LLama2.java: LLM integration with A 100% Pure Java file

airhacks.fm podcast with adam bien

An airhacks.fm conversation with Alfonso Peterssen (@TheMukel) about:

discussion about Alfonso's early programming experience and participation in the IOI competition, studying computer science and functional programming with Martin Odersky, internships at Google and Oracle Labs working on compilers and the Espresso project implementing a JVM in Java, espresso mentioned in "#208 GraalVM: Meta Circularity on Different Levels", "#194 GraalVM, Apple Silicon (M1) and Clouds", "#167 GraalVM and Java 17, Truffle, Espresso and Native Image" and "#157 The Ingredients of GraalVM", porting LLVM to pure Java in one class, integrating Large Language Models (LLMs) in Java by porting the LLAMA model from C to Java, GPU acceleration with tornadovm, TornadoVM appeared at "#282 TornadoVM, Paravox.ai: Java, AI, LLMs and Hardware Acceleration", performance of the Java port being within 10% of the C versions, potential huge opportunities for integrating AI and LLMs with enterprise Java systems for use cases like fraud detection, the Java port being a 1,000 line self-contained implementation with no external dependencies, the need for more resources and support to further develop the Java LLM integration, the llama2.java project

Alfonso Peterssen on twitter: @TheMukel