---
datasets:
- Sao10K/Claude-3-Opus-Instruct-15K
- Ba2han/DollyLlama-5k
language:
- en
license: mit
---

**Supports system messages!**

**Template: Zephyr**

It's far from perfect but I put together a Phi-3-Medium-4k trained mostly on llama-3-70B generated multi-turn conversations with system messages.


![image/png](https://cdn-uploads.huggingface.co/production/uploads/6324eabf05bd8a54c6eb1650/0WT17jqvTJpYYOnzLau0l.png)

GGUF coming soon.

(Via Unsloth)