Thx for sharing llm.java

[otabuzzman]

Your port was quite helpful in terms of file and (expecially) endian handling for my own version of llm.java.

I used Java Stream parallelization as well as you did. In a 2nd step I leveraged TornadoVM to run the methods that implement the layers as CUDA kernels (blog).

Regards Jürgen

[harryjackson]

Hi Jürgen,

Thanks for pinging me about it. I liked your article. I had no idea TornadoVM was even a thing. It looks pretty impressive.

-- Harry

On Wed, Jan 8, 2025 at 1:04 AM Jürgen Schuck @.***> wrote:

Your port was quite helpful in terms of file and (expecially) endian handling for my own version of llm.java https://github.com/otabuzzman/llm.java.

I used Java Stream parallelization as well as you did. In a 2nd step I leveraged TornadoVM https://www.tornadovm.org/ to run the methods that implement the layers as CUDA kernels (blog https://otabuzzman.com/posts/tornado-llmc/).

Regards Jürgen

— Reply to this email directly, view it on GitHub https://github.com/harryjackson/llm.java/issues/1, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAJ24W5JLWZPDWU5Y2JGHZL2JTS2NAVCNFSM6AAAAABUZNVATOVHI2DSMVQWIX3LMV43ASLTON2WKOZSG43TINZVGAZDOOA . You are receiving this because you are subscribed to this thread.Message ID: @.***>