WingNews logo WingNews
top | item 37047081

Llama2.java: Karpathy's llama2.c ported to Java

33 points| mukel | 2 years ago |github.com

18 comments

order

gavinray|2 years ago

The Java code is impressively written, using newer features like MemorySegment.

Looked at the author and realized it's Alfonso from the Graal team -- makes sense.

I wonder whether the "matmul" code could be further optimized with the Vector API and SIMD.

mukel|2 years ago

Author here: I implemented several versions of matmul with different unrolling schemes using the Vector API and I got a ~4X speedup with a single thread, but the speedup fades the more threads you add. I think that performance is constrained by memory bandwidth which is saturated with a small number of threads, regardless of vectorization.

kurhan|2 years ago

Also new virtual threads might be beneficial. I was experimenting using Vector api for matrix multiplication once and effect was pretty good.

atairov|2 years ago

Thanks for sharing this! It's great to have a reference implementation written on java lang. With given original simplicity it's really easy to follow llama architecture logic.

Just in case if anyone interested in Python version, I spend some time on weekend and ported it to pure python -- https://github.com/tairov/llama2.py

I never knew that it would take about 500 lines of core part code to implement inference for such a cutting edge AI technology.

mukel|2 years ago

A Java port of llama2.c that performs very close to C on large models. Llama 2 7B runs at a whooping 1.6 tokens/s.

mike_hearn|2 years ago

Hey man, awesome stuff. Surely any JIT compiler will struggle to vectorize something using IntStream.range, though? Looking at matmul, I'd not expect that to be auto-vectorized. The Panama API can be used to do a matmul vectorization, too bad it seems to never launch.

shortrounddev2|2 years ago

How you all used these things for anything useful? I can't get them to give useful results on my 3060 8gb. If I wanted to get decent results I think I'd need to rent a GPU node somewhere, but chatGPT is still free

SushiHippie|2 years ago

The 4bit quantized 13B models, give really decent answers (not as good as gpt4, but often as good as gpt 3)

nmfisher|2 years ago

I know it might be asking a lot, but it would be great if someone could put up a HF space so I could try all the various flavours/sizes.

jiehong|2 years ago

This makes me wonder: what’s the status of GPU programming on the JVM?

Any abstraction for GPGPU or shaders programming?