Your embedding model can be different from your text generation model
Clearing up a misconception that I had about LLMs, which you may have had too
Hello fellow datanistas!
Have you ever wondered if your embedding model has to be from the same provider when doing retrieval augmented generation? I used to think so too, but it turns out, we were both wrong! I've written a blog post that dives into this topic and dispels this common misconception. I believe it will add value to your data science journey.
In the post, I also share my experiences with different embedding models and why I now use SentenceTransformer. It's a practical, cost-effective choice that works well with my setup. I've even updated LlamaBot with this new knowledge, and I'm excited for you to see how it now freely composes embeddings from various models with text generation performed by other models.
You can read the full post here. If you find it helpful, I encourage you to share it with others who might benefit from this information. Let's continue learning and growing together in this exciting field of data science. Looking forward to hearing your thoughts on the post!
Cheers,
Eric