绘制模型训练损失变化图

跑了个网络，每个epoch的损失等信息会打印在日志里，但是损失变化通过数字来看不够直观。

那个日志的格式是这样的：

import numpy as np
from os import path
import os
import matplotlib.pyplot as plt
from matplotlib import cm,ticker
if __name__ == '__main__':
    path="tt/training_log_2021_3_31_09_50_38.txt"
    out = open(path, encoding='utf-8')
    lines = out.readlines()
    
    #提取trainLoss和validationLoss
    trainLoss=[]
    validationLoss=[]
    for line in lines:
        if "train loss" in line:
            val=np.float(line.split("loss : ")[-1][:-1]) #[:-1]是去除末尾'\n'
            trainLoss.append(val)
        if "validation loss" in line:
            val=np.float(line.split("loss: ")[-1][:-1])
            validationLoss.append(val)

    epochNum=len(trainLoss)
    for i in range(epochNum):
        print("epoch{}: train loss:{}  val loss:{}".format(i,trainLoss[i],validationLoss[i]))


    #绘图
    fig=plt.figure()
    xs=np.arange(epochNum)
    plt.yticks(np.arange(-1,0,0.1))
    plt.plot(xs, trainLoss, color='coral', label="train loss")
    plt.plot(xs, validationLoss, color='g', label="val loss")
    plt.legend()
    plt.show()
    #plt.savefig("loss.png")

得到图示：

OVER