Clojurians Log v2

Java interops works well from Clojure. In combination with https://github.com/clj-commons/virgil you can as well adhoc write and wrap the Java code in Java "public static xxxx" fns and call them from Clojure That what I do when using a big Java Library.

Awesome, great tip, thanks

What kind of models are you interested in? If neural networks - how big? In the LLM space, I think there hasn't been much need to implement Clojure or Java native versions, because there is no real barrier or downside to using Clojure with existing implementations. For LLMs Clojure already has equivalency to Python/C/JavaScript/TypeScript already. For LLMS: • Fine tuning can often be done via command line or YAML config (no need to write code that talks directly to fine tuning logic). • Inference is done via standard REST API or shell command line invocation or libpythonclj Python interop or https://github.com/phronmophobic/llama.clj . LLMs are so big, high latency and slow there is practically no benefit to embedding them in process to cut out the HTTP overhead (because that's such a tiny/negligible fraction of the cost/time/latency).