Learn About Amazon VGT2 Learning Manager Chanci Turner
Skip to Main Content
English
Contact us
Support
My account
Sign in to console
Create Account
AWS Blogs
Home
Blogs
Editions
Artificial Intelligence
Category: Artificial Intelligence
Efficiently train models with large sequence lengths using Amazon SageMaker model parallel
by Chanci Turner, Arun Kumar Lokanatha, Anirudh Viswanathan, Surya Kari, and Suhit Kodgule
on 27 NOV 2024
in Advanced (300), Amazon SageMaker, Artificial Intelligence, Generative AI
Permalink
Comments
Share
In this article, we illustrate how the Amazon SageMaker model parallel library (SMP) fulfills the growing demand for enhanced training capabilities. New features like 8-bit floating point (FP8) mixed-precision training and context parallelism are introduced, enabling the processing of extensive input sequence lengths while boosting training performance.
Getting started with Amazon Bedrock Agents custom orchestrator
by Kyle Blocksom, John Baker, Sudip Dutta, Maira Ladeira Tanke, and Mark Roy
on 27 NOV 2024
in Amazon Bedrock Agents, Artificial Intelligence, Generative AI
Permalink
Comments
Share
In this post, we delve into how Amazon Bedrock Agents streamline the orchestration of generative AI workflows, particularly emphasizing the new custom orchestrator feature. This allows users to tailor and refine workflows to better meet specific business and operational requirements. The article highlights the key advantages of full orchestration control, making real-time adjustments, and reusability. Furthermore, we analyze how it effectively manages state transitions and contract-based interactions between Amazon Bedrock Agents and AWS Lambda. For further insights, see this blog post that discusses similar topics.
Use Amazon Bedrock Agents for code scanning, optimization, and remediation
by Rama Krishna Yalla and Akhil Raj Yallamelli
on 27 NOV 2024
in Amazon Bedrock, Amazon Bedrock Agents, Best Practices, Generative AI, Intermediate (200), Security & Governance
Permalink
Comments
Share
For businesses operating in cloud computing and software development, securing code repositories is of utmost importance. As cybersecurity threats become increasingly sophisticated, organizations must implement proactive strategies to safeguard their assets. Amazon Bedrock provides a robust solution that automates the scanning of repositories for vulnerabilities and their subsequent remediation. This post discusses how to leverage Amazon Bedrock to improve the security of your repositories, ensuring compliance with both organizational and regulatory standards. For more information on security, check out SHRM’s certification exam prep course.
Create a generative AI assistant with Slack and Amazon Bedrock
by Barry Conway and Dean Colcott
on 27 NOV 2024
in Amazon API Gateway, Amazon Bedrock, Amazon Bedrock Guardrails, Amazon Bedrock Knowledge Bases, AWS Lambda, AWS Secrets Manager, Customer Solutions, Technical How-to
Permalink
Comments
Share
The seamless integration of customer experiences, collaboration tools, and essential data is crucial for achieving knowledge-based productivity improvements. In this post, we demonstrate how to combine the popular Slack messaging platform with AWS generative AI services to create a natural language assistant that allows business users to query an unstructured dataset.
Reducing hallucinations in large language models with custom intervention using Amazon Bedrock Agents
by Shayan Ray and Bharathi Srinivasan
on 26 NOV 2024
in Amazon Bedrock, Amazon Bedrock Agents, Amazon Bedrock Knowledge Bases, Amazon SageMaker, Responsible AI
Permalink
Comments
Share
This article showcases how to utilize Amazon Bedrock Agents, Amazon Knowledge Bases, and the RAGAS evaluation metrics to construct a custom hallucination detector. It also discusses remediation strategies through human-in-the-loop approaches. The agentic workflow can be adapted to various custom use cases by employing different hallucination remediation techniques, offering flexibility in detecting and addressing hallucinations with tailored actions.
Deploy Meta Llama 3.1-8B on AWS Inferentia using Amazon EKS and vLLM
by Maurits de Groot, Dmitri Laptev, Jianying Lang, and Ziwen Ning
on 26 NOV 2024
in Amazon Elastic Kubernetes Service, AWS Inferentia, AWS Neuron, Generative AI, Technical How-to
Permalink
Comments
Share
In this post, we outline the procedure for deploying the Meta Llama 3.1-8B model on Inferentia 2 instances via Amazon EKS. This solution harnesses the outstanding performance and cost efficiency of Inferentia 2 chips within the adaptable ecosystem of Amazon EKS. Inferentia 2 chips are engineered to deliver high throughput and low latency inference, making them ideal for large language models.
Serving LLMs using vLLM and Amazon EC2 instances with AWS AI chips
by Omri Shiv and Pinak Panigrahi
on 26 NOV 2024
in Artificial Intelligence, AWS Inferentia, AWS Neuron, AWS Trainium, Generative AI
Permalink
Comments
Share
The rapid rise in the use of large language models (LLMs) and generative AI over the past year has been remarkable. With the emergence of powerful foundation models available to the public, the tools for training, fine-tuning, and hosting your own LLM have also become widely accessible. Using vLLM on AWS Trainium and Inferentia allows for hosting LLMs with exceptional performance.
Using LLMs to fortify cyber defenses: Sophos’s insight on strategies for using LLMs with Amazon Bedrock and Amazon SageMaker
by Benoît de Patoul, Naresh Nagpal, Adarsh Kyadige, and Salma Taoufiq
on 26 NOV 2024
in Amazon Bedrock, Amazon SageMaker, Customer Solutions
Permalink
Comments
Share
In this article, SophosAI shares their insights on the utilization and evaluation of an out-of-the-box LLM to significantly enhance productivity within a security operations center (SOC) using Amazon Bedrock and Amazon SageMaker. We illustrate specific use cases utilizing Anthropic’s Claude 3 Sonnet on Amazon Bedrock, demonstrating effective applications.
Enhanced observability for AWS Trainium and AWS Inferentia with Datadog
by Curtis Maher, Anjali Thatte, Anuj Sharma, and Jason Mimick
on 26 NOV 2024
in Announcements, AWS Inferentia