Model-training games & apps

Skills Creator

NEW: Model Training & Fine-Tuning skill (26KB) — complete engineering playbook covering LoRA/QLoRA/DoRA comparison with real benchmarks, dataset engineering (quality checklist, ChatML format, hard example mining), production training scripts for consumer GPUs, DPO and RLHF, LLM-as-Judge evaluation, catastrophic forgetting detection, and deployment with vLLM/llama.cpp. Plus war stories: $50K model that was worse than base, overfitting disaster, forgetting that broke a chatbot.