This commit is contained in:
guanjihuan 2023-11-29 02:14:00 +08:00
parent 1608ae7437
commit 5a568076c7
3 changed files with 16 additions and 17 deletions

View File

@ -1,7 +1,7 @@
[metadata]
# replace with your username:
name = guan
version = 0.1.58
version = 0.1.59
author = guanjihuan
author_email = guanjihuan@163.com
description = An open source python package

View File

@ -1,6 +1,6 @@
Metadata-Version: 2.1
Name: guan
Version: 0.1.58
Version: 0.1.59
Summary: An open source python package
Home-page: https://py.guanjihuan.com
Author: guanjihuan

View File

@ -1,11 +1,10 @@
# Module: machine_learning
import guan
# 全连接神经网络模型(包含一个隐藏层)
# 全连接神经网络模型(包含一个隐藏层)(模型的类定义成全局的)
@guan.function_decorator
def fully_connected_neural_network_with_one_hidden_layer(input_size=1, hidden_size=10, output_size=1, activation='relu'):
import torch
# 如果在函数中定义模型类,尽量定义成全局的,这样可以防止在保存完整模型到文件时,无法访问函数中的模型类。
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):
@ -28,7 +27,7 @@ def fully_connected_neural_network_with_one_hidden_layer(input_size=1, hidden_si
model = model_class_of_fully_connected_neural_network_with_one_hidden_layer()
return model
# 全连接神经网络模型(包含两个隐藏层)
# 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)
@guan.function_decorator
def fully_connected_neural_network_with_two_hidden_layers(input_size=1, hidden_size_1=10, hidden_size_2=10, output_size=1, activation_1='relu', activation_2='relu'):
import torch
@ -67,7 +66,7 @@ def fully_connected_neural_network_with_two_hidden_layers(input_size=1, hidden_s
model = model_class_of_fully_connected_neural_network_with_two_hidden_layers()
return model
# 全连接神经网络模型(包含三个隐藏层)
# 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)
@guan.function_decorator
def fully_connected_neural_network_with_three_hidden_layers(input_size=1, hidden_size_1=10, hidden_size_2=10, hidden_size_3=10, output_size=1, activation_1='relu', activation_2='relu', activation_3='relu'):
import torch
@ -145,28 +144,28 @@ def train_model(model, x_data, y_data, optimizer='Adam', learning_rate=0.001, cr
print(epoch)
return model, losses
# 保存完整模型到文件
@guan.function_decorator
def save_model(model, filename='./model.pth'):
import torch
torch.save(model, filename)
# 保存模型参数到文件
@guan.function_decorator
def save_model_parameters(model, filename='./model_parameters.pth'):
import torch
torch.save(model.state_dict(), filename)
# 加载完整模型
# 保存完整模型到文件(保存时需要模型的类可访问)
@guan.function_decorator
def load_model(filename='./model.pth'):
def save_model(model, filename='./model.pth'):
import torch
model = torch.load(filename)
return model
torch.save(model, filename)
# 加载模型参数(需要输入模型
# 加载模型参数(需要输入模型,加载后,原输入的模型参数也会改变
@guan.function_decorator
def load_model_parameters(model, filename='./model_parameters.pth'):
import torch
model.load_state_dict(torch.load(filename))
return model
# 加载完整模型(不需要输入模型,但加载时需要原定义的模型的类可访问)
@guan.function_decorator
def load_model(filename='./model.pth'):
import torch
model = torch.load(filename)
return model