Pytorch framework learning record 2 - the use of TensorBoard

Pytorch框架学习记录2——TensorBoard的使用

Tensorboard在训练模型时很有用,可以看训练过程中loss的变化.之前用于Tensorflow框架,自Pytorch1.1之后,Pytorch也加了这个功能.

1. TensorBoard的使用

从torch.utils.tensorboard中导入SummaryWriter类.

1.1 SummaryWriter类使用

SummaryWriter类可以在指定文件夹生成一个事件文件,这个事件文件可以对TensorBoard解析.

首先实例化一个SummaryWriter的类,The parameter represents the name of the saved folder

writer = SummaryWriter("logs")  #文件夹名称

1.2 writer.add_scalar()方法：

def add_scalar(
    self,
    tag,
    scalar_value,
    global_step=None,
    walltime=None,
    new_style=False,
    double_precision=False,
):

The above are the officially defined parameters,This method is meant to add a scalar.

tag：drawn icontitle,str类型,注意引号

scalar_value：需要保存的数值,对应y轴的y值

global_step：全局的步数,对应x轴

这里的tagPlay the role of uniquely identifying the image,If you need two pictures,is rewrittentag！！！

1.4 举例

绘制y = x 的图像.

from torch.utils.tensorboard import SummaryWriter


writer = SummaryWriter("logs")

for i in range(100):
    writer.add_scalar("y = x", i, i)

writer.close()

1.5 add_image()方法

def add_image(self, tag, img_tensor, global_step=None, walltime=None, dataformats='CHW'):
''' Args: tag (string): Data identifier img_tensor (torch.Tensor, numpy.array, or string/blobname): Image data global_step (int): Global step value to record walltime (float): Optional override default walltime (time.time()) seconds after epoch of event Shape: img_tensor: Default is :math:`(3, H, W)`. You can use ``torchvision.utils.make_grid()`` to convert a batch of tensor into 3xHxW format or call ``add_images`` and let us do the job. Tensor with :math:`(1, H, W)`, :math:`(H, W)`, :math:`(H, W, 3)` is also suitable as long as corresponding ``dataformats`` argument is passed, e.g. ``CHW``, ``HWC``, ``HW``. '''

其中,img_tensor参数类型要求为：torch.Tensor、numpy.array或者string类型.global_step为步骤,int类型.

img_tensorThe default image size format is （3,H,W）,But generally our picture format is （H,W,3),Therefore, the image format needs to be adjusted,调整的方法如下：设置dataformats='HWC'即可！！！

from torch.utils.tensorboard import SummaryWriter
from PIL import Image
import numpy as np

writer = SummaryWriter("logs")

img_path = "C:\\Users\\hp\\PycharmProjects\pythonProject\\Pytorch_Learning\\flower_data\\train\daisy\\5547758_eea9edfd54_n.jpg"
img_PIL = Image.open(img_path)
img_array = np.array(img_PIL)
print(type(img_array))
writer.add_image("test", img_array, 1, dataformats='HWC')

for i in range(100):
    writer.add_scalar("y = x", i, i)

writer.close()

1.6 Event file open method：

在PyCharm中,Find where the project isterminal,

tensorboard --logdir=The name of the folder where the event file is located --port=端口号

Just open it on the URL that pops up.

2. Common error cases

2.1 ImportError: TensorBoard logging requires TensorBoard version 1.15 or above

tensorboard版本过低,Reinstall the latest version after uninstalling.

2.2 AttributeError: module ‘tensorflow._api.v1.io’ has no attribute ‘gfile’

tensorflow 的版本和tensorboard的版本不匹配.升级tensorflow版本为2.0版本即可,Execute the following command to automatically upgradetensorflow版本和tensorboard版本：

pip install tensorflow==2.0

2.3 ModuleNotFoundError: No module named ‘grpc’

重新安装grpcio

pip uninstall grpcio

pip install grpcio

2.4 No dashboards are active for the current data set.

The open file location does not exist,Copy the absolute path to the folder location,然后重新执行terminal语句即可