学习深度学习、机器学习、数据分析与科学计算的小伙伴们对数据可视化的需求是比较重视的。所以Python的第三方库种有很多Python可视化工具,比如常用的TensorboardX,而今天我要介绍的python可视化工具是与其性能效果旗鼓相当Visdom,在网上已经有比较多的简单的visdom的使用介绍了,今天俺就结合网上教程系统化地整理一下常用的一些Visdom的使用方法叭。
Visdom:一个灵活的可视化工具,可用来对于 实时,富数据的 创建,组织和共享。支持Torch和Numpy还有pytorch。
Visdom 可以实现远程数据的可视化,对科学实验有很大帮助。我们可以远程的发送图片和数据,并进行在ui界面显示出来,检查实验结果,或者debug.
要用这个先要安装,对于python模块而言,安装都是蛮简单的直接pip 或者 conda 包安装:
pip install visdom
安装完每次要用直接输入代码打开:
python -m visdom.server
然后根据提示在浏览器中输入相应地址即可,默认地址为:http://localhost:8097/
import visdom # 添加visdom库
import numpy as np # 添加numpy库
vis = visdom.Visdom(env='test') # 设置环境窗口的名称,如果不设置名称就默认为main
vis.text('test', win='main') # 使用文本输出
vis.image(np.ones((3, 100, 100))) # 绘制一幅尺寸为3 * 100 * 100的图片,图片的像素值全部为1
其中:
visdom.Visdom(env=‘命名新环境') vis.text(‘文本', win=‘环境名') vis.image(‘图片',win=‘环境名')
一条
import visdom
import numpy as np
vis = visdom.Visdom(env='my_windows') # 设置环境窗口的名称,如果不设置名称就默认为main
x = list(range(10))
y = list(range(10))
# 使用line函数绘制直线 并选择显示坐标轴
vis.line(X=np.array(x), Y=np.array(y), opts=dict(showlegend=True))
vis.line([x], [y], opts=dict(showlegend=True)[展示说明])
两条
import visdom
import numpy as np
vis = visdom.Visdom(env='my_windows')
x = list(range(10))
y = list(range(10))
z = list(range(1,11))
vis.line(X=np.array(x), Y=np.column_stack((np.array(y), np.array(z))), opts=dict(showlegend=True))
vis.line([x], [y=np.column_stack((np.array(y),np.array(z),np.array(还可以增加)))]) np.column_stack(a,b), 表示两个矩阵按列合并
import visdom
import torch
vis = visdom.Visdom(env='sin')
x = torch.arange(0, 100, 0.1)
y = torch.sin(x)
vis.line(X=x,Y=y,win='sin(x)',opts=dict(showlegend=True))
import visdom
import numpy as np
vis = visdom.Visdom(env='my_windows')
# 利用update更新图像
x = 0
y = 0
my_win = vis.line(X=np.array([x]), Y=np.array([y]), opts=dict(title='Update'))
for i in range(10):
x += 1
y += i
vis.line(X=np.array([x]), Y=np.array([y]), win=my_win, update='append')
使用“append”追加数据,“replace”使用新数据,“remove”用于删除“name”中指定的跟踪。
import visdom
import torch
# 新建一个连接客户端
# 指定env = 'test1',默认是'main',注意在浏览器界面做环境的切换
vis = visdom.Visdom(env='test1')
# 绘制正弦函数
x = torch.arange(1, 100, 0.01)
y = torch.sin(x)
vis.line(X=x,Y=y, win='sinx',opts={'title':'y=sin(x)'})
# 绘制36张图片随机的彩色图片
vis.images(torch.randn(36,3,64,64).numpy(),nrow=6, win='imgs',opts={'title':'imgs'})
#绘制loss变化趋势,参数一为Y轴的值,参数二为X轴的值,参数三为窗体名称,参数四为表格名称,参数五为更新选项,从第二个点开始可以更新
vis.line(Y=np.array([totalloss.item()]), X=np.array([traintime]),
win=('train_loss'),
opts=dict(title='train_loss'),
update=None if traintime == 0 else 'append'
)
此代码出自CycleGAN的 utils.py 里一个实现
# 记录训练日志,显示生成图,画loss曲线 的类
class Logger():
def __init__(self, n_epochs, batches_epoch):
'''
:param n_epochs: 跑多少个epochs
:param batches_epoch: 一个epoch有几个batches
'''
self.viz = Visdom() # 默认env是main函数
self.n_epochs = n_epochs
self.batches_epoch = batches_epoch
self.epoch = 1 # 当前epoch数
self.batch = 1 # 当前batch数
self.prev_time = time.time()
self.mean_period = 0
self.losses = {}
self.loss_windows = {} # 保存loss图的字典集合
self.image_windows = {} # 保存生成图的字典集合
def log(self, losses=None, images=None):
self.mean_period += (time.time() - self.prev_time)
self.prev_time = time.time()
sys.stdout.write('
Epoch %03d/%03d [%04d/%04d] -- ' % (self.epoch, self.n_epochs, self.batch, self.batches_epoch))
for i, loss_name in enumerate(losses.keys()):
if loss_name not in self.losses:
self.losses[loss_name] = losses[loss_name].data.item() #这里losses[loss_name].data是个tensor(包在值外面的数据结构),要用item方法取值
else:
self.losses[loss_name] = losses[loss_name].data.item()
if (i + 1) == len(losses.keys()):
sys.stdout.write('%s: %.4f -- ' % (loss_name, self.losses[loss_name]/self.batch))
else:
sys.stdout.write('%s: %.4f | ' % (loss_name, self.losses[loss_name]/self.batch))
batches_done = self.batches_epoch * (self.epoch - 1) + self.batch
batches_left = self.batches_epoch * (self.n_epochs - self.epoch) + self.batches_epoch - self.batch
sys.stdout.write('ETA: %s' % (datetime.timedelta(seconds=batches_left*self.mean_period/batches_done)))
# 显示生成图
for image_name, tensor in images.items(): # 字典.items()是以list形式返回键值对
if image_name not in self.image_windows:
self.image_windows[image_name] = self.viz.image(tensor2image(tensor.data), opts={'title':image_name})
else:
self.viz.image(tensor2image(tensor.data), win=self.image_windows[image_name], opts={'title':image_name})
# End of each epoch
if (self.batch % self.batches_epoch) == 0: # 一个epoch结束时
# 绘制loss曲线图
for loss_name, loss in self.losses.items():
if loss_name not in self.loss_windows:
self.loss_windows[loss_name] = self.viz.line(X=np.array([self.epoch]), Y=np.array([loss/self.batch]),
opts={'xlabel':'epochs', 'ylabel':loss_name, 'title':loss_name})
else:
self.viz.line(X=np.array([self.epoch]), Y=np.array([loss/self.batch]), win=self.loss_windows[loss_name], update='append') #update='append'可以使loss图不断更新
# 每个epoch重置一次loss
self.losses[loss_name] = 0.0
# 跑完一个epoch,更新一下下面参数
self.epoch += 1
self.batch = 1
sys.stdout.write('
')
else:
self.batch += 1
train.py中调用代码是
class Logger():
def __init__(self, n_epochs, batches_epoch):
'''
:param n_epochs: 跑多少个epochs
:param batches_epoch: 一个epoch有几个batches
'''
self.viz = Visdom() # 默认env是main函数
self.n_epochs = n_epochs
self.batches_epoch = batches_epoch
self.epoch = 1 # 当前epoch数
self.batch = 1 # 当前batch数
self.prev_time = time.time()
self.mean_period = 0
self.losses = {}
self.loss_windows = {} # 保存loss图的字典集合
self.image_windows = {} # 保存生成图的字典集合
def log(self, losses=None, images=None):
self.mean_period += (time.time() - self.prev_time)
self.prev_time = time.time()
sys.stdout.write('
Epoch %03d/%03d [%04d/%04d] -- ' % (self.epoch, self.n_epochs, self.batch, self.batches_epoch))
for i, loss_name in enumerate(losses.keys()):
if loss_name not in self.losses:
self.losses[loss_name] = losses[loss_name].data.item() #这里losses[loss_name].data是个tensor(包在值外面的数据结构),要用item方法取值
else:
self.losses[loss_name] = losses[loss_name].data.item()
if (i + 1) == len(losses.keys()):
sys.stdout.write('%s: %.4f -- ' % (loss_name, self.losses[loss_name]/self.batch))
else:
sys.stdout.write('%s: %.4f | ' % (loss_name, self.losses[loss_name]/self.batch))
batches_done = self.batches_epoch * (self.epoch - 1) + self.batch
batches_left = self.batches_epoch * (self.n_epochs - self.epoch) + self.batches_epoch - self.batch
sys.stdout.write('ETA: %s' % (datetime.timedelta(seconds=batches_left*self.mean_period/batches_done)))
# 显示生成图
for image_name, tensor in images.items(): # 字典.items()是以list形式返回键值对
if image_name not in self.image_windows:
self.image_windows[image_name] = self.viz.image(tensor2image(tensor.data), opts={'title':image_name})
else:
self.viz.image(tensor2image(tensor.data), win=self.image_windows[image_name], opts={'title':image_name})
# End of each epoch
if (self.batch % self.batches_epoch) == 0: # 一个epoch结束时
# 绘制loss曲线图
for loss_name, loss in self.losses.items():
if loss_name not in self.loss_windows:
self.loss_windows[loss_name] = self.viz.line(X=np.array([self.epoch]), Y=np.array([loss/self.batch]),
opts={'xlabel':'epochs', 'ylabel':loss_name, 'title':loss_name})
else:
self.viz.line(X=np.array([self.epoch]), Y=np.array([loss/self.batch]), win=self.loss_windows[loss_name], update='append') #update='append'可以使loss图不断更新
# 每个epoch重置一次loss
self.losses[loss_name] = 0.0
# 跑完一个epoch,更新一下下面参数
self.epoch += 1
self.batch = 1
sys.stdout.write('
')
else:
self.batch += 1
到此这篇常见的visdom使用方法的介绍就介绍到这了,希望能给到大家帮助啦。