nlptransformersllm
Finetuning LLMs
Full finetuning, LoRA, QLoRA, instruction tuning — adapting pretrained models to specific tasks.
Уровни глубины
L0Intro~0ч
Knows finetuning adapts a pretrained model on a smaller task-specific dataset.
L1Basics~10ч
Runs a LoRA finetune with HuggingFace PEFT on a classification or generation task.
L2Working~25ч
Applies instruction tuning, chat formatting, DPO/SFT pipelines; evaluates with benchmarks; handles catastrophic forgetting.
L3Advanced~40ч
Designs data pipelines for quality-focused finetuning; applies continual learning, model merging, and multi-task finetuning.
L4Research~80ч
Contributes to alignment, efficient adaptation, or model merging research.
Ресурсы
L2 — Working