3 ms·
Ask HN: What do you find most difficult about fine tuning AI models?
Those of you who have fine tuned NLP (GPT-X/Llama) or diffusion models, what was the most difficult or annoying part about the process?
I'm trying to help solve problems in this field, so I'm curious about your experiences.
- tlack 3y agoOn the LLM side: the cost / amount of resources that need to be brought to bear for long periods of time On the Stable Diffusion side: labeling, best practices, file management, quality control