---
license: llama2
---

13b CodeLlama model finetuned on HQ + self-play augmented data to optimize C++ programs. For more details on how the data was collected and training was done, refer to [this webpage](https://pie4perf.com/).