Model Factory

Model Factory is a powerful feature in OllaMan that allows users to create customized model replicas based on existing local models by adjusting System Prompts and inference parameters.

Features Overview

• Customize Based on Existing Models: Choose any local model pulled via Ollama as a base.
• Custom System Prompts: Set specific personas, rules, or task instructions for the model.
• Micro-tuning Parameters: Precisely control the model's creativity and stability (Temperature, Top P, Top K).
• Preset Templates: Built-in common prompt templates (Coding Assistant, Creative Writing, Data Analysis).

Interface Guide

The Model Factory interface allows for configuration via a left panel and a prompt editor on the right.

1. Model Identity

• Base Model: Select an installed local model as the base. Note: Cloud models are not selectable.
• New Model Name: Give your custom model a unique name.

2. Parameters

Adjust the following parameters to control the output style of the model:

• Temperature (0.0 - 2.0): Controls randomnes.

  • Lower values make the output more focused and deterministic.
  • Higher values make the output more random and creative.
  • Default: 0.7

• Top P (0.0 - 1.0): Nucleus Sampling.

  • Controls the model to consider only the top tokens with cumulative probability P.
  • Default: 0.9

• Top K (1 - 100):

  • Limits the model to select from the top K most probable tokens.
  • Default: 40

3. System Prompt Editor

This is the core area for customizing the model. Enter instructions here to define the model's behavior.

• Markdown Support: The editor supports Markdown syntax highlighting.
• Character Count: Real-time display of prompt character count.
• Quick Presets: If empty, click the preset cards below to quickly fill:
  • Coding Assistant: Focuses on code writing and explanation.
  • Creative Writing: Suitable for story creation and brainstorming.
  • Data Analysis: Focuses on logical reasoning and data processing.

Creation Process

1. Select a Base Model in the left panel.
2. Enter a New Model Name.
3. (Optional) Write or select a System Prompt in the right editor.
4. (Optional) Adjust Inference Parameters.
5. Click the Create button at the bottom left.
6. Wait for completion; a green success message will appear.

Once created, you can find this new model in the "Local Models" list and immediately use it in the "Chat" feature.

Additionally, you can manage this model in the "Local Models" page, where you can push it to the remote Ollama repository to share with others or backup.