File size: 2,287 Bytes
84478e2
 
b1476ae
 
84478e2
 
b1476ae
84478e2
4459505
b1476ae
84478e2
3ba5541
 
 
b1476ae
0ef5846
b1476ae
3ba5541
7f76738
d9adc65
 
 
2a6acfe
3ba5541
 
 
0ef5846
 
2a6acfe
b1476ae
0ef5846
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
---
title: README
emoji: 🦙
colorFrom: blue
colorTo: blue
sdk: static
pinned: true
---
# The Llama Family
*From Meta*

Welcome to the official Hugging Face organization for Llama, Llama Guard, and Prompt Guard models from Meta! 

In order to access models here, please visit a repo of one of the three families and accept the license terms and acceptable use policy. Requests are processed hourly.

In this organization, you can find models in both the original Meta format as well as the Hugging Face transformers format. You can find:

Current:
* **Llama 3.2:** The Llama 3.2 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction-tuned generative models in 1B and 3B sizes (text in/text out). 
* **Llama 3.2 Vision:** The Llama 3.2-Vision collection of multimodal large language models (LLMs) is a collection of pretrained and instruction-tuned image reasoning generative models in 11B and 90B sizes (text + images in / text out)

History:
* **Llama 3.1:** a collection of pretrained and fine-tuned text models with sizes ranging from 8 billion to 405 billion parameters pre-trained on ~15 trillion tokens.
* **Llama 3.1 Evals:** a collection that provides detailed information on how we derived the reported benchmark metrics for the Llama 3.1 models, including the configurations, prompts and model responses used to generate evaluation results.
* **Llama Guard 3:** a Llama-3.1-8B pretrained model, aligned to safeguard against the MLCommons standardized hazards taxonomy and designed to support Llama 3.1 capabilities.
* **Prompt Guard:** a mDeBERTa-v3-base (86M backbone parameters and 192M word embedding parameters) fine-tuned multi-label model that categorizes input strings into 3 categories - benign, injection, and jailbreak. It is suitable to run as a filter prior to each call to an LLM in an application.
* **Llama 2:** a collection of pretrained and fine-tuned text models ranging in scale from 7 billion to 70 billion parameters.
* **Code Llama:** a collection of code-specialized versions of Llama 2 in three flavors (base model, Python specialist, and instruct tuned).
* **Llama Guard:** a 8B Llama 3 safeguard model for classifying LLM inputs and responses.

Learn more about the models at https://ai.meta.com/llama/