---
license: apache-2.0
---

Experimental Test Model of Llama-3-8B (base)

Finetuned using the ChatML formatting. Only 50% of epoch 1 was done out of two epochs.

Extended to 1 Million context using the PoSE technique.