Tencent Cloud App Deployment

Currently, the SwanLab self-hosted version is available on the Tencent Cloud App Market. Trainers are welcome to use it out-of-the-box through Tencent Cloud~

• SwanLab Tencent Cloud App

Prerequisites

1. First, you need a Tencent Cloud account and ensure that the account has permission to install cloud applications, reference: Tencent Cloud App Purchase & Installation Instruction

2. In the Tencent Cloud Console - VPC, create a default VPC to provide the target network for the cloud application. Currently supported regions are as follows:

   • China Mainland: Nanjing; Beijing; Guangzhou; Chengdu; Shanghai; Chongqing
   • Other Regions: Hong Kong SAR; Singapore; Silicon Valley; Sao Paulo; Frankfurt

Taking the Nanjing region as an example, the CIDR and subnet can be modified as needed. The only required fields are Name, Subnet Name, and Availability Zone.

Installation Tutorial

1. Go to the SwanLab Tencent Cloud App page, check I have read and agree to the "Tencent Cloud App Universal Product User Agreement", and click Install App to jump to the console interface.

2. In the console interface, you only need to configure three cloud resource settings: Target Network, Cloud Server Type, and Data Disk Size:

The meaning of each cloud resource is as follows:

Configuration	Description	Requirements
Target Network	The region where the cloud service is hosted	You can choose based on the region where the VPC was previously created
Cloud Server Type	The cloud server instance configuration for hosting the SwanLab self-hosted version	Minimum configuration: - vCPU: ≥ 4 cores - Memory: ≥ 8GB - System Storage Space: Default 40GB
Data Disk Size	The hard disk size used by the SwanLab self-hosted version to record experimental data	Default is 100GB, minimum is 40GB

After configuring the cloud resources, click Next: Confirm Resources.

3. Next, you will enter the Confirm Order Information interface. Tencent Cloud will generate a bill based on the cloud resources selected in the previous step. At this time, you need to ensure that there is a certain balance in your Tencent Cloud account. After confirming that the order is correct, click Allow service role to call other cloud service interfaces, and click Next: Install App.

4. Next, you will enter the application installation interface. You need to wait for all resources to be created and started, which will take about 5 minutes.

5. After completion, you can see the created cloud application in the Tencent Cloud console interface. Click Open App to use the self-hosted version of SwanLab.

Tip

After the application is created, if you open the application immediately, you may see a 404 page. This is because the cloud server instance needs to perform some container initialization operations after it is created. Wait 1~2 minutes before opening it.

Activate Main Account

Now, you can use the self-hosted version of SwanLab on Tencent Cloud.

Individual users can apply for a free License on the SwanLab official website, located in "Settings" - "Account and License".

Enter your account, password, and License to activate the self-hosted version of SwanLab.

Start Experiment

Log in with the Python SDK:

swanlab login --host <IP Address>

If you have logged in to swanlab before and want to log in again, please use: swanlab login --host <IP Address> --relogin.

Press Enter, fill in the API Key to complete the login. After that, your SwanLab experiments will be uploaded to the privately deployed SwanLab by default.

---

Test script:

import swanlab
import random

# Create a SwanLab project
swanlab.init(
    # Set project name
    project="my-awesome-project",
    
    # Set hyperparameters
    config={
        "learning_rate": 0.02,
        "architecture": "CNN",
        "dataset": "CIFAR-100",
        "epochs": 10
    }
)

# Simulate a training session
epochs = 10
offset = random.random() / 5
for epoch in range(2, epochs):
  acc = 1 - 2 ** -epoch - random.random() / epoch - offset
  loss = 2 ** -epoch + random.random() / epoch + offset

  # Log training metrics
  swanlab.log({"acc": acc, "loss": loss})

# [Optional] Finish training, this is necessary in a notebook environment
swanlab.finish()

After running, you can view the experiment on the webpage.

Tip

If you no longer need to use it, please destroy the application in the Tencent Cloud App Console in time to avoid continued billing.