
https://ai.meta.com/blog/meta-llama-3/
Meta developed and released the Meta Llama 3 family of large language models (LLMs). It is the most capable openly available LLM to date.
Providers
LLaMA 3 models will soon be available on a range of platforms, including AWS, Databricks, Google Cloud, Hugging Face, Kaggle, IBM WatsonX, Microsoft Azure, NVIDIA NIM, and Snowflake, with support from hardware platforms offered by AMD, AWS, Dell, Intel, NVIDIA, and Qualcomm.
Getting Started
This blog post provides a comprehensive resource guide to help you get started with LLaMA 3 quickly. Below, you'll find links to essential information and tools to help you explore the capabilities of LLaMA 3.
What is LLaMA 3?
Meta's LLaMA 3 family of large language models consists of pretrained and instruction-tuned generative text models in 8B and 70B sizes. The instruction-tuned models are optimized for dialogue use cases and outperform many open-source chat models on common industry benchmarks.
Llama 3 comes in two sizes — 8B and 70B parameters — in pre-trained and instruction tuned variants.
| Model | Training Data | Params | Context Length | GQA | Token Count | Knowledge Cutoff |
|---|---|---|---|---|---|---|
| LLaMA 3 | New mix of publicly available online data | 8B | 8k | Yes | 15T+ | March, 2023 |
| LLaMA 3 | New mix of publicly available online data | 70B | 8k | Yes | 15T+ | December, 2023 |
Playgrounds
Get hands-on with LLaMA 3 Chat in these interactive playgrounds:
- HuggingChat: https://huggingface.co/chat/
- Perplexity: https://labs.perplexity.ai/
- LMSYS Chatbot Arena: https://chat.lmsys.org/
- Meta AI: https://www.meta.ai
- Groq: https://groq.com/
Downloading the Model
You can download the LLaMA 3 model from:
- Hugging Face: https://huggingface.co/meta-llama
- Meta: https://llama.meta.com/llama-downloads/
Benchmarks
Evaluate the performance of LLaMA 3 using the following benchmarks:
- LMSYS Chatbot Arena leaderboard: https://chat.lmsys.org/?leaderboard
- Meta official benchmark numbers: https://ai.meta.com/blog/meta-llama-3/
- GitHub: https://github.com/meta-llama/llama3/blob/main/MODEL_CARD.md#benchmarks
Prompting the Chat Model
Engage in a multiturn conversation with LLaMA 3 using the following prompt template:
<|begin_of_text|><|start_header_id|>system<|end_header_id|>
{{ system_prompt }}<|eot_id|><|start_header_id|>user<|end_header_id|>
{{ user_message_1 }}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
{{ model_answer_1 }}<|eot_id|><|start_header_id|>user<|end_header_id|>
{{ user_message_2 }}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
For more information on prompting LLaMA 3, visit:
- https://github.com/meta-llama/llama-recipes
- https://llama.meta.com/docs/model-cards-and-prompt-formats/meta-llama-3/#special-tokens-used-with-meta-llama-3
How to finetune LLaMA 3
With LLaMA 3 being openly available, fine-tuning is a breeze. Plus, you'll find a wealth of resources at your disposal to train your own customized versions of LLaMA 3, giving you the flexibility to tailor it to your specific needs.
- https://huggingface.co/blog/mlabonne/orpo-llama-3
- https://llama.meta.com/docs/how-to-guides/fine-tuning/
- https://www.together.ai/blog/together-ai-partners-with-meta-to-release-meta-llama-3-for-inference-and-fine-tuning
- https://huggingface.co/blog/llama3
- https://predibase.com/
- https://github.com/huggingface/autotrain-advanced
Additional Resources
Stay up-to-date with the latest developments and learn more about LLaMA 3 with these additional resources:
Blog
- Meta blog: https://ai.meta.com/blog/
- Meta LLaMA 3 blog post: https://ai.meta.com/blog/meta-llama-3/
- LLaMA 3 website: https://llama.meta.com/llama3
- Use Policy: https://llama.meta.com/llama3/use-policy
Community resources
- https://medium.com/@AITutorMaster/fine-tuning-llama-3-with-unsloth-a-comprehensive-guide-b2e0feabe4a5
- https://twitter.com/svpino/status/1781665372619390984
- https://www.confident-ai.com/blog/the-ultimate-guide-to-fine-tune-llama-2-with-llm-evaluations
- https://www.linkedin.com/feed/update/urn:li:activity:7186854795163439105/
Informative tweets
- https://twitter.com/abhi1thakur/status/1781567706577219607
- https://twitter.com/ylecun/status/1780999054719127840
- https://twitter.com/Thom_Wolf/status/1781061969354649826
- https://twitter.com/lvwerra/status/1780998032420475169
- https://twitter.com/_lewtun/status/1781340847201460413
- https://twitter.com/karpathy/status/1781028605709234613
- https://twitter.com/karpathy/status/1781047292486914189
- https://twitter.com/_philschmid/status/1781613940067152369
A huge thank you to the open-source community and individual contributors who have created and shared these valuable resources. Your generosity and dedication to advancing AI and NLP are truly appreciated.
Thanks for joining me on this LLaMA 3 adventure! I will keep updating this list . If you're still curious about the possibilities, let's keep the conversation going on LinkedIn or X - and who knows, maybe we'll create something amazing together!
Love this post? Help spread the word about LLaMA 3 by sharing it with your network.
Comments