Fine-tune with LoRA
Upload a dataset, pick a base model, and queue a low-rank training run.
No file selected
Training starts once your account is verified.
1
Upload a JSONL dataset
One JSON object per line:
{"prompt": "…", "completion": "…"}
2
Choose base model & rank
Rank 16–64 covers most tasks. Higher rank = more capacity and more VRAM.
3
Run queues automatically
Training takes 15–60 min. The adapter lands in your private registry and is ready to pull or serve.