使用TensorBoard进行可视化

1 定义用于TensorBoard日志的目录

import os
root_logdir = os.path.join(os.curdir,"my_logs")

def get_run_logdir():
    import time
    run_id = time.strftime("run_%Y_%m_%d-%H_%M_%S")
    return os.path.join(root_logdir,run_id)

run_logdir = get_run_logdir() # 例：'./my_logs/run_2022_03_09-15_53_22'

2 类与函数

2.1 SummaryWriter类

torch.utils.tensorboard.writer.SummaryWriter(log_dir=无，注释=''，purge_step=无，max_queue=10，flush_secs=120，filename_suffix='')
功能：创建一个摘要编写器，该脚本将事件和摘要写出到事件文件中。

from torch.utils.tensorboard import SummaryWriter

# create a summary writer with automatically generated folder name.
writer = SummaryWriter()
# folder location: runs/May04_22-14-54_s-MacBook-Pro.local/

# create a summary writer using the specified folder name.
writer = SummaryWriter("my_experiment")
# folder location: my_experiment

# create a summary writer with comment appended.
writer = SummaryWriter(comment="LR_0.1_BATCH_16")
# folder location: runs/May04_22-14-54_s-MacBook-Pro.localLR_0.1_BATCH_16/

writer = SummaryWriter(run_logdir)

2.2 add_scalar()函数

add_scalar(标记中， scalar_value， global_step=无， walltime=无， new_style=假， double_precision=假)

功能：将标量数据添加到摘要。

from torch.utils.tensorboard import SummaryWriter
writer = SummaryWriter()
x = range(100)
for i in x:
    writer.add_scalar('y=2x', i * 2, i)
writer.close()

2.3 add_scalars()函数

add_scalars(main_tag，tag_scalar_dict，global_step=无，挂页时间=无)

功能：将许多标量数据添加到摘要中。

from torch.utils.tensorboard import SummaryWriter
writer = SummaryWriter()
r = 5
for i in range(100):
    writer.add_scalars('run_14h', {'xsinx':i*np.sin(i/r),
                                    'xcosx':i*np.cos(i/r),
                                    'tanx': np.tan(i/r)}, i)
writer.close()
# This call adds three values to the same scalar plot with the tag
# 'run_14h' in TensorBoard's scalar section.

2.4 分组画图

按层次结构命名来对图进行分组。

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

writer = SummaryWriter()

for n_iter in range(100):
    writer.add_scalar('Loss/train', np.random.random(), n_iter)
    writer.add_scalar('Loss/test', np.random.random(), n_iter)
    writer.add_scalar('Accuracy/train', np.random.random(), n_iter)
    writer.add_scalar('Accuracy/test', np.random.random(), n_iter)

3 启动TensorBoard服务器

$ tensorboard --logdir=./my_logs --port=6006