Introducing LoRA fine-tuning

Learn what LoRA is, when it makes sense, and how to run your first training job on Mycelis.

When LoRA is worth it

• when your agent must use consistent domain language
• when style and behavior need to stay stable
• when prompt-only approaches are not enough

Prepare data

1. Collect high-quality example dialogs.
2. Keep style and format consistent.
3. Remove duplicates and noisy samples.

Start training

1. Open Fine-Tuning in Mycelis.
2. Select base model and dataset.
3. Start the job and monitor logs.

Evaluate output

1. Compare answers against base model.
2. Test hard prompts and edge cases.
3. Deploy only when quality criteria are met.