"No Dependencies" Java 21 / LLM / Llama3 Integration

llama3.java (github.com/mukel/llama3.java) walkthrough.
Also checkout the airhacks.fm podcast with Alfonso: the creator of llama2.java: "LLama2.java: LLM integration with A 100% Pure Java file" 👉 airhacks.fm/#episode_294 📻📡
Java used in this short: openjdk version "21" 2023-09-19 LTS
Visual Studio Code used for editing: code.visualstudio.com/
See you at live, virtual workshops: airhacks.live
About me: airhacks.industries
#airhacks

[@sergeyshvyrev969]

Мега круто!!!! Спасбо, Адам!!!! ))))

[@bienadam]

I only started the chat 😀 . The integration was built by Alfonso Peterssen. Please "star" Alfonso's repository: github.com/mukel/llama3.java

[@brunofsleite]

Top

[@bienadam]

Thank you for watching! ...and this is just a start. A lot is going on in Java space right now.

[@billelRedouane86]

Good job it's works fine! but I have this error messages and not sure why?!
- 1.05 tokens/s (27) Ran out of context length...
- it's very very slow

[@AlfonsoPeterssen]

By default --max-tokens=512 but you can set it to -1 to match the max context length of the model (8192).
Shouldn't be that slow ~1t/s is what I get on a Raspberry PI 5.
The Vector API is only supported on vanilla OpenJDK, please make sure that's the one you use.

[@flaviuvanca9289]

Hi Adam, I came across your channel last week, and I have to say there are some great videos on it, of course not for beginners.
I started learning Java a while back, and achieved ocp 8 back in Nov 2021. I still feel at beginning of my journey and can't seem to find my place in the industry yet. I want to dive in the world of microservices to pursue a specialization. Would you have suggestions?