Model HQ

Testing Models in Model HQ

This guide explains how to test models within Model HQ. You will learn how to launch the testing interface, select a model, configure test options, and use additional features such as mapping and sample generation.

 

1. Launching the Test Interface

  1. From the main menu, click on Models.
    models
  2. Then, click on the Test button.
    test

 

2. Selecting a Model

After clicking Test, you will be prompted to select a model.

selectModel

  • Use the dropdown menu to view available models. They are arranged from lower to higher parameter sizes.
  • If you are unsure which model to choose, click Choose for me, and Model HQ will automatically select the most suitable model based on your system specification.

For demonstration, the llama-3.2-1b model has been selected.

 

3. Configuring Test Options

Once a model is selected, you will be taken to the Test Options interface.

testOptions

Here, you can choose from three test types:

  1. Sandbox – Manual testing through direct queries.
  2. Standard – Predefined test samples provided by Model HQ.
  3. Custom – Upload your own dataset in JSON or CSV format.

Sandbox Testing

  1. Select Sandbox Testing (auto-selected by default).
  2. Enter a query in the provided text area.
    sandbox
  3. Click > to run the test.
    sandbox

If the output is unexpected, please reach out to us at support@aibloks.com.

Standard Testing

  1. Select Standard Testing.
    standard
  2. Tests will run automatically using preloaded sample data in Model HQ.
    standardResults

Custom Testing

  1. Select Custom Testing.
    customTesting
  2. Upload your dataset (JSON or CSV).
  3. Click > to begin testing.
    customResults

Custom Dataset Requirements

  • CSV: Must include headers with columns query, context, and answer.
  • JSON: Must include rows with query, context, and answer keys.
    Example:
    {
      "query": "What is the capital of France?",
      "answer": "Paris",
      "context": "France is a country in Europe."
    }
  • The Mapper tool allows you to align your dataset schema with Model HQ's expected format.

 

Additional Features

Mapper

If you are using a custom dataset, click Mapper to align your schema to Model HQ's required format.

  • Required field: query
  • Optional fields: answer, context
mapper

 

Generate Sample

The Generate Sample feature allows you to automatically create test cases.

generateSample

Keep sample sets under 20 entries for optimal results.

Steps to Generate:

  1. Enter a prompt such as:
    "Write 10 samples about U.S. states asking for their capitals."
    samplePrompt
  2. Click > to generate samples in JSON format.
    generatedSample
  3. Once generated, download the dataset.
  4. The generated samples are already mapped and ready to use.
    mappedSample
  5. Click > to begin testing with the generated dataset.

 

Tips & Best Practices

  1. After configuring tests, click >.
    • If the model is not yet downloaded, it will be fetched from the Model HQ Model Catalog (requires internet).
    • If already available locally, it will load directly.
      loadModel
  2. You can also download test results for offline review and analysis.

 

Conclusion

Model HQ makes it easy to test models using sandbox queries, built-in samples, or your own datasets. You can further enhance testing with mapping tools and automated sample generation.

For questions, issues, or feedback, please contact us at support@aibloks.com.