I wanted to share this little flowchart I made 🙂 Let me know if you have suggestions or comments! Curious how everyone else is thinking about this

This flowchart is a great visual guide! It perfectly balances the decision-making process between product design tweaks and more technical adjustments like fine-tuning. Really appreciate the emphasis on understanding the problem deeply before jumping into fine-tuning. Super helpful for ensuring LLM resources are used wisely! 😊

This is awesome. Thanks for sharing :)

So much wisdom in this block:

@Roheet Bakare @Gavin @Don Alvarez thanks for the kind thoughts! Glad you all enjoyed it 🙂

I’ve been considering finetuning llama3.170b for my specific use case, which involves highly specialized Salesforce logic. This area has a smaller online presence compared to widely-used open-source databases like PostgreSQL. We’ve successfully implemented a pipeline that incorporates RLHF (Reinforcement Learning from Human Feedback) to assist with our fine-tuning process. However, I have one main concern: What if Sam releases a superior model in November? Would that render my investment of time, effort, and money in fine-tuning obsolete?