Entry Point AI Docs
  • Getting Started
    • Introduction
    • Quickstart
  • Key Concepts
    • Model Providers
    • Templates
    • Fields
    • Examples
    • Models
    • Evaluation
    • Transforms
  • Guides
    • Evaluate a Prompt
      • Turn a Prompt into a Template
      • Prepare Validation Examples
      • Create a Templated Model
      • Review & Rate Outputs
      • Iterating on the Prompt
    • Build a Dataset
      • Import a CSV
      • Transform Data
      • Synthesize Examples
      • Quantity & Quality of Data
    • Fine-tune a Model
      • Start a Fine-tune
      • Fine-tuning Hyperparameters
      • Compare Fine-tunes
      • Next Steps
    • Generate Completions
      • Playground
      • Inference Parameters
      • Shareable Link
      • Deploy a Model
Powered by GitBook
On this page
  • Test Different Models
  • Try Different Templates
  • Expand Your Dataset
  • Share the Model
  • Deploy It
  1. Guides
  2. Fine-tune a Model

Next Steps

PreviousCompare Fine-tunesNextGenerate Completions

Last updated 9 months ago

Now that you have a fine-tuned model and have an evaluation score, you may be asking what to do next. Here are a few ideas.

Test Different Models

Experiment with smaller and larger models. Usually it's best to choose the smallest model that gets you the desired quality of output, because it will be cheaper and faster to run.

Try Different Templates

You may want to try templates that include instructions, basic labels, or no content besides your field references. These can all have an impact on how the model performs.

Expand Your Dataset

Use the tool to increase the size of your training and validation datasets. Use the alignment text to instruct the model on edge cases to generate data for.

Share the Model

Want to get a better idea of how other people will use your model in the wild? You can easily generate a shareable link under the model details. All completions from the shared model will be tracked on the Completions page.

Deploy It

Your model is hosted by the you selected and can be used for inference through the various APIs that model providers offer. In some cases, depending on the model provider, you can download the weights and take your model anywhere you like for inference.

Synthesis
model provider