Llama has been arguably the most powerful open source large model in the AI community. However, due to the open source agreement, it has not been free for commercial use. On July 19th, Meta released the long-awaited free commercial version Llama 2. Overnight, the pattern of large-scale models changed drastically again.
As one of the first partners announced by Meta, Amazon cloud technology customers can now use the Llama 2 base model developed by Meta through Amazon SageMaker JumpStart!
The Llama 2 Large Language Model (LLM) family is a pre-trained and fine-tuned generative text model with parameters ranging from 7 billion to 70 billion. Fine-tuned large language models called Llama-2-chat, which are optimized for conversational use cases, are now supported in SageMaker JumpStart to try and use these models; Access to learning solutions so you can quickly start using machine learning.
Call 2 minutes
Llama 2 is an autoregressive language model using an optimized Transformer architecture. Llama 2 is intended for commercial and research use in the English domain. The model provides a range of parameter sizes, 7 billion, 13 billion, and 70 billion, as well as pre-trained and fine-tuned variants.
The fine-tuned version adopts supervised fine-tuning (FT) and reinforcement learning based on human feedback (RLHF), which is consistent with human preferences, thus ensuring practicality and safety, Meta said. Llama 2 was pretrained using 2 terabytes of data from publicly available sources. Fine-tuned models are used for assisted chat, while pre-trained models can be applied to various natural language generation tasks. Regardless of which version of the model a developer uses, Meta's responsible usage guidelines can help guide any additional fine-tuning that may be required to customize and optimize the model with appropriate security mitigations.
Introduction to Amazon SageMaker JumpStart
Amazon Cloud Technology is committed to continuously lowering the threshold for using machine learning. Up to now, Amazon cloud technology has helped more than 100,000 customers of different sizes from all walks of life to use machine learning to innovate.
To this end, Amazon Cloud Technology has launched Amazon SageMaker, a one-stop development platform for machine learning, which provides maximum convenience for all developers to build, train and deploy basic models, and provides more customers with the ability to access basic models. As one of the core components of Amazon SageMaker, SageMaker JumpStart has built-in 300+ open source models and 10+ preset scene solutions, which can help users "one-click" deployment and fine-tuning such as Stable Diffusion, Bloom, FLAN-T5, Alexa TM And other mainstream generative AI basic models, users can use SageMaker JumpStart to deploy or fine-tune many pre-trained models with one click, easily develop high-quality models and shorten deployment time.
Now, with just a few clicks in Amazon SageMaker Studio, or programmatically discover and deploy Llama 2 via the SageMaker Python SDK, you can experience model performance and MLOps control with SageMaker features such as Amazon SageMaker Pipelines, Amazon SageMaker Debugger, or container logs. The model is deployed in the secure environment of Amazon cloud technology and controlled by VPC, which helps to ensure data security. Currently, Amazon SageMaker Studio supports the Llama 2 model (us-east 1 and us-west 2 regions only).
epilogue
In April 2023, Amazon Cloud Technology launched the generative AI "family bucket" represented by Amazon Bedrock, officially entering the global generative AI competition battlefield, aiming to work with industry-leading partners to continuously reduce the cost of generative AI applications. The development threshold helps enterprises unleash the productivity and creativity of generative AI.
Starting from the end, standing at the turning point of the development of generative AI, as a leader in the field of cloud computing, Amazon cloud technology will continue to help every enterprise, everyone has the opportunity to obtain, use and enjoy the new generative AI technology brought benefits.