A hands-on Quarkus tutorial using MLX LM on a Mac to train a small adapter, compare base and adapted behavior, and make a local model honor a private agent contract.
Thanks a lot Markus for this post, it did help me understand how this works and tie together both java, python and LLMs easier than i could have thought of. Going to test other models ( qwen2.5 is great for this ) to compare.
Would be nice that qlora stuff were performed with jbang and Quarkus but Java probably still lacks proper model training libraries, thanks for this tutorial, it’s good to know when to use model training and when retrieval or tools are a better options, another day, another thing learned
Yeah. Multiple reasons. I had to get this out quickly because I had promised this to someone. On the other hand, I’m a big proponent of using the right tool for the right job. It’s not a religious decision to do ML with Python. 🤷🏼♂️ could sure have done it with Java though. Might do a follow up.
Thanks a lot Markus for this post, it did help me understand how this works and tie together both java, python and LLMs easier than i could have thought of. Going to test other models ( qwen2.5 is great for this ) to compare.
Thanks for reading and the feedback!
Would be nice that qlora stuff were performed with jbang and Quarkus but Java probably still lacks proper model training libraries, thanks for this tutorial, it’s good to know when to use model training and when retrieval or tools are a better options, another day, another thing learned
Yeah. Multiple reasons. I had to get this out quickly because I had promised this to someone. On the other hand, I’m a big proponent of using the right tool for the right job. It’s not a religious decision to do ML with Python. 🤷🏼♂️ could sure have done it with Java though. Might do a follow up.
Thanks for the feedback!