size())
# 将不是标签的全部元素组成列表
"""
M = [c for c in data.columns if c!= "left"] 
print("M:\t", X_data)
"""
X_data =data[[c for c in data.columns if c != 'left']].values # 取出列表的中各个元素所对应的值
"""
    两种方式进行数据类型转换
    如果numpy上转换 则用.astype(np.float32)
    如果torch上转换 则用.type(torch.float32)
"""
X = torch.from_numpy(X_data.astype(np.float32))
#X =torch.from_numpy(X_data).type(torch.float32)
print("X:\t", X)

print("X.size():\t", X.shape)   # X.size()和X.shape等价
"""""""""""""""""""""""""""""""""""""""""""""""
创建模型：
from torch import nn
自定义模型：
nn.Module: 继承这个类
__init__：初始化所有的层
forward: 定义模型的运算过程 （前向传播的过程）
"""""""""""""""""""""""""""""""""""""""""""""""
"""
# 自定义类 方法一
class Model(nn.Module):
    def __init__(self):
        super().__init__()
        self.liner_1 = nn.Linear(20, 64)
        self.liner_2 = nn.Linear(64, 64)
        self.liner_3 = nn.Linear(64, 1)
        self.relu = nn.ReLU()  # 初始化relu
        self.sigmoid = nn.Sigmoid() # 初始化sigmoid

    def forward(self, input):
        x = self.Liner_1(input)
        x = self.relu(x)
        x = self.Liner_2(x)
        x = self.rele(x)
        x = self.Liner_3(x)
        x = self.sigmod(x)
        return x
"""

"""""""""""""""""""""""""""""""""""
方法的改写： 方法二
import torch.nn.functional as F
class Model(nn.Module):
    def __init__(self):
        super().__init__()
        self.liner_1 = nn.Linear(20, 64)
        self.liner_2 = nn.Linear(64, 64)
        self.liner_3 = nn.Linear(64, 1)
      

    def forward(self, input):
        x = F.relu(self.Liner_1(input))
        x = F.relu(self.Liner_2(x))
        x = F.sigmoid(self.Liner_3(x))
        return x
"""""""""""""""""""""""""""""""""""
import torch.nn.functional as F
class Model(nn.Module):
    def __init__(self):
        super().__init__()
        self.Liner_1 = nn.Linear(20, 64)
        self.Liner_2 = nn.Linear(64, 64)
        self.Liner_3 = nn.Linear(64, 1)
    def forward(self, input):
        x = F.relu(self.Liner_1(input))
        x = F.relu(self.Liner_2(x))
        x = F.sigmoid(self.Liner_3(x))
        return x
"""
model = Model() # 模型的实例化
print("model:\t", model)
"""
lr = 0.0001
def get_model():
    model = Model()
    opt = torch.optim.Adam(model.parameters(), lr=lr)
    return model, opt

model, optim = get_model() # return  返回model、optim
"""
定义损失函数
"""
loss_fn = nn.BCELoss()
# 定义优化器

batch = 64
no_of_batch = len(data)//batch
epochs = 100

for epoch in range(epochs):
    for i in range(no_of_batch):
        start = i*batch
        end = start+batch
        x = X[start: end]
        y = Y[start: end]
        y_pred=model(x)
        loss =loss_fn(y_pred, y)
        optim.zero_grad()
        loss.backward()
        optim.step()    # 每一个参数进行优化
        """
        因为以下步骤不需要计算损失
        torch类型 查询它的值用.data  item()代表将单一元素的Tensor类型转换成数值型的标量
        numpy和pandas类型的数据查询它的值用.values
        """
    with torch.no_grad():
        print("epoch:",  epoch, "loss:", loss_fn(model(X), Y).data.item())

# 查看损失值
loss_fn(model(X), Y)
print("loss_fn(model(X), Y):\t", loss_fn(model(X), Y))

data.head():	    satisfaction_level  last_evaluation  ...   part  salary
0                0.38             0.53  ...  sales     low
1                0.80             0.86  ...  sales  medium
2                0.11             0.88  ...  sales  medium
3                0.72             0.87  ...  sales     low
4                0.37             0.52  ...  sales     low

[5 rows x 10 columns]
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 14999 entries, 0 to 14998
Data columns (total 10 columns):
 #   Column                 Non-Null Count  Dtype  
---  ------                 --------------  -----  
 0   satisfaction_level     14999 non-null  float64
 1   last_evaluation        14999 non-null  float64
 2   number_project         14999 non-null  int64  
 3   average_montly_hours   14999 non-null  int64  
 4   time_spend_company     14999 non-null  int64  
 5   Work_accident          14999 non-null  int64  
 6   left                   14999 non-null  int64  
 7   promotion_last_5years  14999 non-null  int64  
 8   part                   14999 non-null  object 
 9   salary                 14999 non-null  object 
dtypes: float64(2), int64(6), object(2)
memory usage: 1.1+ MB
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 14999 entries, 0 to 14998
Data columns (total 10 columns):
 #   Column                 Non-Null Count  Dtype  
---  ------                 --------------  -----  
 0   satisfaction_level     14999 non-null  float64
 1   last_evaluation        14999 non-null  float64
 2   number_project         14999 non-null  int64  
 3   average_montly_hours   14999 non-null  int64  
 4   time_spend_company     14999 non-null  int64  
 5   Work_accident          14999 non-null  int64  
 6   left                   14999 non-null  int64  
 7   promotion_last_5years  14999 non-null  int64  
 8   part                   14999 non-null  object 
 9   salary                 14999 non-null  object 
dtypes: float64(2), int64(6), object(2)
memory usage: 1.1+ MB
data.info():	 None
data.part.unique() ['sales' 'accounting' 'hr' 'technical' 'support' 'management' 'IT'
 'product_mng' 'marketing' 'RandD']
data.salary.unique() ['low' 'medium' 'high']
data.groupby(['salary', 'part']).size() salary  part       
high    IT               83
        RandD            51
        accounting       74
        hr               45
        management      225
        marketing        80
        product_mng      68
        sales           269
        support         141
        technical       201
low     IT              609
        RandD           364
        accounting      358
        hr              335
        management      180
        marketing       402
        product_mng     451
        sales          2099
        support        1146
        technical      1372
medium  IT              535
        RandD           372
        accounting      335
        hr              359
        management      225
        marketing       376
        product_mng     383
        sales          1772
        support         942
        technical      1147
dtype: int64
pd.get_dummies(data.salary):	        high  low  medium
0         0    1       0
1         0    0       1
2         0    0       1
3         0    1       0
4         0    1       0
...     ...  ...     ...
14994     0    1       0
14995     0    1       0
14996     0    1       0
14997     0    1       0
14998     0    1       0

[14999 rows x 3 columns]
data.head()):	    satisfaction_level  last_evaluation  number_project  ...  high  low  medium
0                0.38             0.53               2  ...     0    1       0
1                0.80             0.86               5  ...     0    0       1
2                0.11             0.88               7  ...     0    0       1
3                0.72             0.87               5  ...     0    1       0
4                0.37             0.52               2  ...     0    1       0

[5 rows x 13 columns]
data.head():	    satisfaction_level  last_evaluation  ...  support  technical
0                0.38             0.53  ...        0          0
1                0.80             0.86  ...        0          0
2                0.11             0.88  ...        0          0
3                0.72             0.87  ...        0          0
4                0.37             0.52  ...        0          0

[5 rows x 21 columns]
data.left.value_counts():
 0    11428
1     3571
Name: left, dtype: int64
Y_data.shape:	 (14999, 1)
Y.shape torch.Size([14999, 1])
X:	 tensor([[0.3800, 0.5300, 2.0000,  ..., 1.0000, 0.0000, 0.0000],
        [0.8000, 0.8600, 5.0000,  ..., 1.0000, 0.0000, 0.0000],
        [0.1100, 0.8800, 7.0000,  ..., 1.0000, 0.0000, 0.0000],
        ...,
        [0.3700, 0.5300, 2.0000,  ..., 0.0000, 1.0000, 0.0000],
        [0.1100, 0.9600, 6.0000,  ..., 0.0000, 1.0000, 0.0000],
        [0.3700, 0.5200, 2.0000,  ..., 0.0000, 1.0000, 0.0000]])
X.size():	 torch.Size([14999, 20])
E:\Professional Software\Anconda\envs\pytracking\lib\site-packages\torch\nn\functional.py:1350: UserWarning: nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.
  warnings.warn("nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.")
epoch: 0 loss: 0.7312576770782471
epoch: 1 loss: 0.7727270722389221
epoch: 2 loss: 0.7783536314964294
epoch: 3 loss: 0.7793934345245361
epoch: 4 loss: 0.7771381139755249
epoch: 5 loss: 0.7743889689445496
epoch: 6 loss: 0.7647398114204407
epoch: 7 loss: 0.7545432448387146
epoch: 8 loss: 0.7476876974105835
epoch: 9 loss: 0.7408889532089233
epoch: 10 loss: 0.7297168374061584
epoch: 11 loss: 0.7265633344650269
epoch: 12 loss: 0.8101921081542969
epoch: 13 loss: 0.7332798838615417
epoch: 14 loss: 0.716731607913971
epoch: 15 loss: 0.7073214650154114
epoch: 16 loss: 0.696769654750824
epoch: 17 loss: 0.690093457698822
epoch: 18 loss: 0.6791982054710388
epoch: 19 loss: 0.6695009469985962
epoch: 20 loss: 0.6602796912193298
epoch: 21 loss: 0.6475608944892883
epoch: 22 loss: 0.639624297618866
epoch: 23 loss: 0.6317117214202881
epoch: 24 loss: 0.6239936351776123
epoch: 25 loss: 0.6263017058372498
epoch: 26 loss: 0.6041481494903564
epoch: 27 loss: 0.5972230434417725
epoch: 28 loss: 0.5907231569290161
epoch: 29 loss: 0.585003137588501
epoch: 30 loss: 0.5898972153663635
epoch: 31 loss: 0.5837881565093994
epoch: 32 loss: 0.5768670439720154
epoch: 33 loss: 0.5715934038162231
epoch: 34 loss: 0.5673941969871521
epoch: 35 loss: 0.5641562342643738
epoch: 36 loss: 0.5610777139663696
epoch: 37 loss: 0.5590518116950989
epoch: 38 loss: 0.5567962527275085
epoch: 39 loss: 0.5620077252388
epoch: 40 loss: 0.5546135306358337
epoch: 41 loss: 0.5531746745109558
epoch: 42 loss: 0.55274897813797
epoch: 43 loss: 0.5511655807495117
epoch: 44 loss: 0.5524853467941284
epoch: 45 loss: 0.5519489645957947
epoch: 46 loss: 0.5520744323730469
epoch: 47 loss: 0.5514076352119446
epoch: 48 loss: 0.5520205497741699
epoch: 49 loss: 0.5527095794677734
epoch: 50 loss: 0.5532509684562683
epoch: 51 loss: 0.5529757738113403
epoch: 52 loss: 0.5533396005630493
epoch: 53 loss: 0.5538538098335266
epoch: 54 loss: 0.5545904636383057
epoch: 55 loss: 0.5555252432823181
epoch: 56 loss: 0.5529634952545166
epoch: 57 loss: 0.5549248456954956
epoch: 58 loss: 0.5557812452316284
epoch: 59 loss: 0.5557355284690857
epoch: 60 loss: 0.5564326643943787
epoch: 61 loss: 0.5556958317756653
epoch: 62 loss: 0.5536842942237854
epoch: 63 loss: 0.5552572011947632
epoch: 64 loss: 0.5563111305236816
epoch: 65 loss: 0.5552616715431213
epoch: 66 loss: 0.5558472275733948
epoch: 67 loss: 0.5550410151481628
epoch: 68 loss: 0.5725806951522827
epoch: 69 loss: 0.5583928227424622
epoch: 70 loss: 0.5544316172599792
epoch: 71 loss: 0.5551984906196594
epoch: 72 loss: 0.5552375912666321
epoch: 73 loss: 0.5546104907989502
epoch: 74 loss: 0.5539920926094055
epoch: 75 loss: 0.5583476424217224
epoch: 76 loss: 0.575699508190155
epoch: 77 loss: 0.556276798248291
epoch: 78 loss: 0.5547029972076416
epoch: 79 loss: 0.5523102879524231
epoch: 80 loss: 0.5545598864555359
epoch: 81 loss: 0.5762888193130493
epoch: 82 loss: 0.5546205043792725
epoch: 83 loss: 0.5516649484634399
epoch: 84 loss: 0.5519871115684509
epoch: 85 loss: 0.550129234790802
epoch: 86 loss: 0.5504892468452454
epoch: 87 loss: 0.5495496988296509
epoch: 88 loss: 0.5545125603675842
epoch: 89 loss: 0.5482568144798279
epoch: 90 loss: 0.566870927810669
epoch: 91 loss: 0.5604091882705688
epoch: 92 loss: 0.5596952438354492
epoch: 93 loss: 0.5566151738166809
epoch: 94 loss: 0.5397318601608276
epoch: 95 loss: 0.5474585890769958
epoch: 96 loss: 0.5286720395088196
epoch: 97 loss: 0.5305213332176208
epoch: 98 loss: 0.5286723375320435
epoch: 99 loss: 0.525821328163147
loss_fn(model(X), Y):	 tensor(0.5258, grad_fn=<BinaryCrossEntropyBackward>)