36 lines
1.9 KiB
Markdown
36 lines
1.9 KiB
Markdown
---
|
|
license: apache-2.0
|
|
pipeline_tag: text-generation
|
|
language:
|
|
- it
|
|
- en
|
|
tags:
|
|
- chat
|
|
- minerva-7b
|
|
- gguf
|
|
- instruct
|
|
- dpo
|
|
base_model:
|
|
- sapienzanlp/Minerva-7B-instruct-v1.0
|
|
library_name: transformers
|
|
---
|
|
|
|
<div style="text-align: center; display: flex; flex-direction: column; align-items: center;">
|
|
<img src="https://huggingface.co/sapienzanlp/Minerva-7B-instruct-v1.0/resolve/main/minerva-logo.png" style="max-width: 550px; height: auto;">
|
|
</div>
|
|
|
|
# Model Card for Minerva-7B-instruct-v1.0 in GGUF Format
|
|
|
|
Minerva is the first family of **LLMs pretrained from scratch on Italian** developed by [Sapienza NLP](https://nlp.uniroma1.it)
|
|
in the context of the [Future Artificial Intelligence Research (FAIR)](https://fondazione-fair.it/) project, in collaboration with [CINECA](https://www.cineca.it/) and with additional contributions from [Babelscape](https://babelscape.com) and the [CREATIVE](https://nlp.uniroma1.it/creative/) PRIN Project.
|
|
Notably, the Minerva models are truly-open (data and model) Italian-English LLMs, with approximately half of the pretraining data
|
|
including Italian text. The full tech is available at [https://nlp.uniroma1.it/minerva/blog/2024/11/26/tech-report](https://nlp.uniroma1.it/minerva/blog/2024/11/26/tech-report).
|
|
|
|
## Description
|
|
|
|
This is the model card for the GGUF conversion of [**Minerva-7B-instruct-v1.0**](https://huggingface.co/sapienzanlp/Minerva-7B-instruct-v1.0), a 7 billion parameter model trained on almost 2.5 trillion tokens (1.14 trillion in Italian,
|
|
1.14 trillion in English and 200 billion in code). This repository contains the model weights in float32 and float16 formats, as well as quantized versions in 8-bit, 6-bit, and 4-bit precision.
|
|
|
|
**Important**: This model is compatible with [llama.cpp](https://github.com/ggerganov/llama.cpp) updated to at least commit `6fe624783166e7355cec915de0094e63cd3558eb` (5 November 2024).
|
|
|