Is LoRA always the answer for your task and base model? Do better trade-offs exist?
However, covering the dozen PEFT techniques without a unified framework is ... difficult!
So, we put ๐ค PEFT lib to work and found out the obvious -- LoRA isn't always the best ๐ฅฒ
My information consumption is now 1/4 X, 1/4 podcast interviews of the smartest practitioners, 1/4 talking to the leading AI models, and 1/4 reading old books. The opportunity cost of anything else is far too high, and rising daily.
@MokadyRon I agree, I feel like there is a missing middle between very general instructions (text prompts), and extremely precise instructions, like controlnet canny, where every line in the input needs to be matched by the output.