Creating a Virtual Machine

For first-time AutoDL users, after entering the AutoDL website, you first need to register an account. If you are a student, remember to complete student verification, which can save some money. These operations are relatively simple, so I won’t go into detail here.

Creating an Instance

After registration, click the Console button in the upper right corner to enter the console interface.

Then, enter the instance container.

Here, all rented instances are displayed. Click Rent New Instance to enter the instance creation interface.

You can see that you can choose the appropriate billing method, region, GPU type, number of GPUs, data disk size, image to use, etc.

Here, we choose Pay-as-you-go, Chongqing Zone A, 4090D×1, 500GB data disk, and the image as shown.

Then, click Create Instance, wait for a while, and the instance will be created successfully.

Connecting to the Virtual Machine

Although AutoDL provides several different connection methods, I recommend using the SSH connection method, relying on key-based authentication.

Setting up config

First, copy the login command, and you will get a command in the form of ssh -p 12345 root@xxx.mmm.nnn.com. Then, enter the .ssh folder under your user directory.

Enter the config file and write in the following format:

Host autodl
    HostName xxx.mmm.nnn.com
    Port 12345
    User root
    IdentityFile ~/.ssh/id_rsa_autodl

Then you can connect using ssh autodl.

Environment Configuration

After connecting, you can configure the conda environment as needed.

For example, create a new environment:

conda create -n gene python=3.10

Activate the environment:

conda activate gene

You can also modify the environment prompt:

conda config --set env_prompt '({default_env})'

Then use conda info --envs to see that the gene environment has been activated.