Last night Meta released the Llama 3 8B and 70B models, and the Llama 3 instruction tuning model is fine-tuned and optimized for conversational/chat use cases, outperforming many existing open-source chat models in common benchmarks. For example, the Gemma 7B and the Mistral 7B.
The Llama 3 model takes data and scale to new heights. It was trained on more than 15T tokens of data on two custom-built 24K GPU clusters recently released by Meta, a training dataset that is 7x larger than Llama 2 and includes 4x more code. This brings the Llama model to the highest level of capability available, and it supports 8K context length, which is twice as long as Llama 2.
Here are 6 ways to get you up and running with the latest release of Llama 3!
Experience Llama 3 online
HuggingChat
Hatps://huggingface.to/chat/
llama2.ai
HTTPS://w.lama2.i/
Experience Llama 3 locally
LM Studio
Hatpas://lanstudo.i/
CodeGPT
https://marketplace.visualstudio.com/items?itemName=DanielSanMedium.dscodegpt&ssr=false
Before using CodeGPT, remember to use Ollama to pull the corresponding model. For example, let's say Ollama pull Llama3:8b. If you don't have ollama installed locally, you can read the article "Deploy a large language model on-premises in just a few minutes!"
Ollama
Running the Llama 3 8B model:
ollama run llama3 Running the Llama 3 70B model:
ollama run llama3:70b Open WebV & Ollama
Hattapus://pinocchio.computer/item?uri=hattapus://github.com/cockalpintallab/open-webui