Alfonso² Peterssen

icon company Oracle Labs icon location Zürich, Switzerland icon location @graalvm padawan at Oracle Labs. Working on Espresso: a meta-circular Java bytecode interpreter for GraalVM. Java in Java by day, LLMs in Java by night.

Results 5 issues of Alfonso² Peterssen

Support multipart requests

7 comment

Very slick library, but can't send multipart requests out-of-the-box. Other libraries I've tried do support it but have other caveats: - [RosHttp](https://github.com/hmil/RosHTTP) but not as nice as this one (and...

enhancement

Sanitize class files

CC: @manoskouk By adding duplicated methods/fields is possible to generate ill-formed class files without any warnings; failing at runtime with ClassFormatError. Tracking duplicated methods/fields is a bit hard since methods...

Update to Scala 2.13

1 comment

Please publish artifacts for the latest Scala 2.13.0, it should be rather straightforward to update.

Improve matrix multiplication using the Java Vector API on Apple silicon.

1 comment

llama.cpp runs incredibly fast on Apple silicon, I ran a build with pure CPU, and it is closer to the memory bandwidth e.g. 28 tokens/s on an M3 Pro. llama3.java...

Bundle all standalone models in a single project.

So far I've ported the following models to Java: Llama 3 & 3.1, Mistral/Codestral/Mathstral/Nemostral (+ Tekken tokenizer), Qwen2, Phi3 and Gemma 1 & 2 ... All models are bundled as...