README.md

---
license: llama3
---

Pruning strategy to delete optimal layers of Llama-8B instruct model. Discarded 25% of the layers, model still produces legible text. Next steps include healing PEFT steps