Fine-Tuning Guide

What is fine-tuning?

Fine-tuning adapts a pretrained model to your domain and style. Mycelis uses LoRA for efficient adaptation.

Data format

Use JSONL with chat messages:

{"messages":[{"role":"user","content":"What is our return policy?"},{"role":"assistant","content":"Returns are allowed within 30 days..."}]}

Recommendations

Start training

1. Dashboard → Fine-Tuning → New Training
2. Select base model
3. Upload JSONL
4. Set hyperparameters (epochs, LR, LoRA rank, batch size)
5. Start

You can also use deployment wizard for train + auto-deploy.

Next steps

• Create a dedicated deployment
• Configure an Agent