LoRA Desk

Check a LoRA or QLoRA training config before it burns a GPU-day. Paste Unsloth or PEFT + TRL Python, an Axolotl YAML, JSON or CLI flags: free in your browser, every adapter and trainer knob is read with its line and checked against the LoRA Without Regret recipe - all-linear targets, lora_alpha = 2 x r, rank by task, LoRA learning rates, effective batch under 32, bf16 never fp16, the current TRL API and Unsloth's traps - with the adapter's size and a per-GPU memory floor. A metered run reviews it knob by knob, or rewrites it and the page rescans the rewrite. Built on the SkillSafe skill @wshobson/lora-qlora-recipes.