PyTorch’s Post

AWS, Arm, Meta, and others helped optimize the performance of PyTorch 2.0 inference for Arm-based processors. As a result, Arm-based AWS Graviton instance inference performance for PyTorch 2.0 is up to 3.5x the speed for ResNet-50 compared to the previous PyTorch release, and up to 1.4x the speed for BERT. Find out how to take advantage of these optimizations and run an inference: https://bit.ly/44Qf7kt

Did you miss re:Invent? "What's New in Amazon OpenSearch Service" breakout is now available to stream. Mukul Karnik and myself give an overview of the major advancements on OpenSearch and Amazon OpenSearch Service in 2023 as well as the new re:Invent announcements like OpenSearch zero-ETL with S3, DynamoDB zero-ETL with OpenSearch, OpenSearch multi-modal neural search with Amazon Bedrock, the open source of OpenSearch generative-AI assistant, the new OpenSearch optimized OR1 instance family, vector engine on Amazon OpenSearch Serverless and much more. Special thanks to Bill Burket from Electronic Arts for sharing their journey to Amazon OpenSearch Service. Mukul Karnik OpenSearch Project #OpenSearch Amazon Web Services (AWS) #awsreinvent2023 #amazonbedrock

In this post, we share how the Amazon Search M5 team optimize their use of accelerated compute resources using AWS Batch fair-share scheduling to schedule distributed deep learning workloads, while saving money. https://buff.ly/3Epqepd

Deep dive: PyTorch 2.0 on AWS Graviton. Watch this AWS Online TechTalk video which will talk through the price performance benefits of your PyTorch 2.0 workloads on AWS. AWS Graviton based instance inference performance for PyTorch 2.0 is up to 3.5x faster than the previous PyTorch release.

Exciting news from AWS! They have recently introduced Guardrails for Amazon Bedrock to enhance the safety of Large Language Models (LLMs). We are all aware of the remarkable benefits that come with utilizing LLMs, but it is crucial to address the concerns surrounding them. By implementing Guardrails, AWS offers an effective solution to limit the responses generated by these models, providing a level of control over their immense power. Discover more about this significant development in the blog post below. https://ift.tt/pGujK5a

AWS Bedrock. It wasn’t just a service; it was a symphony of generative magic, orchestrated by quantum minds and powered by neural stardust. #aws #bedrock #cloudcomputing

This week, TNS VoxPop is asking, "What news from AWS re:Invent will have the most impact on you?" Chime in here: https://lnkd.in/gbzYT2me #awsreinvent2023 #AWSreInvent #reInvent

It was a crazy lesson today, I learned how to set up Auto Scaling Groups in AWS. It's fascinating how you can use Machine Learning to predict your future traffic and have it scale your instances automatically. #1dayatatime

Looking to improve the price performance of your PyTorch 2.0 workloads? AWS Graviton-based instance inference performance for PyTorch 2.0 is up 3.5 times than the previous PyTorch release 🚀 Check out this free webinar from my colleague Hahnara Hyun to learn more → https://lnkd.in/e6cBAQnz #amazon #aws #ec2 #graviton #cloud #developer #aiml

#reinvent2023 by Amazon Web Services (AWS) just wrapped last week. As Adam Selipsky said re:Invent is a learning event and there were a lot of talks: 5 keynotes, 17 Innovation talks by AWS leaders and 2400+ breakout sessions by customers, partners and AWS employees. It's a LOT of information! Here (https://lnkd.in/eB_6fAyf) is a link to the videos of the keynotes and innovation talks. The innovation talks are pretty amazing as senior leaders share their insights, roadmap on a range of topics from infrastructure to ops to data to AI to developer experience and security. One tip in case you are overwhelmed by 30+ hours of content: -- Use a GenAI tool to summarize a Youtube transcript and get the big picture before diving in at 2x speed! I use it with Claude2 https://lnkd.in/e9hG2P36 Happy watching and share your best talks.