LLM fine-tuning from a web interface
No-code LLM fine-tuning with LoRA/QLoRA
- Video memory
- 16 GB
- Setup
- ~10 min
- Access
- port 7860
- Billing
- hourly, in BRL
LLaMA-Factory makes fine-tuning approachable: pick the base model, point at your dataset and watch the run on screen. LoRA and QLoRA let you adapt large models on a single card.
LLaMA-Factory is a web UI (LLaMA Board) for fine-tuning open LLMs (Llama, Qwen, Mistral, etc.) with LoRA/QLoRA, no code required. QLoRA of 7B-13B models fits a 24GB GPU (RTX 4090). Includes example datasets, evaluation and adapter export.
What it is for
- Teach the model your company's vocabulary and tone
- Fine-tune on your own support conversations
- Train with LoRA on a single card and cut the cost
- Export the weights and serve them with vLLM right after
How to deploy
- Create your account and add balance (card or Pix, no subscription).
- In the console, pick the LLaMA-Factory template and a machine — the console hides the ones that do not meet the requirement.
- In about 10 minutes the setup finishes and the access address shows up in the panel, on port 7860.
Done? Just destroy the machine and billing stops with it. No contract, no minimum commitment.
FAQ
How much video memory do I need?
With QLoRA, 16 GB already fine-tunes mid-sized models. Larger models or full training want 40 GB or more.
How many examples do I need?
A few thousand well-chosen examples usually beat tens of thousands of poorly filtered ones.
Where do the trained weights end up?
On your machine, ready to download or serve right there.