4 ms·
For me personally, being able to fine-tune the local LLM's at a much higher rank and training more layers is very useful for (somewhat unreliably) embedding inf
by muttled 3y ago
For me personally, being able to fine-tune the local LLM's at a much higher rank and training more layers is very useful for (somewhat unreliably) embedding information. AFAIK the OpenAI fine-tuning is more geared towards formatting the output.
- bigfudge 3y agoAs I understand it, fine tuning is never really about adding content. RAG and related techniques are likely cheaper/better if that’s what you want.
- antupis 3y agoyup how I understand fine tuning is more about adding context and bigger picture. RAG is more about adding actual content. Good system probably needs both in long run.