llm.f

f is for Fortran. Here I'm implementing language models from scratch in Fortran. Inspired by llm.c

I contribute general purpose layers to neural-fortran deep learning library. Therefore, it is among this project deps (but it's still tiny).

Installation

fpm build

Progress

ML

Layer	Status	Forward	Backward	Llama	Qwen
Llama Attention	✅	✅	✅	✅	✅
Silu MLP	✅	✅	✅	✅	✅
RMSNorm	✅	✅	✅	✅	✅
Decoder Layer	✅	✅	✅	✅	✅
Llama Model	⌛	✅	⌛
KV Caching	❌	❌	-
Text Generation	❌	❌	-
Training	❌	❌	❌

Infrastructure

1. Code fo loading safetensors. Will need to make Rust and Fortran friends
2. BPE Tokenizer. In progress
3. Graphics cards. Need to choose between C CUDA, Fortran CUDA and OpenACC
4. CMakeList