|
--- |
|
license: apache-2.0 |
|
language: |
|
- es |
|
--- |
|
|
|
Llama 2 (7B) fine-tuned on a [own Spanish instructions dataset](https://huggingface.co/datasets/dariolopez/Llama-2-databricks-dolly-oasst1-es). |
|
|
|
On this repo you can find 4-bit and 5-bit quantized versions of the [Llama 2 (7B) Spanish fine-tuned](https://huggingface.co/dariolopez/Llama-2-databricks-dolly-oasst1-es-axolotl). |
|
|
|
# How to use |
|
|
|
```sh |
|
git clone https://github.com/ggerganov/llama.cpp |
|
cd llama.cpp && git pull && make clean && make |
|
git clone https://huggingface.co/dariolopez/Llama-2-databricks-dolly-oasst1-es-axolotl-GGUF |
|
./main -m ./llama-2-databricks-dolly-oasst1-es-axolotl.gguf.q4_k_m.bin -n 2048 --color --temp 0 -ngl 35 -p "<s>[INST] Describe 5 lugares para visitar en España: [/INST]" |
|
``` |
|
|
|
# Based on |
|
|
|
https://mlabonne.github.io/blog/posts/Quantize_Llama_2_models_using_ggml.html |
