Aquin LogoAquinLabs

Recipe

recipe.yaml is the full train API for built-in methods. You do not write Python unless you drop methods/<name>.py. The kernel reads the recipe; aq does not override it.

PrerequisiteA train folder with recipe.yaml · kernel deps: pip install -r aq/kernel/requirements.txt

recipe.yaml · tabular

family: tabular with method linear | logistic | ridge | lasso | elasticnet | tree | forest | boosting | gp. Backed by scikit-learn (boosting prefers XGBoost → LightGBM → CatBoost → sklearn).

example
family: tabularmethod: lineardata:  path: data.csv  target: yeval:  metric: mse  min_score: null

recipe.yaml · llm / lora / qlora

family: llm with method llm | lora | qlora. model: is required (hub id or path). size: is a label only. QLoRA needs CUDA + bitsandbytes.

example
family: llmmethod: loramodel: meta-llama/Llama-3.2-1B-Instructobjective: lorarank: 16alpha: 32steps: 100lr: 2.0e-4data:  path: data.jsonl  prompt: prompt  completion: completioneval:  metric: loss

Objectives that work: next-token, sft, full-ft, lora/qlora, fim, mlm, span, continued-pretrain. Fail closed (not faked): formats, speculative, paged_kv, objective: mtp.

recipe.yaml · transformer

method: transformer with arch encoder | decoder | encoder-decoder. model: required.

example
method: transformermodel: bert-base-uncasedarch: encodersteps: 50data:  path: data.csv  text: text  target: labeleval:  metric: accuracy

recipe.yaml · guard

Opt-in fail-closed watches mid-train. See Metrics & guard.

example
guard:  safety: true   # NaN / blow-up  leak: true     # train vs evals overlap