---
library_name: transformers
tags: []
---

# Malaysian Llama-3 8B 65536 context length

65536 context length and 15300000 RoPE Theta.

WanDB, https://wandb.ai/huseinzol05/EasyContext-65536?nw=nwuserhuseinzol05

Source code, https://github.com/mesolitica/malaya/tree/master/session/llama3#extend-1m-context-length

Special thanks to https://github.com/jzhang38/EasyContext for wrapping https://github.com/zhuzilin/ring-flash-attention for distributed training!