epscylonb/1386.ai.rocm

A lightweight transformer language model built from scratch in PyTorch, trained on a single consumer GPU with a full pipeline for data processing, pretraining, and instruction tuning.

1 best points · 0 comments · 1 Hacker News post · 7 GitHub stars · View on GitHub

Hacker News discussions

• Porting a Scratch-Built 500M LLM Training Pipeline to ROCm on Strix Halo (1 points, 0 comments, by thomasfromcdnjs, 2026-04-28)