head.py

from pyedl.main import Tensor
from pyedl.nn import Sequential, Linear, MSELoss, Sigmoid, Tanh, CrossEntropyLoss
from pyedl.optim import SGD
from matplotlib import pyplot as plt
import numpy as np


def drawLossLs(lossList: list, title="PYEDL!!!"):
    """
    函数名称：drawLossLs

    函数功能：绘制损失列表的折线图

    参数列表：
    lossList: list，损失列表
    title: str，图表标题，默认为"PYEDL!!!"

    返回值：无

    示例：
    lossList = [0.5, 0.4, 0.3, 0.2, 0.1]
    drawLossLs(lossList, "训练损失变化图")

    """
    plt.rcParams["font.sans-serif"] = ["SimHei"]  # 设置字体为黑体
    plt.rcParams["axes.unicode_minus"] = False  # 解决负号'-'显示为方块的问题
    x = range(len(lossList))
    try:

        # 绘制折线图
        plt.plot(x, lossList, marker="*")
        # 添加标题和标签
        plt.title(title, fontproperties="KaiTi", fontsize=12)
        plt.xlabel("Epoch")
        plt.ylabel("Loss")
        # 显示图表
        plt.show()
    except TypeError:
        print("\n\n Wraning!!! 绘图时类型出错,程序正在尝试修复!!! \n\n")

        try:
            raise Exception
            lossList = [int(lossList[i].data) for i in range(len(lossList))]
            # 绘制折线图
            plt.plot(x, lossList, marker="*")
            print("看来修复成功了")
            # 添加标题和标签
            plt.title(title, fontproperties="KaiTi", fontsize=12)
            plt.xlabel("Epoch")
            plt.ylabel("Loss")

            # 显示图表
            plt.show()
            print("程序修复成功")

        except KeyboardInterrupt:
            exit(114514)
        except Exception as e:
            print("没办法了,自己修复吧")
            print(e)
            print(
                f"""
|--------------------------------------------------               
|    作者给你一点有用的信息:                         
|    type(lossList)={type(lossList)}               
|    type(lossList[0])={type(lossList[0])}    
|    lossList[0].data=  {lossList[0].data}  
|    type(lossList[0].data)={type(lossList[0].data)}
|--------------------------------------------------
                  """
            )


def easyTrain(
    data,
    target,
    model,
    criterion=MSELoss(),
    optim=None,
    epoch=10,
    returnLossList=True,
    printLoss=False,
):
    """
    函数名称：easyTrain

    函数功能：简单训练模型并返回损失列表

    参数列表：

    data: 输入数据
    target: 目标数据
    model: 模型
    criterion: 损失函数，默认为MSELoss()
    optim: 优化器，默认为None
    epoch: 训练轮数，默认为10
    returnLossList: 是否返回损失列表，默认为False
    printLoss: 是否打印损失，默认为False
    返回值：如果returnLossList为True，则返回损失列表，否则返回None

    示例：
    data = [1, 2, 3, 4, 5]
    target = [0.5, 1.5, 2.5, 3.5, 4.5]
    model = NeuralNetwork()
    criterion = MSELoss()
    loss = easyTrain(data, target, model, criterion, epoch=20, returnLossList=True, printLoss=True)
    """
    lossList = []
    if optim is None:
        optim = SGD(parameters=model.get_parameters(), alpha=0.01)
    print("============================Training================================")
    for i in range(epoch):
        pred = model.forward(data)  # 预测
        loss = criterion.forward(pred, target)  # 比较
        loss.backward(Tensor(np.ones_like(loss.data)))  # 学习
        optim.step()
        if printLoss:
            print(loss)
        if criterion is MSELoss():
            lossList.append(loss.data[0])
        elif criterion is CrossEntropyLoss():
            lossList.append(loss.data)
        else:
            lossList.append(loss.data)
    if returnLossList:
        return lossList