概述
YOLOv5(v4.0 release开始)已经在本地集成了Weights & Biases,也就是可视化的工具wandb,可方便的追踪模型训练的整个过程,包括模型的性能、超参数、GPU的使用情况、模型预测,还有数据集。
软硬件环境
OS:CentOS 7.7.1908
[king@cam yolov5-docker-image]$ cat /etc/redhat-release
CentOS Linux release 7.7.1908 (Core)
[king@cam yolov5-docker-image]$ uname -a
Linux cam 3.10.0-1062.18.1.el7.x86_64 #1 SMP Tue Mar 17 23:49:17 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux
[king@cam yolov5-docker-image]$ |
Python 3.8.5 && Python 2.7.5
[king@cam yolov5-docker-image]$ python3 --version
Python 3.8.5
[king@cam yolov5-docker-image]$ python --version
Python 2.7.5
[king@cam yolov5-docker-image]$ |
CUDA: 10.0, V10.0.130
[king@cam yolov5-docker-image]$ nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2018 NVIDIA Corporation
Built on Sat_Aug_25_21:08:01_CDT_2018
Cuda compilation tools, release 10.0, V10.0.130 |
YOLOv5 v4.0
NVIDIA GeForce GTX 2080Ti
[king@cam yolov5-docker-image]$ nvidia-smi
Sun Feb 21 10:33:18 2021
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 440.64 Driver Version: 440.64 CUDA Version: 10.2 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 GeForce RTX 208... Off | 00000000:1A:00.0 Off | N/A |
| 40% 54C P2 180W / 260W | 8100MiB / 11019MiB | 19% Default |
+-------------------------------+----------------------+----------------------+
| 1 GeForce RTX 208... Off | 00000000:1E:00.0 Off | N/A |
| 41% 33C P8 36W / 260W | 0MiB / 11019MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 0 82207 C python3 8087MiB |
+-----------------------------------------------------------------------------+
|
安装wandb
配置wandb
注册账号 ======> 终端输入API Key(在浏览器中访问站点 https://wandb.ai/authorize,复制后贴到终端中)
wandb有在线和本地两种使用方式。
在线使用方式需要在https://wandb.ai/home,注册一个账号。注册后新建项目,名字叫yolov5,然后本地安装配置wandb, 按提示输入必要的信息(API Key)。
/usr/local/python385/bin/wandb login ***6ef350c8****** |
不过wandb网站挺卡,wandb也有本地使用方式。参考:https://docs.wandb.ai/self-hosted/local, 配置好后也可以本地访问了。
训练模型
开始训练了,这个过程跟之前训练是一模一样的
python3 train.py --device 0 |
查看训练过程
在模型训练的过程中,登录网站 https://wandb.ai/home,在自己的项目中就可以看到训练的状态了。
所有评论(0)