# GAAI model adaptation and multimodal evaluation lab

Run `python 11-Blog-Programme/labs/gaai-adaptation/adaptation_cases.py` from the repository root. The lab executes twenty deterministic decisions and calculations spanning adaptation choice, split integrity, trainable parameter scope, paired evaluation, quantization selection, resource admission, licensing, multimodal metrics, human correction and experiment reporting.

No checkpoint is downloaded or loaded. No training, fine-tuning, quantization, model inference, OCR, speech transcription or image processing executes. Quality, memory, latency, cost and output records are authored fixtures unless described as theoretical parameter arithmetic. They do not establish performance of LoRA, QLoRA, a checkpoint, a provider or NeuraPath.

The verifier must pass all 20 checks and save `adaptation-verification.json`. License, legal, instructor and editorial review remains pending.