Llama

🚀 Llama 2

Llama 2 is a collection of generative text models with parameter scales ranging from 7 billion to 70 billion. It offers both pretrained and fine - tuned versions, with fine - tuned models optimized for dialogue.

🚀 Quick Start

To access Llama 2 on Hugging Face, you need to be granted access from Meta first. Then, fill out this form. Before submitting, visit the Meta website and accept the License Terms and Acceptable Use Policy. Requests will be processed within 1 - 2 days.

⚠️ Important Note

Your Hugging Face account email address MUST match the email you provide on the Meta website, or your request will not be approved.

✨ Features

• Multiple Parameter Sizes: Available in 7B, 13B, and 70B parameter sizes, suitable for different application scenarios.
• Optimized for Dialogue: Fine - tuned models, Llama - 2 - Chat, are optimized for dialogue use cases, outperforming many open - source chat models on most benchmarks.
• High - Quality Output: In human evaluations for helpfulness and safety, it is on par with some popular closed - source models like ChatGPT and PaLM.

📚 Documentation

Model Details

Note: Use of this model is governed by the Meta license. In order to download the model weights and tokenizer, please visit the website and accept our License before requesting access here.

Meta developed and publicly released the Llama 2 family of large language models (LLMs). These are a collection of pretrained and fine - tuned generative text models with parameter scales from 7 billion to 70 billion. The fine - tuned LLMs, Llama - 2 - Chat, are optimized for dialogue.

Model Developers: Meta Variations: Llama 2 comes in 7B, 13B, and 70B parameter sizes, with both pretrained and fine - tuned variations. Input: The models accept text input only. Output: The models generate text output only. Model Architecture: Llama 2 is an auto - regressive language model using an optimized transformer architecture. The tuned versions use supervised fine - tuning (SFT) and reinforcement learning with human feedback (RLHF) to align with human preferences for helpfulness and safety.

Llama 2 family of models. Token counts refer to pretraining data only. All models are trained with a global batch - size of 4M tokens. Bigger models (70B) use Grouped - Query Attention (GQA) for improved inference scalability.

Model Dates: Llama 2 was trained between January 2023 and July 2023. Status: This is a static model trained on an offline dataset. Future versions of the tuned models will be released as we improve model safety with community feedback. License: A custom commercial license is available at: https://ai.meta.com/resources/models-and-libraries/llama-downloads/ Research Paper: "Llama - 2: Open Foundation and Fine - tuned Chat Models"

Intended Use

Intended Use Cases: Llama 2 is intended for commercial and research use in English. Tuned models are for assistant - like chat, while pretrained models can be adapted for various natural language generation tasks. Out - of - scope Uses:

• Use in any way that violates applicable laws or regulations (including trade compliance laws).
• Use in languages other than English.
• Use in any other way prohibited by the Acceptable Use Policy and Licensing Agreement for Llama 2.

Hardware and Software

Training Factors: Custom training libraries, Meta's Research Super Cluster, and production clusters were used for pretraining. Fine - tuning, annotation, and evaluation were also performed on third - party cloud compute.

Carbon Footprint: Pretraining used a cumulative 3.3M GPU hours of computation on A100 - 80GB hardware (TDP of 350 - 400W). Estimated total emissions were 539 tCO2eq, 100% of which were offset by Meta’s sustainability program.

Training Data

Overview: Llama 2 was pretrained on 2 trillion tokens of data from publicly available sources. The fine - tuning data includes publicly available instruction datasets and over one million new human - annotated examples. Neither the pretraining nor the fine - tuning datasets include Meta user data. Data Freshness: The pretraining data has a cutoff of September 2022, but some tuning data is more recent, up to July 2023.

Evaluation Results

In this section, we report the results for the Llama 1 and Llama 2 models on standard academic benchmarks. For all evaluations, we use our internal evaluations library.

Overall performance on grouped academic benchmarks.

• Code: We report the average pass@1 scores of our models on HumanEval and MBPP.
• Commonsense Reasoning: We report the average of PIQA, SIQA, HellaSwag, WinoGrande, ARC easy and challenge, OpenBookQA, and CommonsenseQA. We report 7 - shot results for CommonSenseQA and 0 - shot results for all other benchmarks.
• World Knowledge: We evaluate the 5 - shot performance on NaturalQuestions and TriviaQA and report the average.
• Reading Comprehension: For reading comprehension, we report the 0 - shot average on SQuAD, QuAC, and BoolQ.
• MATH: We report the average of the GSM8K (8 shot) and MATH (4 shot) benchmarks at top 1.

Evaluation of pretrained LLMs on automatic safety benchmarks. For TruthfulQA, we present the percentage of generations that are both truthful and informative (the higher the better). For ToxiGen, we present the percentage of toxic generations (the smaller the better).

Evaluation of fine - tuned LLMs on different safety datasets. Same metric definitions as above.

Ethical Considerations and Limitations

Llama 2 is a new technology with risks. Testing to date has been in English and cannot cover all scenarios. Therefore, the model's potential outputs cannot be predicted in advance, and it may produce inaccurate, biased, or other objectionable responses. Developers should perform safety testing and tuning tailored to their specific applications before deploying Llama 2.

Please see the Responsible Use Guide available at https://ai.meta.com/llama/responsible-use-guide/

Reporting Issues

Please report any software “bug,” or other problems with the models through one of the following means:

• Reporting issues with the model: github.com/facebookresearch/llama
• Reporting problematic content generated by the model: developers.facebook.com/llama_output_feedback
• Reporting bugs and security concerns: facebook.com/whitehat/info

Llama Model Index