Create a GPU Container Job

In this tutorial, you will create your first GPU Container job using the CosmicAC CLI. By the end, you will have a running container ready to accept a shell session.

Prerequisites:

A CosmicAC account
CosmicAC CLI installed. See Installation.

Log in

Authenticate with your CosmicAC account:

cosmicac login

This opens a browser window. Complete the login steps to continue. If the browser does not open, copy the URL from the terminal and paste it into your browser.

Skip this step if you are already authenticated.

Initialize the job configuration

Run the interactive setup to generate a job.config.json file:

cosmicac jobs init

Follow the prompts:

Prompt	Description
Project name	A name for your job
Tags	Comma-separated labels to organize the job
Type	Select `GPU_CONTAINER`
GPU type	The GPU hardware to use (e.g., `H100`)
GPU count	The number of GPUs to allocate
Country code	The 2-letter region code where your container runs (e.g., `us`, `in`, `gb`)
Container image	The base image in `repository:tag` format (e.g., `ubuntu:24.04`)
Cost limit	The maximum spend in USD
Alerts	Select `Cost Exceeded`, `Errors`, or both

The CLI saves your answers to job.config.json in the current directory.

Review the configuration

Open job.config.json and confirm your settings before submitting:

{
  "name": "train-image-model",
  "tags": ["training", "image-recognition"],
  "type": "GPU_CONTAINER",
  "gpu": {
    "type": "H100",
    "count": "2"
  },
  "location": "in",
  "container": {
    "image": {
      "repository": "ubuntu",
      "tag": "24.04"
    }
  },
  "params": {},
  "cost_limit": 100,
  "alerts": {
    "cost_exceeded": true,
    "errors": true
  }
}

If anything is incorrect, edit the file directly or re-run cosmicac jobs init to regenerate it.

Create the container

Submit the job:

cosmicac jobs create

CosmicAC provisions and starts your GPU Container job. This takes a few minutes.

Verify the container is running

Check the status of your job:

cosmicac jobs list

Your container is ready when the status shows Running.

You have created your first GPU Container job. It is now running and ready to accept a shell session.