Amazon Bedrock Tutorial: Generative AI on AWS
HTML-код
- Опубликовано: 25 июн 2024
- Learn how to use Amazon Bedrock for generative AI (text and images) on AWS.
This tutorial will guide you through the setup process, showing you how to generate text and images via code. You'll learn how to generate text with the Anthropic Claude model, and images with the Stable Diffusion model. All you need is an AWS account and the AWS CLI installed and configured.
🔗 Code: github.com/pixegami/aws-bedro...
🔗 Get Started on AWS: • How to Get Started wit...
🔗 Amazon Bedrock: aws.amazon.com/bedrock/
📚 Chapters
00:00 Introduction to Amazon Bedrock
01:12 Setting Up Amazon Bedrock
02:54 Text Playground
03:41 Image Playground
04:47 Generating Text with Code
07:43 Generating Images with Code
09:40 Amazon Bedrock Pricing
12:13 Amazon Bedrock vs OpenAI
#pixegami #aws
![Text Summarization with AWS Amazon Bedrock - Cohere FM, AWS Lambda, API Gateway [HANDS-ON]](/img/1.gif)
This is a great tutorial! I have already set up my AWS account and CLI. I run your code and it works for the first time. Thank you so much!
Awesome! Glad to hear it went well :)
Great tutorial!
Thank you!
Just discovered the channel. Great content and production quality ❤ subd
Thank you! Welcome aboard!
Thanks, easy to follow. Just FYI Amazon offers an image model now: Titan Image Generator G1
Thank you :) Yes I just saw that! And Stable Diffusion is upgraded to a newer version too. AI moves too fast!
Will AWS charge for trying out things in playground
Can you do a tutorial on comparing 2 or more documents via file upload?
Yup! I'm working on that and will publish a video soon (using Langchain, RAG and OpenAI).
Thanks for the awesome video.. i was wondering if you know if i can import other models and lora's and i plan to use the python output by comfyui. I've had trouble finding this info for bedrock. Any insights? Thanks again. To me, it looks like just the plain vanilla foundation models... I'm hoping you maybe tried an import from huggingface on your own.
I haven't really gone deep into using Stable Diffusion and ComfyUI before, but just looking at the website, I think it's a different use-case to what you'd wanna use AWS Bedrock for.
I think that use-case is if you have your own models (like SDXL) and are self-hosting it and want to set up a more complex image generation pipeline with more control (from what I read on github.com/comfyanonymous/ComfyUI).
But the use case for AWS Bedrock is more where you don't mind having less control on the image generation itself, but just want to access via an API and have it easily integrated with a web-app or other service.
@@pixegami Thanks for the response... I did a bit more research and I totally agree with you. Baseten has a pretty thorough blog and video showing what I'm after, I'm going to give that a try. Merry christmas!
excuse my naivete but why would I write all that long code if I can literally just type in the image prompt what I want instead and get the same result?
The web UI is great for one-off generations. But the “long code” is a programmatic way to control your use of GenAI. You only need to type it once, but it gives you the power to automate future generations.
For example you might want it to generate a image automatically once per day for your app. Or you want it to generate 1000 different character designs at once, based on a set of descriptions. Doing those manually is going to cost more in the long run than a well written automation.
@@pixegami thanks! Great response and great content
How can we do a AI search within a data source internally instead of Global? Otherwords invokemodel should fetch the results from the data source that I point.
The easiest most common way to do this is to use a technique called "RAG" (Retrieval Augmented Generation). This is what OpenAI uses as well when you build a custom GPT.
Here's some info on how to get started: python.langchain.com/docs/expression_language/cookbook/retrieval
Here's a video tutorial I recorded about it (it uses OpenAI instead of AWS though, but you can change that): ruclips.net/video/tcqEUSNCn8I/видео.html
Thanks for tutorial! I’m really confused I cannot enable any model. I’m root user and I chose us region. What can be wrong?
What error do you see when you try to enable them? If you are in us-east-1 and are logged in as an Admin, it should just work :(
Hey I think it was some transient. All is good now and I get super cool material :) thanks for your tutorials !!
When trying to enable the models with root user I get this error message:
Access request for 1 models failed
• SDXL 1.0 - Operation not allowed
Hmm, are you attempting this with an Admin user, and also in one of the supported regions (us-east-1, us-west-2)?
@@pixegami yes, as root user and us-east-1. I created my aws account many years ago. So my workaround was to create another free-tier account.
My request failing for accessing models