1. Tensorboard 简介

TensorBoard 是Google开发的一个机器学习可视化工具，它原本是TensorFlow中的模块，不过现在已经集成到了Pytorch中。它的功能主要包括
1. 跟踪和可视化损失及准确率等指标
2. 可视化模型图（操作和层）
3. 查看权重、偏差或其他张量随时间变化的直方图
4. 将嵌入投射到较低的维度空间
5. 显示图片、文字和音频数据
6. 剖析 TensorFlow 程序
TensorBoard 的工作原理和 Wandb 基本相同，本质也是一个网页服务，分成前端和后台两部分，两部分间是异步I/O的
- 后台程序将数据写入到本地文件中
- 前端程序读取本地文件中的数据来进行显示
由于 TensorBoard 已经集成到 Pytorch，无需再单独安装，直接 torch.utils.tensorboard 即可找到

2. 快速入门

2.1 运行方法

可以把 Tensorboard 的运行分成两步
1. 记录数据：使用 SummaryWriter 类实例数据要追踪的数据。每次运行时，该类对象首先会在给定目录log_dir中创建 “事件文件”（本次运行的数据仓库），然后在训练过程中我们可以利用其提供的一系列高级 API 向事件文件中异步地添加数据，从而实时地追踪数据变化。该类的定义如下
```
torch.utils.tensorboard.writer.SummaryWriter(log_dir=None, comment='', purge_step=None, max_queue=10, flush_secs=120, filename_suffix='')
```
  - log_dir (str) – 事件文件保存的目录位置。默认为runs / CURRENT_DATETIME_HOSTNAME，每次运行后都会更改。推挤使用分层文件夹结构，例如对于每个新实验，可以传递“runs / exp1”，“runs / exp2”等来进行比较。
  - comment (str) –在默认的log_dir后缀中的注释。如果已设置log_dir，则此参数无效。
  - purge_step (int) – 若在运行到第 $T + X$ 个 step 的时候由于各种原因（内存溢出）发生崩溃，那么当服务重启之后，就回退 $X$ 个 step，从 $T$ 时刻重新开始将数据写入文件。purge_step 参数就是设置的 $X$ ，这一段数据将被重新写入。注意，崩溃和恢复的实验应该具有相同的log_dir
  - max_queue (int) – 记录事件和摘要时在内存中开的队列的长度，当队列慢了之后就会把数据写入磁盘（文件）中。
  - flush_secs (int) – 将待处理事件和摘要刷新到磁盘的频率，以秒为单位，默认为每两分钟一次。
  - filename_suffix (str) – 添加到log_dir目录中所有事件文件名的后缀。有关文件名构造的更多详细信息，请参阅tensorboard.summary.writer.event_file_writer.EventFileWriter。
2. 启动网页服务显示数据：使用 tensorboard --logdir 数据文件夹 命令运行网页服务，其中 “数据文件夹” 应设置为之前实验时设置的 log_dir。若看到了如下输出 TensorBoard 2.8.0 at http://localhost:6006/ (Press CTRL+C to quit) 则说明启动成功，在浏览器打开相应 url 就能进入TensorBoard界面看到数据显示了。注意默认端口是 6006，如果想进一步指定网页服务端口，可以用 tensorboard --logdir=数据文件夹 --port=端口 命令

2.2 常用 API

我们使用 SummaryWriter 类提供的一系列 API 记录数据，比较常用的包括
1. add_hparams：以表格形式添加一组要比较的超参数
2. add_scalar：将标量数据添加到摘要中，用来画一条折线
3. add_scalars：将一组标量数据添加到摘要中，可以在同一张图内画多条折线
4. add_histogram：绘制直方图
5. add_image：将一张图片添加到摘要，需要 pillow 包
6. add_images：将一组图片添加到摘要，需要 pillow 包
7. add_figure：将matplotlib图形渲染到图像中，并将其添加到摘要中，需要 matplotlib 包
8. add_video：将视频数据添加到摘要中，需要 moviepy 包
9. add_audio：将音频数据添加到摘要中
10. add_text：将文本数据添加到摘要中
11. add_graph：将图表数据添加到摘要中，这个常用来显示模型结构
12. add_embedding：将词嵌入向量数据添加到摘要中，这个可以交互式显示一组词向量在三维空间的投影
13. add_pr_curve：添加精确召回曲线。绘制精确召回曲线可让您了解模型在不同阈值设置下的性能。此函数可以为每个目标提供真实标签(T/F)和预测置信度(通常是模型的输出)。利用 TensorBoard UI 可以交互式地选择阈值
14. add_custom_scalars：通过在 “scalar” 中收集的图表 tag 来创建特殊图表。注意对于每个 SummaryWriter 对象该函数只能调用一次，因为它只向tensorboard提供元数据，所以可以在训练循环之前或之后调用该函数
15. add_mesh：向TensorBoard添加网格或3D点云。可视化基于Three.js，因此它允许用户与呈现的对象进行交互。除了顶点、面等基本定义外，用户还可以提供相机参数、光照条件等

综合测试代码参考这里，运行效果可以参考这里：

# 引入SummaryWriter
import numpy as np
import torch
from torch.utils.tensorboard import SummaryWriter
import torchvision
from PIL import Image
import matplotlib
import matplotlib.pyplot as plt
matplotlib.use('Agg')##### 1、add_scalar实例
def add_scalar_demo():# 将信息写入logs文件夹，可以供TensorBoard消费，来可视化writer = SummaryWriter("logs/add_scalar")# 绘制 y = 2x 实例x = range(100)for i in x:writer.add_scalar('add_scalar实例：y=2x', i * 2, i)# 关闭writer.close()
add_scalar_demo()##### 2、add_scalars 实例
def add_scalars_demo():# 将信息写入logs文件夹，可以供TensorBoard消费，来可视化writer = SummaryWriter("logs/add_scalars")r = 5for i in range(100):writer.add_scalars('add_scalars实例', {'xsinx':i*np.sin(i/r),'xcosx':i*np.cos(i/r),'tanx': np.tan(i/r)}, i)# 关闭writer.close()add_scalars_demo()##### 3、add_text 实例
def add_text_demo():# 将信息写入logs文件夹，可以供TensorBoard消费，来可视化writer = SummaryWriter("logs/add_text")writer.add_text('lstm', 'This is an lstm', 0)writer.add_text('rnn', 'This is an rnn', 10)# 关闭writer.close()add_text_demo()##### 4、add_graph 实例def add_graph_demo():# 将信息写入logs文件夹，可以供TensorBoard消费，来可视化writer = SummaryWriter("logs/add_graph")img = torch.rand([1, 3, 64, 64], dtype=torch.float32)model = torchvision.models.AlexNet(num_classes=10)# print(model)writer.add_graph(model, input_to_model=img)# 关闭writer.close()add_graph_demo()##### 5、add_image 实例def add_image_demo():# 将信息写入logs文件夹，可以供TensorBoard消费，来可视化writer = SummaryWriter("logs/add_image")img1 = np.random.randn(1, 100, 100)writer.add_image('add_image 实例：/imag1', img1)img2 = np.random.randn(100, 100, 3)writer.add_image('add_image 实例：/imag2', img2, dataformats='HWC')img = Image.open('../imgs/1.png')img_array = np.array(img)writer.add_image('add_image 实例：/cartoon', img_array, dataformats='HWC')# 关闭writer.close()add_image_demo()##### 6、add_images 实例def add_images_demo():# 将信息写入logs文件夹，可以供TensorBoard消费，来可视化writer = SummaryWriter("logs/add_images")imgs1 = np.random.randn(8, 100, 100, 1)writer.add_images('add_images 实例/imgs1', imgs1, dataformats='NHWC')imgs2 = np.zeros((16, 3, 100, 100))for i in range(16):imgs2[i, 0] = np.arange(0, 10000).reshape(100, 100) / 10000 / 16 * iimgs2[i, 1] = (1 - np.arange(0, 10000).reshape(100, 100) / 10000) / 16 * iwriter.add_images('add_images 实例/imgs2', imgs2)  # Default is :math:`(N, 3, H, W)`img = Image.open('../imgs/1.jpg')img3 = np.array(img)imgs4= np.zeros((5, img3.shape[0], img3.shape[1], img3.shape[2]))for i in range(5):imgs4[i] = img3//(i+1)writer.add_images('add_images 实例/imgs4', imgs4, dataformats='NHWC')  # Default is :math:`(N, 3, H, W)`# 关闭writer.close()add_images_demo()##### 7、add_figure 实例def add_figure_demo():# 将信息写入logs文件夹，可以供TensorBoard消费，来可视化writer = SummaryWriter("logs/add_figure")# First create some toy data:x = np.linspace(0, 2 * np.pi, 400)y = np.sin(x ** 2)# Create just a figure and only one subplotfig, ax = plt.subplots()ax.plot(x, y)ax.set_title('Simple plot')writer.add_figure("add_figure 实例：figure", fig)# 关闭writer.close()add_figure_demo()##### 8、add_pr_curve 实例def add_pr_curve_demo():# 将信息写入logs文件夹，可以供TensorBoard消费，来可视化writer = SummaryWriter("logs/add_pr_curve")labels = np.random.randint(2, size=100)  # binary labelpredictions = np.random.rand(100)writer.add_pr_curve('add_pr_curve 实例：pr_curve', labels, predictions, 0)# 关闭writer.close()add_pr_curve_demo()##### 9、add_embedding 实例def add_embedding_demo():# 将信息写入logs文件夹，可以供TensorBoard消费，来可视化writer = SummaryWriter("logs/add_embedding")import tensorflow as tfimport tensorboard as tbtf.io.gfile = tb.compat.tensorflow_stub.io.gfileimport keywordimport torchmeta = []while len(meta) < 100:meta = meta + keyword.kwlist  # get some stringsmeta = meta[:100]for i, v in enumerate(meta):meta[i] = v + str(i)label_img = torch.rand(100, 3, 10, 32)for i in range(100):label_img[i] *= i / 100.0writer.add_embedding(torch.randn(100, 5), metadata=meta, label_img=label_img)# 关闭writer.close()add_embedding_demo()

3. 使用 TensorBoard 记录 PPO 运行情况

将 TensorBoard 相关代码添加到前文 RL 实践（7）—— CartPole【TPRO & PPO】的 PPO 代码中，观察 RL 的收敛过程

import gym
import torch
import random
import torch.nn.functional as F
import numpy as np
import matplotlib.pyplot as plt
from tqdm import tqdm
from gym.utils.env_checker import check_env
from gym.wrappers import TimeLimit 
from datetime import datetime
from torch.utils.tensorboard import SummaryWriterclass PolicyNet(torch.nn.Module):''' 策略网络是一个两层 MLP '''def __init__(self, input_dim, hidden_dim, output_dim):super(PolicyNet, self).__init__()self.fc1 = torch.nn.Linear(input_dim, hidden_dim)self.fc2 = torch.nn.Linear(hidden_dim, output_dim)def forward(self, x):x = F.relu(self.fc1(x))             # (1, hidden_dim)x = F.softmax(self.fc2(x), dim=1)   # (1, output_dim)return xclass VNet(torch.nn.Module):''' 价值网络是一个两层 MLP '''def __init__(self, input_dim, hidden_dim):super(VNet, self).__init__()self.fc1 = torch.nn.Linear(input_dim, hidden_dim)self.fc2 = torch.nn.Linear(hidden_dim, 1)def forward(self, x):x = F.relu(self.fc1(x))x = self.fc2(x)return xclass PPO(torch.nn.Module):def __init__(self, state_dim, hidden_dim, action_range, actor_lr, critic_lr, lmbda, epochs, eps, gamma, device):super().__init__()self.actor = PolicyNet(state_dim, hidden_dim, action_range).to(device)self.critic = VNet(state_dim, hidden_dim).to(device) self.actor_optimizer = torch.optim.Adam(self.actor.parameters(), lr=actor_lr)self.critic_optimizer = torch.optim.Adam(self.critic.parameters(), lr=critic_lr)self.device = deviceself.gamma = gammaself.lmbda = lmbda      # GAE 参数self.epochs = epochs    # 一条轨迹数据用来训练的轮数self.eps = eps          # PPO 中截断范围的参数self.device = device        def take_action(self, state):state = torch.tensor(state, dtype=torch.float).to(self.device)state = state.unsqueeze(0)probs = self.actor(state)action_dist = torch.distributions.Categorical(probs)action = action_dist.sample()return action.item()def compute_advantage(self, gamma, lmbda, td_delta):''' 广义优势估计 GAE '''td_delta = td_delta.detach().numpy()advantage_list = []advantage = 0.0for delta in td_delta[::-1]:advantage = gamma * lmbda * advantage + deltaadvantage_list.append(advantage)advantage_list.reverse()return torch.tensor(np.array(advantage_list), dtype=torch.float)def update(self, transition_dict):states = torch.tensor(np.array(transition_dict['states']), dtype=torch.float).to(self.device)actions = torch.tensor(transition_dict['actions']).view(-1, 1).to(self.device)rewards = torch.tensor(transition_dict['rewards'], dtype=torch.float).view(-1, 1).to(self.device)next_states = torch.tensor(np.array(transition_dict['next_states']), dtype=torch.float).to(self.device)dones = torch.tensor(transition_dict['dones'], dtype=torch.float).view(-1, 1).to(self.device)td_target = rewards + self.gamma * self.critic(next_states) * (1-dones)td_delta = td_target - self.critic(states)advantage = self.compute_advantage(self.gamma, self.lmbda, td_delta.cpu()).to(self.device)old_log_probs = torch.log(self.actor(states).gather(1, actions)).detach()# 用刚采集的一条轨迹数据训练 epochs 轮for _ in range(self.epochs):log_probs = torch.log(self.actor(states).gather(1, actions))ratio = torch.exp(log_probs - old_log_probs)surr1 = ratio * advantagesurr2 = torch.clamp(ratio, 1 - self.eps, 1 + self.eps) * advantage  # 截断actor_loss = torch.mean(-torch.min(surr1, surr2))                   # PPO损失函数critic_loss = torch.mean(F.mse_loss(self.critic(states), td_target.detach()))# 更新网络参数self.actor_optimizer.zero_grad()self.critic_optimizer.zero_grad()actor_loss.backward()critic_loss.backward()self.actor_optimizer.step()self.critic_optimizer.step()if __name__ == "__main__":def moving_average(a, window_size):''' 生成序列 a 的滑动平均序列 '''cumulative_sum = np.cumsum(np.insert(a, 0, 0)) middle = (cumulative_sum[window_size:] - cumulative_sum[:-window_size]) / window_sizer = np.arange(1, window_size-1, 2)begin = np.cumsum(a[:window_size-1])[::2] / rend = (np.cumsum(a[:-window_size:-1])[::2] / r)[::-1]return np.concatenate((begin, middle, end))def set_seed(env, seed=42):''' 设置随机种子 '''env.action_space.seed(seed)env.reset(seed=seed)random.seed(seed)np.random.seed(seed)torch.manual_seed(seed)state_dim = 4               # 环境观测维度action_range = 2            # 环境动作空间大小actor_lr = 1e-3critic_lr = 1e-2num_episodes = 200hidden_dim = 64gamma = 0.98lmbda = 0.95epochs = 10eps = 0.2device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")# build environmentenv_name = 'CartPole-v0'env = gym.make(env_name, render_mode='rgb_array')check_env(env.unwrapped)    # 检查环境是否符合 gym 规范set_seed(env, 0)# build agentagent = PPO(state_dim, hidden_dim, action_range, actor_lr, critic_lr, lmbda, epochs, eps, gamma, device)# TensorBoard writerTIMESTAMP = "{0:%Y-%m-%dT%H-%M-%S/}".format(datetime.now())writer = SummaryWriter(f"logs/PPO")#writer = SummaryWriter(f"logs/PPO/{TIMESTAMP}")# start trainingreturn_list = []for i in range(10):with tqdm(total=int(num_episodes / 10), desc='Iteration %d' % i) as pbar:for i_episode in range(int(num_episodes / 10)):episode_return = 0transition_dict = {'states': [],'actions': [],'next_states': [],'next_actions': [],'rewards': [],'dones': []}state, _ = env.reset()# 以当前策略交互得到一条轨迹while True:action = agent.take_action(state)next_state, reward, terminated, truncated, _ = env.step(action)next_action = agent.take_action(next_state)transition_dict['states'].append(state)transition_dict['actions'].append(action)transition_dict['next_states'].append(next_state)transition_dict['next_actions'].append(next_action)transition_dict['rewards'].append(reward)transition_dict['dones'].append(terminated or truncated)state = next_stateepisode_return += rewardif terminated or truncated:break#env.render()# 用当前策略收集的数据进行 on-policy 更新agent.update(transition_dict)# 更新进度条return_list.append(episode_return)pbar.set_postfix({'episode':'%d' % (num_episodes / 10 * i + i_episode + 1),'return':'%.3f' % episode_return,'ave return':'%.3f' % np.mean(return_list[-10:])})pbar.update(1)writer.add_scalars(main_tag='return', tag_scalar_dict={f'hidden{hidden_dim}':episode_return}, global_step=i*int(num_episodes / 10) + i_episode)writer.add_hparams(hparam_dict={'actor_lr': actor_lr, 'critic_lr': critic_lr,'hidden_dim': hidden_dim,'gamma': gamma,'lmbda': lmbda,'eps': eps,'num_episodes': num_episodes},metric_dict={'hparam/ave return': np.mean(return_list), })writer.close()# show policy performencemv_return_list = moving_average(return_list, 29)episodes_list = list(range(len(return_list)))plt.figure(figsize=(12,8))plt.plot(episodes_list, return_list, label='raw', alpha=0.5)plt.plot(episodes_list, mv_return_list, label='moving ave')plt.xlabel('Episodes')plt.ylabel('Returns')plt.title(f'{agent._get_name()} on CartPole-V0')plt.legend()plt.savefig(f'./result/{agent._get_name()}.png')plt.show()