Chad Brewbaker

You need to release a suffix array of the training corpus to do it properly. This is also useful in designing hypothetical copyright filters.

Perfect ticket for @gdb. Looks like it is based on [TinyEMU](https://bellard.org/tinyemu/). To get this project FedRamp certified, might make sense to work with https://github.com/amazonlinux/amazon-linux-2023/issues/102 and have FedRAMP approved AWSLinux2023 VM...

@copy project seems far better maintained than jslinux. Just need a stripped down Ubuntu 20.04 x86 image to be binary compatible with OpenAI. https://github.com/copy/v86 Digging into the guts of jslinux,...

Seems NVIDIA acquired them? https://www.pgroup.com/index.htm now at https://developer.nvidia.com/hpc-sdk . New corp email is [email protected] . New license https://docs.nvidia.com/hpc-sdk/eula/index.html

I was discussing this yesterday with @jonmasters. Ideally this would be a script that takes llm.c and transforms it into specialized but still legible C code for a particular architecture....

I have this in mind for the Mojo target issue - which is really about having the Makefile support composability like the one for llama.cpp. Probably copy-pasta most of what...

I would do MPI-2 as MPI IO is all you need and it is most widely supported.

Make a fork. Happy to help. https://github.com/tairov/llama2.mojo The way to do it would be to add to the Makefile like https://github.com/ggerganov/llama.cpp/blob/master/Makefile Then as part of the build you pass LLM_USEMOJO...

Minimal fork just to show the Makefile. I'll work on porting matmul_backward() and linking it. https://github.com/chadbrewbaker/llm.c/tree/mojo

Lol. "Fast math" does crazy rounding modes - best to avoid unless you know what you are doing. Just did a deep dive and Modular/Mojo still don't have their C...