image-20230826213922965

起因

OpenAI 3.5 發布 Fine-Tuning API 後,另外一邊 Andrew Ng 馬上上線相關課程: https://deeplearning.ai/short-courses/finetuning-large-language-models/

Why Finetune

附上大家容易有疑慮的比較圖:為何弄 Fine-Tune 為何不透過 Vector Prompting

Image

image-20230826220610545

  • 確認原本 LLM 不能完成
  • 尋找 gold samples
  • 確認可以比較好結果
  • Finetune it.

Buy Me A Coffee

Evan

Attitude is everything