Create from template

Learn how to create applications from templates in Release

Release simplifies the process of creating and deploying applications by using preset templates. These templates are pre-configured environments tailored to specific use cases, allowing you to get your applications up and running quickly without needing to handle the intricate details of setup and configuration.

Release templates are designed to handle the configurations required for various environments and optimized for specific tasks like running Jupyter Notebooks, deploying language models, or fine-tuning machine learning models. This guide provides step-by-step instructions for creating an application from a template in Release.

Step 1: Create an application

Log in to your Release account and navigate to Applications. Click Create Application and select Create from Template from the dropdown to start the setup process.

Create App

Step 2: Choose your execution type

Release offers two types of templates: Server and Runnable. The type you choose depends on the nature of your application.

Select Server or Runnable

• Server: Server templates are meant for applications that need a persistent server environment. For instance, running a Jupyter Notebook server with GPU acceleration requires a stable, long-running server. Learn more about server applications.

• Runnable: Runnable templates are for applications that can be executed on demand. This is ideal for tasks like fine-tuning models, where the application does not need to run continuously. Learn more about runnable applications.

Step 3: Select a template

Choose the template that best suits your application requirements. Release offers a range of templates optimized for specific tasks. To learn more about the available templates, check out the template library.

Select a server template

The stack icons in the top-right corner of each template card represent the core technologies and integrations included in the template.

Server Templates:

• Jupyter: For GPU-accelerated Jupyter Notebooks.

• Ollama: For large language models.

• Ollama-WebUI: A feature-rich, self-hosted WebUI.

• GenAI: For building GenAI applications.

• Anything-LLM: Other language model-based applications.

Alternatively, you can choose a runnable template for your application by selecting the Runnable tickbox.

Select a Runnable Template

Runnable Templates:

• NeMo: For fine-tuning models with NeMo.

• Pytorch: A container with PyTorch and NVIDIA GPU drivers.

• Axolotl: Another option for fine-tuning models.

To learn more about runnables, check out the runnable documentation.

Step 4: Configure your application

Release will automatically generate a name for your application based on the template you choose. You can change this name to something more descriptive if you wish.

Configure your Application Template

Edit the template configuration settings to suit your specific requirements. These settings include the application’s name, a description, and the resources allocated to it. You can also choose the environment in which the application will run, such as a CPU or GPU environment.

To learn more about configuring your application, check out the configuration documentation for templates.

Step 5: Finish and create environment

Once you have configured your application, click Finish to save your app template. You can now create an environment for your application by going to the application’s page and clicking Create Environment.

Learn more about creating environments in Release.