Show HN: TurboGPT: train 22KiB transformer in 13s
TurboGPT is a 22KiB transformer model that can be trained in 13 seconds, requiring CUDA 13.4 and a Windows-only build script. It is inspired by minGPT, a model designed for home experiments. The comment section has 1 post and 8 points.
Read the full article at github.com →