python生成loss/acc训练曲线

根据训练模型的工作日志生成loss(acc)曲线

准备数据

利用excel将工作日志中的iter、loss、acc分别提出来单独放置文本文件中，获得iter.txt、acc.txt和loss.txt

loss/acc-iter曲线

编写代码

import matplotlib.pyplot as plt
import pylab as pl
import numpy as np

iter_text = np.loadtxt("data/iter.txt")
loss_text = np.loadtxt("data/loss.txt")

x = iter_text[:]
y = loss_text[:]

fig = plt.figure(figsize = (7, 5)) # figsize调节创建窗口的大小

p1 = pl.plot(x, y, 'r', label = u'loss')
pl.legend()

pl.xlabel(u'iters')
pl.ylabel(u'loss')

plt.title('Compare the data relationship')
plt.savefig('data/training.png')
pl.show()

生成图像

acc-loss镜像曲线

编写代码

import matplotlib.pyplot as plt
import pylab as pl
import numpy as np

iter_text = np.loadtxt("data/iter.txt")
acc_text = np.loadtxt("data/acc.txt")
loss_text = np.loadtxt("data/loss.txt")

x = iter_text[:]
y1 = loss_text[:]
y2 = acc_text[:]

fig , ax1 = plt.subplots(figsize = (9, 5))  # figsize调节创建窗口的尺寸大小
ax2 = ax1.twinx() # 共享x轴

p1 = ax1.plot(x, y1, 'r', label = u'loss')
ax1.legend()

p2 = ax2.plot(x, y2, 'b', label = u'acc')
ax2.legend()

pl.xlabel(u'iters')
ax1.set_ylabel(u'loss', color = 'r')  
ax2.set_ylabel(u'acc', color = 'b')

plt.title('Compare the data relationship')
plt.savefig('data/training1.png')
pl.show()

生成图像

参考文章

链接: link.
链接: link.