guanjihuan/py.guanjihuan.com
1
1
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

0.1.110-显式传递参数至局域变量

Browse Source
This commit is contained in:
guanjihuan
2024-06-12 00:55:40 +08:00
parent 170cd45bb6
commit ddd2234689
3 changed files with 56 additions and 49 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
PyPI/src/guan.egg-info/PKG-INFO
View File

@@ -1,6 +1,6 @@
Metadata-Version: 2.1
Name: guan
Version: 0.1.109
Version: 0.1.110
Summary: An open source python package
Home-page: https://py.guanjihuan.com
Author: guanjihuan

101
PyPI/src/guan/machine_learning.py
View File

@@ -5,28 +5,29 @@ def fully_connected_neural_network_with_one_hidden_layer(input_size=1, hidden_si
import torch
global model_class_of_fully_connected_neural_network_with_one_hidden_layer
class model_class_of_fully_connected_neural_network_with_one_hidden_layer(torch.nn.Module):
def __init__(self):
def __init__(self, input_size, hidden_size, output_size, activation):
super().__init__()
self.hidden_layer = torch.nn.Linear(input_size, hidden_size)
self.output_layer = torch.nn.Linear(hidden_size, output_size)
self.activation = activation
def forward(self, x):
if activation == 'relu':
if self.activation == 'relu':
hidden_output = torch.nn.functional.relu(self.hidden_layer(x))
elif activation == 'leaky_relu':
elif self.activation == 'leaky_relu':
hidden_output = torch.nn.functional.leaky_relu(self.hidden_layer(x))
elif activation == 'sigmoid':
elif self.activation == 'sigmoid':
hidden_output = torch.nn.functional.sigmoid(self.hidden_layer(x))
elif activation == 'tanh':
elif self.activation == 'tanh':
hidden_output = torch.nn.functional.tanh(self.hidden_layer(x))
elif activation == 'gelu':
elif self.activation == 'gelu':
hidden_output = torch.nn.functional.gelu(self.hidden_layer(x))
elif activation == 'silu':
elif self.activation == 'silu':
hidden_output = torch.nn.functional.silu(self.hidden_layer(x))
else:
hidden_output = self.hidden_layer(x)
output = self.output_layer(hidden_output)
return output
model = model_class_of_fully_connected_neural_network_with_one_hidden_layer()
model = model_class_of_fully_connected_neural_network_with_one_hidden_layer(input_size, hidden_size, output_size, activation)
return model
# 全连接神经网络模型（包含两个隐藏层）（模型的类定义成全局的）
@@ -34,45 +35,47 @@ def fully_connected_neural_network_with_two_hidden_layers(input_size=1, hidden_s
import torch
global model_class_of_fully_connected_neural_network_with_two_hidden_layers
class model_class_of_fully_connected_neural_network_with_two_hidden_layers(torch.nn.Module):
def __init__(self):
def __init__(self, input_size, hidden_size_1, hidden_size_2, output_size, activation_1, activation_2):
super().__init__()
self.hidden_layer_1 = torch.nn.Linear(input_size, hidden_size_1)
self.hidden_layer_2 = torch.nn.Linear(hidden_size_1, hidden_size_2)
self.output_layer = torch.nn.Linear(hidden_size_2, output_size)
self.activation_1 = activation_1
self.activation_2 = activation_2
def forward(self, x):
if activation_1 == 'relu':
if self.activation_1 == 'relu':
hidden_output_1 = torch.nn.functional.relu(self.hidden_layer_1(x))
elif activation_1 == 'leaky_relu':
elif self.activation_1 == 'leaky_relu':
hidden_output_1 = torch.nn.functional.leaky_relu(self.hidden_layer_1(x))
elif activation_1 == 'sigmoid':
elif self.activation_1 == 'sigmoid':
hidden_output_1 = torch.nn.functional.sigmoid(self.hidden_layer_1(x))
elif activation_1 == 'tanh':
elif self.activation_1 == 'tanh':
hidden_output_1 = torch.nn.functional.tanh(self.hidden_layer_1(x))
elif activation_1 == 'gelu':
elif self.activation_1 == 'gelu':
hidden_output_1 = torch.nn.functional.gelu(self.hidden_layer_1(x))
elif activation_1 == 'silu':
elif self.activation_1 == 'silu':
hidden_output_1 = torch.nn.functional.silu(self.hidden_layer_1(x))
else:
hidden_output_1 = self.hidden_layer_1(x)
if activation_2 == 'relu':
if self.activation_2 == 'relu':
hidden_output_2 = torch.nn.functional.relu(self.hidden_layer_2(hidden_output_1))
elif activation_2 == 'leaky_relu':
elif self.activation_2 == 'leaky_relu':
hidden_output_2 = torch.nn.functional.leaky_relu(self.hidden_layer_2(hidden_output_1))
elif activation_2 == 'sigmoid':
elif self.activation_2 == 'sigmoid':
hidden_output_2 = torch.nn.functional.sigmoid(self.hidden_layer_2(hidden_output_1))
elif activation_2 == 'tanh':
elif self.activation_2 == 'tanh':
hidden_output_2 = torch.nn.functional.tanh(self.hidden_layer_2(hidden_output_1))
elif activation_2 == 'gelu':
elif self.activation_2 == 'gelu':
hidden_output_2 = torch.nn.functional.gelu(self.hidden_layer_2(hidden_output_1))
elif activation_2 == 'silu':
elif self.activation_2 == 'silu':
hidden_output_2 = torch.nn.functional.silu(self.hidden_layer_2(hidden_output_1))
else:
hidden_output_2 = self.hidden_layer_2(hidden_output_1)
output = self.output_layer(hidden_output_2)
return output
model = model_class_of_fully_connected_neural_network_with_two_hidden_layers()
model = model_class_of_fully_connected_neural_network_with_two_hidden_layers(input_size, hidden_size_1, hidden_size_2, output_size, activation_1, activation_2)
return model
# 全连接神经网络模型（包含三个隐藏层）（模型的类定义成全局的）
@@ -80,61 +83,64 @@ def fully_connected_neural_network_with_three_hidden_layers(input_size=1, hidden
import torch
global model_class_of_fully_connected_neural_network_with_three_hidden_layers
class model_class_of_fully_connected_neural_network_with_three_hidden_layers(torch.nn.Module):
def __init__(self):
def __init__(self, input_size, hidden_size_1, hidden_size_2, hidden_size_3, output_size, activation_1, activation_2, activation_3):
super().__init__()
self.hidden_layer_1 = torch.nn.Linear(input_size, hidden_size_1)
self.hidden_layer_2 = torch.nn.Linear(hidden_size_1, hidden_size_2)
self.hidden_layer_3 = torch.nn.Linear(hidden_size_2, hidden_size_3)
self.output_layer = torch.nn.Linear(hidden_size_3, output_size)
self.activation_1 = activation_1
self.activation_2 = activation_2
self.activation_3 = activation_3
def forward(self, x):
if activation_1 == 'relu':
if self.activation_1 == 'relu':
hidden_output_1 = torch.nn.functional.relu(self.hidden_layer_1(x))
elif activation_1 == 'leaky_relu':
elif self.activation_1 == 'leaky_relu':
hidden_output_1 = torch.nn.functional.leaky_relu(self.hidden_layer_1(x))
elif activation_1 == 'sigmoid':
elif self.activation_1 == 'sigmoid':
hidden_output_1 = torch.nn.functional.sigmoid(self.hidden_layer_1(x))
elif activation_1 == 'tanh':
elif self.activation_1 == 'tanh':
hidden_output_1 = torch.nn.functional.tanh(self.hidden_layer_1(x))
elif activation_1 == 'gelu':
elif self.activation_1 == 'gelu':
hidden_output_1 = torch.nn.functional.gelu(self.hidden_layer_1(x))
elif activation_1 == 'silu':
elif self.activation_1 == 'silu':
hidden_output_1 = torch.nn.functional.silu(self.hidden_layer_1(x))
else:
hidden_output_1 = self.hidden_layer_1(x)
if activation_2 == 'relu':
if self.activation_2 == 'relu':
hidden_output_2 = torch.nn.functional.relu(self.hidden_layer_2(hidden_output_1))
elif activation_2 == 'leaky_relu':
elif self.activation_2 == 'leaky_relu':
hidden_output_2 = torch.nn.functional.leaky_relu(self.hidden_layer_2(hidden_output_1))
elif activation_2 == 'sigmoid':
elif self.activation_2 == 'sigmoid':
hidden_output_2 = torch.nn.functional.sigmoid(self.hidden_layer_2(hidden_output_1))
elif activation_2 == 'tanh':
elif self.activation_2 == 'tanh':
hidden_output_2 = torch.nn.functional.tanh(self.hidden_layer_2(hidden_output_1))
elif activation_2 == 'gelu':
elif self.activation_2 == 'gelu':
hidden_output_2 = torch.nn.functional.gelu(self.hidden_layer_2(hidden_output_1))
elif activation_2 == 'silu':
elif self.activation_2 == 'silu':
hidden_output_2 = torch.nn.functional.silu(self.hidden_layer_2(hidden_output_1))
else:
hidden_output_2 = self.hidden_layer_2(hidden_output_1)
if activation_3 == 'relu':
if self.activation_3 == 'relu':
hidden_output_3 = torch.nn.functional.relu(self.hidden_layer_3(hidden_output_2))
elif activation_3 == 'leaky_relu':
elif self.activation_3 == 'leaky_relu':
hidden_output_3 = torch.nn.functional.leaky_relu(self.hidden_layer_3(hidden_output_2))
elif activation_3 == 'sigmoid':
elif self.activation_3 == 'sigmoid':
hidden_output_3 = torch.nn.functional.sigmoid(self.hidden_layer_3(hidden_output_2))
elif activation_3 == 'tanh':
elif self.activation_3 == 'tanh':
hidden_output_3 = torch.nn.functional.tanh(self.hidden_layer_3(hidden_output_2))
elif activation_3 == 'gelu':
elif self.activation_3 == 'gelu':
hidden_output_3 = torch.nn.functional.gelu(self.hidden_layer_3(hidden_output_2))
elif activation_3 == 'silu':
elif self.activation_3 == 'silu':
hidden_output_3 = torch.nn.functional.silu(self.hidden_layer_3(hidden_output_2))
else:
hidden_output_3 = self.hidden_layer_3(hidden_output_2)
output = self.output_layer(hidden_output_3)
return output
model = model_class_of_fully_connected_neural_network_with_three_hidden_layers()
model = model_class_of_fully_connected_neural_network_with_three_hidden_layers(input_size, hidden_size_1, hidden_size_2, hidden_size_3, output_size, activation_1, activation_2, activation_3)
return model
# 卷积神经网络模型（包含两个卷积层和两个全连接层）（模型的类定义成全局的）
@@ -142,7 +148,7 @@ def convolutional_neural_network_with_two_convolutional_layers_and_two_fully_con
import torch
global model_class_of_convolutional_neural_network_with_two_convolutional_layers_and_two_fully_connected_layers
class model_class_of_convolutional_neural_network_with_two_convolutional_layers_and_two_fully_connected_layers(torch.nn.Module):
def __init__(self):
def __init__(self, in_channels, out_channels_1, out_channels_2, kernel_size_1, kernel_size_2, stride_1, stride_2, padding_1, padding_2, pooling, pooling_kernel_size, pooling_stride, input_size, hidden_size_1, hidden_size_2, output_size):
super().__init__()
self.convolutional_layer_1 = torch.nn.Conv2d(in_channels=in_channels, out_channels=out_channels_1, kernel_size=kernel_size_1, stride=stride_1, padding=padding_1)
self.convolutional_layer_2 = torch.nn.Conv2d(in_channels=out_channels_1, out_channels=out_channels_2, kernel_size=kernel_size_2, stride=stride_2, padding=padding_2)
@@ -150,8 +156,9 @@ def convolutional_neural_network_with_two_convolutional_layers_and_two_fully_con
self.hidden_layer_1 = torch.nn.Linear(input_size, hidden_size_1)
self.hidden_layer_2 = torch.nn.Linear(hidden_size_1, hidden_size_2)
self.output_layer = torch.nn.Linear(hidden_size_2, output_size)
self.pooling = pooling
def forward(self, x):
if pooling == 1:
if self.pooling == 1:
channel_output_1 = torch.nn.functional.relu(self.pooling_layer(self.convolutional_layer_1(x)))
channel_output_2 = torch.nn.functional.relu(self.pooling_layer(self.convolutional_layer_2(channel_output_1)))
else:
@@ -162,7 +169,7 @@ def convolutional_neural_network_with_two_convolutional_layers_and_two_fully_con
hidden_output_2 = torch.nn.functional.relu(self.hidden_layer_2(hidden_output_1))
output = self.output_layer(hidden_output_2)
return output
model = model_class_of_convolutional_neural_network_with_two_convolutional_layers_and_two_fully_connected_layers()
model = model_class_of_convolutional_neural_network_with_two_convolutional_layers_and_two_fully_connected_layers(in_channels, out_channels_1, out_channels_2, kernel_size_1, kernel_size_2, stride_1, stride_2, padding_1, padding_2, pooling, pooling_kernel_size, pooling_stride, input_size, hidden_size_1, hidden_size_2, output_size)
return model
# 从损失函数的变化情况中获取是否停止训练的信号
@@ -260,7 +267,7 @@ def save_model_with_torch_jit_script(model, filename='model_scripted_with_torch_
scripted_model = torch.jit.script(model)
scripted_model.save(filename)
# 以字典的形式保存模型的所有信息到文件（保存时需要模型的类可访问，此外还要输入模型的实例化函数）
# 以字典的形式保存模型的所有信息到文件（保存时需要模型的类可访问，此外还要输入模型的实例化函数。需要注意的是：该方法要求类和实例化函数都是独立可直接运行的模块）
def save_model_with_all_information(model, model_class, model_instantiation, note='', filename='./model_with_all_information.pth'):
import torch
import guan
@@ -293,7 +300,7 @@ def load_model_with_torch_jit_script(filename='model_scripted_with_torch_jit.pth
scripted_model = torch.jit.load(filename)
return scripted_model
# 加载包含所有信息的模型（包含了模型的类和实例化函数等，返回的是模型对象）
# 加载包含所有信息的模型（包含了模型的类和实例化函数等，返回的是模型对象。需要注意的是：该方法要求类和实例化函数都是独立可直接运行的模块）
def load_model_with_all_information(filename='./model_with_all_information.pth', note_print=0):
import torch
checkpoint = torch.load(filename)