The A.I. Daily
TurboGPT
Train a 22KiB byte-level GPT in 13s in CUDA C++, for home experiments.
TurboGPT
2 storiesEvery TurboGPT story collected by The AI Daily, 2 so far, newest first, refreshed hourly.
Related topics
TurboGPT trains 22KiB transformer in 13 seconds
TurboGPT, an open-source project inspired by minGPT, trains a 22KiB byte-level tiny GPT in 13 seconds using CUDA C++, under the MIT license. It requires CUDA 13.4 with a Windows-only build script, and reaches 2.5295 BPB after 1.5G training tokens on the hn1g dataset.
That is everything