This is a GPT-2 model trained in llm.c for 100K steps (of 1M batch size) on FineWeb-EDU.
A lot more detailed information is here: https://github.com/karpathy/llm.c/discussions/677 .
This model is exactly as the post above, except changing -x 100000 to run 100K steps. The model achieves HellaSwag of 57.7