Skip to main content

How do I use my model(s)?

Download models in HF or GGUF format, access them via Minibase Cloud APIs, or test and chat with them in your browser.

N
Written by Niko McCarty
Updated this week

Overview

After you’ve trained a model in Minibase, there are three main ways to start using it:

  1. Download it and run it locally in your own codebase.

  2. Use Minibase Cloud, sending requests through our API (with examples for Python and JavaScript).

  3. Test and chat with it directly in your browser on the Model Details page.

These options give you flexibility depending on whether you want fast prototyping, browser-based testing, or full local deployment.

How to Use

Option 1: Download your model

  • Go to the Models tab → select your model → Download.

  • Task-based models are provided in Hugging Face (HF) format.

  • Chat, Language, and Micro models are provided in GGUF format.

  • You can integrate these into your own codebase or deploy them on local devices and servers.

Option 2: Use Minibase Cloud via API

  • Access your model without managing infrastructure by using our Cloud API.

  • On the Model Details page, you’ll find code snippets for making API calls.

  • We provide examples in both Python and JavaScript so you can get started quickly.

  • Simply insert your API key, send requests to your model, and get back responses.

Option 3: Chat with your model in the browser

  • From the Model Details page, you can test your model directly by chatting with it.

  • This is the fastest way to experiment and see how your model behaves.

  • You can adjust Temperature (to make outputs more or less random) and Max Tokens (to control output length) while testing.

Use this feature to explore your model’s capabilities, then decide if more fine-tuning is needed.

Tips & Best Practices

  • Prototype quickly in the browser: Use chat mode to see if your fine-tuned model behaves as expected.

  • Use Cloud APIs for easy integration: Great for apps and prototypes that require internet access but no infrastructure overhead.

  • Download for full control: Ideal for production, edge devices, or offline use cases.

  • Experiment with settings: Temperature and Max Tokens can dramatically change how your model responds—use them to fine-tune behavior before retraining.

Troubleshooting

I don't know which format to download

Task-based models use HF format, while Chat, Language, and Micro models use GGUF format. Choose based on the tools and frameworks you plan to use.

The model’s responses aren’t what I expected

Experiment with different Temperature and Max Tokens settings. If results are still off, consider refining your dataset and running another fine-tune.

The chat in the browser cuts off my model’s response

Increase the Max Tokens setting so the model has more space to generate longer outputs.

My API request isn’t working

Check that you’ve added your API key correctly and are using the sample code as a reference. If problems persist, reach out to [email protected].

Did this answer your question?