Using Training Lab

You can create a Training Lab for the Simple AI Training service, select an AI training method, and proceed with training.

Create Training Lab

To create a Training Lab, follow these steps.

1. Click the All Services > AI/ML > Simple AI Training menu. 1. Navigate to the Service Home page of Simple AI Training.

2. On the Service Home page, click the Training Lab menu. 2. Go to the Training Lab List page.

3. On the Training Lab List page, click the Create Training Lab button. 3. Go to the Create Training Lab page.

4. Enter the information required to create the service and select detailed options.

   • In the Basic Information Input area, please enter or select the information required to create a Training Lab.

     Category	Required status	Detailed description
     Training Lab name	Required	Enter the Training Lab name Start with a lowercase English letter and use numbers and the special character (-) to enter within 3 ~ 60 characters The last character of the name must be a lowercase letter or a number

     Table. Training Lab Basic Information Input Items
   • Service Information Input area, enter or select the information needed to create a Training Lab.

     Category	Required	Detailed description
     Endpoint URL	Select	Set whether to use the Endpoint URL For detailed information about using the Endpoint URL, see Training Lab Endpoint 사용하기
     Resource allocation	Required	Select the number of GPUs and memory size to use for training Up to 4 GPUs can be allocated
     Shared Memory	Select	Set the memory to be shared between processes for distributed training and distributed data processing Enter a value between 1 ~ 8

     Table. Training Lab Service Information Input Items
   • In the AI Training Image Information Input area, select the options required to create the service.

     Category	Required	Detailed description
     AI Lab Image URL	Required	Enter the URL address of the container repository (SCR, Docker Hub, etc.) for the image (Python, PyTorch, CUDA, Nvidia GPU, etc.) that defines the environment where the training script will run.
     User ID	Select	Enter the user ID of the image repository
     Password	Select	Enter the password for the image repository

     Table. Training Lab AI Training Image Information Input Items
   • Enter Training Command and Volume Information area, enter or select the required information.

     Category	Required	Detailed description
     Volume	Select	When using an additional Volume, select the Volume’s Mount path and the Volume to connect Mount Path: Enter the data path to use when connecting File Storage (e.g., /root) Volume: Select the Volume to connect Add button can be clicked to add up to 10 Volumes
     Command	Required	Enter the command information within 4,096 characters vllm, sglang, ollama. You must enter an execution command for one of them Environment variables can be specified before the execution command Refer to the command input guidelines and examples below

     Table. Training Lab Training Command and Volume Information Input Items

     Command input criteria and examples

     • Environment variables are specified in the NAME=value format before the command, and multiple ones are separated by spaces.
     • Environment variable names and values cannot be left empty, and the name cannot start with -.
     • There is no whitelist for environment variable names, and duplicate names are also allowed.
     • After setting the environment variables, one of the execution commands vllm, sglang, ollama must follow.
     • Arguments following Runtime are not subject to any additional restrictions beyond port validation and prohibited special character limits.
     • Line breaks and the following special characters cannot be used.

     ; & | ` $ ( ) < > \ " ' { } [ ] * ? ~ ! #
     

     • OLLAMA_ORIGINS=* cannot be entered because it contains the prohibited character *.

     • An example of running vllm with specified environment variables is as follows.

       • Example: VLLM_LOGGING_LEVEL=INFO vllm serve Qwen/Qwen2.5-1.5B-Instruct –port 8000

     • The default port is for internal Runtime communication and is not an external Endpoint port.

       • Default ports: vllm 8000 / sglang 30000 / ollama 11434

     • vllm·sglang is changed to the –port <포트번호> option.

     • ollama is configured via the OLLAMA_HOST environment variable placed before the execution command; an example is shown below.

       • Example: OLLAMA_HOST=0.0.0.0:11434

     • When specifying a port, enter an integer in the range 1 ~ 65535.

   • In the Additional Information Input area, enter or select the required information.

     Category	Required	Detailed description
     tag	Select	Add Tag Up to 50 can be added per resource After clicking the Add Tag button, enter or select Key, Value values

     Table. Additional information entry fields

5. Summary Review the detailed information generated in the panel, then click the Create button.

6. When the popup indicating creation opens, click the Confirm button.

   • When creation is complete, check the created resources on the Training Lab List page.

Training Lab View Detailed Information

You can view the complete list of resources and detailed information for the Training Lab service. To view the detailed information of the Training Lab, follow these steps.

1. Click the All Services > AI/ML > Simple AI Training menu. 1. Go to the Service Home page of Simple AI Training.
2. On the Service Home page, click the Training Lab menu. 2. Go to the Training Lab List page.
3. On the Training Lab List page, click the resource to view detailed information. 3. Navigate to the Training Lab Details page.
   • Training Lab Details page consists of Details, Monitoring, Logs, Tags tabs.

     Category	Detailed description
     Service status	Training Lab status Pending: Creating / Waiting due to insufficient resources / Creation not possible due to incorrect input Running: Creation completed Failed: Creation failed Terminating: Deleting
     Delete Training Lab	Service cancellation button

     Table. Training Lab detail page items

Detailed Information

On the Training Lab List page, you can view detailed information of the selected resource.

Monitoring

You can view the monitoring information of the selected resource on the Training Lab List page.

log

On the Training Lab list page, you can view the log information of the selected resource.

Tag

On the Training Lab List page, you can view the tag information of the selected resource and add, modify, or delete it.

Checking Training Lab Logs

You can view the Training Lab logs in the ServiceWatch service. To view the logs of the Training Lab, follow these steps.

1. All Services > AI/ML > Simple AI Training Click the menu. 1. Navigate to the Service Home page of Simple AI Training.
2. On the Service Home page, click the Training Lab menu. 2. Go to the Training Lab List page.
3. Training Lab List page, select the resource to view logs. 3. Go to the Training Lab Details page.
4. On the Training Lab Details page, click the Log tab. 4. The log information for this Lab is displayed.
5. In the log information, click the name at the bottom of the Training Lab log. 5. Navigate to the ServiceWatch Log Group Details page.
6. On the Log Group Details page, click the Log Stream tab. 6. The log stream list is displayed.
7. Click the log stream name to check (e.g., training-lab). 7. The logs of the stream are displayed in chronological order.

Using the Training Lab Endpoint

When creating a Training Lab, if you configure it to use an Endpoint URL, you can call the inference runtime (vLLM, SGLang, Ollama) running in the Lab from outside using the issued Endpoint URL and token.

Prerequisite

• When creating a Training Lab, you must set the Endpoint URL in the Service Information Input area to Use.
  Caution

  After creating a Training Lab, you cannot change its usage status.
• The status of the Training Lab must be Running.
• The runtime entered in Command must be started correctly.

Endpoint URL structure

The endpoint URL is issued in the following format.

https://<Endpoint 도메인>/<Account ID>/<Training Lab명>

Calling an Endpoint

To call the endpoint, follow these steps.

1. All Services > AI/ML > Simple AI Training Click the menu. 1. Go to the Service Home page of Simple AI Training.
2. On the Service Home page, click the Training Lab menu. 2. Training Lab List Navigate to the page.
3. On the Training Lab List page, select the resource to call the Endpoint. 3. Go to the Training Lab Details page.
4. Check the Endpoint URL on the Detailed Information tab of the Training Lab Details page.
5. Prepare the Token issued when creating a Training Lab.
6. Add x-sat-token: to the request header, and append the runtime’s API path after the Endpoint URL when calling.
   • vLLM, SGLang: Provides an OpenAI-compatible API(/v1/…)

# 모델 목록 조회
curl https://<Endpoint 도메인>/<Account ID>/<Training Lab명>/v1/models \
  -H "x-sat-token: <Token>"

# 채팅 완성 요청
curl https://<Endpoint 도메인>/<Account ID>/<Training Lab명>/v1/chat/completions \
  -H "x-sat-token: <Token>" \
  -H "Content-Type: application/json" \
  -d '{
        "model": "<모델명>",
        "messages": [{"role": "user", "content": "안녕하세요"}]
      }'

* **Ollama**: Provides the Ollama API (<code>/api/...</code>) and the OpenAI-compatible API (</code>/v1/...<code>).

# 모델 목록 조회
curl https://<Endpoint 도메인>/<Account ID>/<Training Lab명>/api/tags \
  -H "x-sat-token: <Token>"

# 생성 요청
curl https://<Endpoint 도메인>/<Account ID>/<Training Lab명>/api/generate \
  -H "x-sat-token: <Token>" \
  -H "Content-Type: application/json" \
  -d '{"model": "<모델명>", "prompt": "안녕하세요"}'

OLLAMA_HOST=0.0.0.0:11434 ollama serve

Regenerate Token

If you have lost the token or are concerned about its exposure, you can reissue the token.

1. Click the All Services > AI/ML > Simple AI Training menu. 1. Navigate to the Service Home page of Simple AI Training.
2. Click the Training Lab menu on the Service Home page. 2. Go to the Training Lab List page.
3. On the Training Lab list page, click the resource to regenerate the Token. 3. Go to the Training Lab Details page.
4. On the Training Lab Details page, click the Create button in the Token Regeneration item of the Detailed Information tab.
5. When the popup indicating regeneration opens, click the Confirm button. 5. The newly issued Token is displayed.

Delete Training Lab

You can delete unused Training Labs. To delete the Training Lab, follow these steps.

1. Click the All Services > AI/ML > Simple AI Training menu. 1. Go to the Service Home page of Simple AI Training.
2. On the Service Home page, click the Training Lab menu. 2. Go to the Training Lab List page.
3. On the Training Lab list page, select the resource to delete, then click the Delete button at the top of the list.
   • You can click the Delete button of the Training Lab you want to remove from the Training Lab list, or click the Delete Training Lab button on the Training Lab’s detail page to delete it individually.
4. When the deletion notification popup appears, click the Confirm button.