README.md

---
license: apache-2.0
---

# ggml versions of OpenLLaMa 3B

- Version: 600b token preview
- Project: [OpenLLaMA: An Open Reproduction of LLaMA](https://github.com/openlm-research/open_llama)
- Model: [openlm-research/open_llama_3b_600bt_preview](https://huggingface.co/openlm-research/open_llama_3b_600bt_preview)
- llama.cpp: [#1588](https://github.com/ggerganov/llama.cpp/pull/1588)

## Use with llama.cpp

Since it is experimental and not merged yet, you can use the PR:

```sh
git clone https://github.com/ggerganov/llama.cpp
cd llama.cpp
curl -L https://github.com/ggerganov/llama.cpp/pull/1588.diff | git apply -
make main
./main -m open-llama-3b-q4-0.bin ...
```