Ollama Model Training Guide¶
A Docker Compose environment for creating, customizing, and managing Ollama language models — no GPU cluster required. Customize models with Modelfiles (like Dockerfiles, but for LLMs), experiment through a web UI, and deploy your creations to other instances.
Quickstart¶
make preflight # check system requirements
make setup # create .env and directories
make up # start Ollama + web UI
make create-model # build your first custom model (interactive)
Then open:
- Chat UI — http://localhost:8080
- Modelfile Wizard — http://localhost:8080/wizard
- Spreadsheet → JSONL Converter — http://localhost:8080/converter
- Ollama API — http://localhost:11434
Where to go next¶
| If you want to... | Read |
|---|---|
| Understand models, Modelfiles, and parameters | Concepts |
| Install and configure the environment | Installation |
| Look up a command quickly | Quick Reference |
| Create a model without touching a terminal | Web UI |
| Pick between prompting, few-shot, and fine-tuning | Choosing a Method |
| Train on your own Q&A dataset | Dataset Training Example |
| Fix something that broke | Troubleshooting |
The full source, issue tracker, and README live on GitHub.