1.karpathy / llm.cLLM training in simple, raw C/CUDACudaMIT31K total3.8K forks+747 starsfrom 30,640 · score 6.62
2.HigherOrderCO / HVM1HVM1 (2022): a massively parallel, optimal functional runtime in RustCudaMIT11K total437 forks+54 starsfrom 11,297 · score 4.01