What is this?

This is an open-source fine-tuned model that is fine-tuned using a dataset of selected writings by Ziping Liu. The purpose here is to create a chat AI that can respond in a specialized way based on Ziping’s writings and, in this case, in a way that responds like Ziping Liu himself and can reason out new ideas based on Ziping’s writings. Given the possibility of responding incorrectly, this model is in the development phases and does not represent Ziping Liu.

Why did I do this? Well, I first fine-tuned models via OpenAI’s API but found that it was too restrictive in fine-tuning given that it only allows certain things and requires a specific format for the dataset used in fine-tuning, and uses OpenAI hardware from the cloud; thus, with no ability in doing things locally. So I then sought out open-source variants and found an open-source model that could match GPT-4’s abilities to some extent and then fine-tuned it. This was all done on my hardware, in which case the fine-tuning was done using my RTX Titan 24GB GPU. For inferencing, I utilize a cheaper Tesla P40 GPU. You can read more about my experience using Tesla GPUs here.

This model is currently undergoing maintenance for upgrades to a newer version, is currently in testing phases, and is useable right below as a pre-production release. Read more about the update at linkedinliu.com

This AI, nicked-named “synthethics,” responds and emulates Ziping’s written works but does not represent Ziping’s official views or positions. It is provided for educational and/or research purposes only. Learn more here.