Fast & efficient BPE tokenizer written in C & python for LLM tranining

Carbon is a pure C++ Transformer framework inspired by GPT, featuring SIMD-optimized tensor math, multi-head attention, feedforward layers, and BPE tokenization. It’s a fully self-contained system for training and running language models without external modules or libraries.

A c++ framework on efficient training & fine-tuning LLMs

MobileFineTuner: Native C++ framework for fine-tuning LLMs directly on mobile devices. Features: LoRA/Full-FT, ZeRO-inspired parameter sharding, energy-aware throttling, custom autograd engine. Keep your data on-device.

Real-time, low-overhead visualization of LLM internals during training.

Quantized LLM training in pure CUDA/C++.