【PyTorch】11：TensorBoard 可视化

---

1. TensorBoard 安装

TensorBoard：TensorFlow 中强大的可视化工具，它可以用来展示网络图、张量的指标变化、张量的分布情况等。在训练网络时，可以设置不同的参数（比如：权重W、偏置B、卷积层数、全连接层数等），使用TensorBoader可以很直观的辅助选择参数。

---

激活虚拟环境，直接命令行安装即可：

pip install tensorboard

测试：

import numpy as np
from torch.utils.tensorboard import SummaryWriterwriter = SummaryWriter(comment='test_tensorboard')for x in range(100):writer.add_scalar('y=2x', x * 2, x)writer.add_scalar('y=pow(2, x)', 2 ** x, x)writer.add_scalars('data/scalar_group', {"xsinx": x * np.sin(x),"xcosx": x * np.cos(x),"arctanx": np.arctan(x)}, x)
writer.close()

启动：

tensorboard --logdir=./runs

查看：

注意：事件文件与代码是同级目录。

---

2. TensorBoard 使用

2.1 SummaryWriter()

SummaryWriter 官方文档：LINK。

torch.utils.tensorboard.writer.SummaryWriter(log_dir=None, comment='', purge_step=None, max_queue=10, flush_secs=120, filename_suffix='')

功能：提供创建 event file 的高级接口。

主要属性：

• log_dir：event file 输出文件夹；
• comment：不指定 log_dir 时，文件夹后缀名；
• filename_suffix：event file 文件名后缀名。

实例：

log_dir = "./train_log/test_log_dir"
writer = SummaryWriter(log_dir=log_dir, comment='_scalars', filename_suffix="123456")
writer = SummaryWriter(comment='_scalars', filename_suffix="123456")

---

2.2 add_scalar()

add_scalar(tag, scalar_value, global_step=None, walltime=None)

功能：记录标量。

• tag：图像的标签名，图的唯一标识；
• scalar_value：要记录的标量；
• global_step：x 轴。

---

add_scalars(main_tag, tag_scalar_dict, global_step=None, walltime=None)

• main_tag ：该图的标签；
• tag_scalar_dict： key 是变量的 tag，value 是变量的值。

from torch.utils.tensorboard import SummaryWritermax_epoch = 100
writer = SummaryWriter(comment='test_comment', filename_suffix="test_suffix")for x in range(max_epoch):writer.add_scalar('y=2x', x * 2, x)writer.add_scalar('y=pow_2_x', 2 ** x, x)writer.add_scalars('data/scalar_group', {"xsinx": x * np.sin(x),"xcosx": x * np.cos(x)}, x)writer.close()

---

2.3 add_histogram()

add_histogram(tag, values, global_step=None, bins='tensorflow', walltime=None, max_bins=None)

writer = SummaryWriter(comment='test_comment', filename_suffix="test_suffix")for x in range(2):np.random.seed(x)data_union = np.arange(100)data_normal = np.random.normal(size=1000)writer.add_histogram('distribution union', data_union, x)writer.add_histogram('distribution normal', data_normal, x)plt.subplot(121).hist(data_union, label="union")plt.subplot(122).hist(data_normal, label="normal")plt.legend()plt.show()writer.close()

---

2.4 add_image()

add_image(tag, img_tensor, global_step=None, walltime=None, dataformats='CHW')

功能：记录图像。

• tag：图像的标签名，图的唯一标识；
• img_tensor：图像数据，注意尺度；
• global_step： x 轴；
• dataformats：数据形式 CHW、HWC、HW。

add_images(tag, img_tensor, global_step=None, walltime=None, dataformats='NCHW')

---

2.5 add_graph()

add_graph(model, input_to_model=None, verbose=False)

功能：可视化模型计算图。

参数说明：

• model ：模型，必须是 nn.Module
• input_to_model ：输出给模型的数据
• verbose ：是否打印计算图结构信息

---

2.6 add_video()

add_video(tag, vid_tensor, global_step=None, fps=4, walltime=None)

---

2.7 make_grid()

torchvision.utils.make_grid(tensor: Union[torch.Tensor, List[torch.Tensor]], nrow: int = 8, padding: int = 2, normalize: bool = False,range: Optional[Tuple[int, int]] = None, scale_each: bool = False, pad_value: int = 0)

功能：制作网格图像。

参数说明：

• tensor：图像数据, BCH*W 形式；
• nrow：行数（列数自动计算）；
  -padding：图像间距（像素单位）；
• normalize：是否将像素值标准化；
• range：标准化范围；
  -scale_each：是否单张图维度标准化；
• pad_value：padding的像素值。

---

实例：

import os
import torch
import time
import torchvision.models as models
import torchvision.transforms as transforms
import torchvision.utils as vutils
from torch.utils.tensorboard import SummaryWriter
from torch.utils.data import DataLoaderimport sys
hello_pytorch_DIR = os.path.abspath(os.path.dirname(__file__)+os.path.sep+".."+os.path.sep+"..")
sys.path.append(hello_pytorch_DIR)
from tools.my_dataset import RMBDataset
from tools.common_tools import set_seed
from model.lenet import LeNetset_seed(1)  # 设置随机种子writer = SummaryWriter(comment='test_your_comment', filename_suffix="_test_your_filename_suffix")split_dir = os.path.join("D:\Project\Pytorch Project\data\CatAndDog", "CatAndDogSplit")
train_dir = os.path.join(split_dir, "train")
# train_dir = "path to your training data"transform_compose = transforms.Compose([transforms.Resize((32, 64)), transforms.ToTensor()])
train_data = RMBDataset(data_dir=train_dir, transform=transform_compose)
train_loader = DataLoader(dataset=train_data, batch_size=16, shuffle=True)
data_batch, label_batch = next(iter(train_loader))img_grid = vutils.make_grid(data_batch, nrow=4, normalize=True, scale_each=True)
# img_grid = vutils.make_grid(data_batch, nrow=4, normalize=False, scale_each=False)
writer.add_image("input img", img_grid, 0)writer.close()

---

2.8 torchsummary

torchsummary，仓库地址：LINK。

功能：查看模型信息，便于调试。

参数说明：

• model： pytorch 模型；
• input_size ：模型输入 size；
• batch_size： batch size；
• device： cuda or cpu。

---

3. 权重和特征图可视化

writer = SummaryWriter(comment='test_your_comment', filename_suffix="_test_your_filename_suffix")alexnet = models.alexnet(pretrained=True)kernel_num = -1
vis_max = 1for sub_module in alexnet.modules():if isinstance(sub_module, nn.Conv2d):kernel_num += 1if kernel_num > vis_max:breakkernels = sub_module.weightc_out, c_int, k_w, k_h = tuple(kernels.shape)for o_idx in range(c_out):kernel_idx = kernels[o_idx, :, :, :].unsqueeze(1)   # make_grid 需要 BCHW，这里拓展C维度kernel_grid = vutils.make_grid(kernel_idx, normalize=True, scale_each=True, nrow=c_int)writer.add_image('{}_Convlayer_split_in_channel'.format(kernel_num), kernel_grid, global_step=o_idx)kernel_all = kernels.view(-1, 3, k_h, k_w)  # 3, h, wkernel_grid = vutils.make_grid(kernel_all, normalize=True, scale_each=True, nrow=8)  # c, h, wwriter.add_image('{}_all'.format(kernel_num), kernel_grid, global_step=322)print("{}_convlayer shape:{}".format(kernel_num, tuple(kernels.shape)))writer.close()

writer = SummaryWriter(comment='test_your_comment', filename_suffix="_test_your_filename_suffix")# 数据
path_img = "./lena.png"     # your path to image
normMean = [0.49139968, 0.48215827, 0.44653124]
normStd = [0.24703233, 0.24348505, 0.26158768]norm_transform = transforms.Normalize(normMean, normStd)
img_transforms = transforms.Compose([transforms.Resize((224, 224)),transforms.ToTensor(),norm_transform
])img_pil = Image.open(path_img).convert('RGB')
if img_transforms is not None:img_tensor = img_transforms(img_pil)
img_tensor.unsqueeze_(0)    # chw --> bchw# 模型
alexnet = models.alexnet(pretrained=True)# forward
convlayer1 = alexnet.features[0]
fmap_1 = convlayer1(img_tensor)# 预处理
fmap_1.transpose_(0, 1)  # bchw=(1, 64, 55, 55) --> (64, 1, 55, 55)
fmap_1_grid = vutils.make_grid(fmap_1, normalize=True, scale_each=True, nrow=8)writer.add_image('feature map in conv1', fmap_1_grid, global_step=322)
writer.close()