DigiNews

Tech Watch by Johan Denoyer

← Back to articles

Show HN: NanoEuler – GPT-2 scale model in pure C/CUDA from scratch

Quality: 8/10 Relevance: 9/10

Summary

NanoEuler is a GPT-2-class language model built entirely from scratch in C/CUDA with hand-written forward and backward passes and a full training pipeline. It runs a CUDA engine capable of training ~116M parameters on a single RTX 4070 and is presented as an educational, open-source artifact rather than a production-ready model.

🚀 Service construit par Johan Denoyer