Train Llama 3 models from scratch. Any scale, any personality. By Arianna Method.

A minimal, hackable Vision-Language Model built on Karpathy’s nanochat — add image understanding and multimodal chat for under $200 in compute.

Train your own ChatGPT on Apple Silicon — MLX port of nanochat

nanochat's inference engine re-vibed in C++ with GGML.

Ascend NPU fork of nanochat for LLM training with torch_npu/HCCL (experimental)

I built this repo to prove to my granny that I can implement GPT.

Rust-first continuation of nanochat-rs, benchmarked against karpathy/nanochat for faster, reproducible LLM training, evaluation, and inference.

Implementing Karpathy's nanochat in rust.

I see now that the path I choose through the maze makes me what I am.

implementing a Language Model Chat from scratch

This script uses Modal to run Karpathy's nanochat training speedrun on cloud GPUs (8× H100s).

Run nanochat training efficiently on Huawei Ascend NPUs with minimal code changes, supporting tokenizer, pretraining, and evaluation workflows.