A working AWS account with credentials, and Bedrock enabled. If you don't have an AWS account you can create one here:
https://signin.aws.amazon.com/signup?request_type=register
The rest of this page deals with creating credentials and enabling Bedrock.
You would expect to be given a compute credit at the workshop which should cover the costs of everything you would do at the workshop. Be sure to undeploy when you are done to avoid racking up extra costs.
This process uses an LLM hosted on AWS Bedrock. These models need to be switched on for use in the Bedrock part of the AWS console.
We'll be using the us-west-2 region. Different regions offer different models, and us-west-2 is less busy than other regions and has a full set of models available.
- Log into the AWS console
- Select us-west-2 region (Oregon). The region selector is at the top, near the right-hand side of the screen.
- Search for Bedrock in the resource search bar, at the top of the screen, left hand side, select "Amazon Bedrock".
- On the Bedrock screen look for Model Access on the LHS.
- Select Modify Model Access
- If offered 'Enable All Models' use that.
- Otherwise, search for Anthropic and enable all the models you see.
- Continue through the dialogues until you Submit. You may have to submit some information to get model access.
- The model should change from requested to available, takes a couple of minutes. Maybe go and set some other stuff up, but leave this page open so you can refresh and see that it got enabled.
This bit needs your concentration, as some credentials are going to be created which provide access to your AWS account. If you accidentally send these credentials to the wrong person or check them into Github and publish them on the internet, it gives other people access to your account, so treat these as passwords.
If you're using an AWS account you already owned, and have a profile setup, you can skip this bit and just use the profile you have set up. Just make sure to set AWS_PROFILE to the name of your AWS profile before launching the deployment.
Here's a video of someone accidentally live-streaming their crypto seed phrase and losing $100k: https://www.youtube.com/watch?v=4RALeCRTw6Y
Do yourself a favor, and set yourself a reminder to delete the IAM User once the workshop is done.
This is done by searching for IAM in the resource search box.
Select "Users" and "Create User". Give the user a name. The user does not need access to the AWS console.
Select "Attach policies directly" and give the user the AdministratorAccess policy. This lets the user deploy to AWS.
You are shown a summary; go ahead and create the user.
At the IAM user console, find the user you just created and select it.
You can see an access keys panel which will be empty; select Create Access Key.
A set of use cases is shown, select Other.
Put something useful in the description tag if you want, and create the access key.
Finally, the access credentials are available on this screen. There are two parts, the access key and secret access key, which is not displayed on this screen.
If you haven't used AWS credentials before, you're going to need to set up the credentials file:
mkdir ~/.aws
touch ~/.aws/credentials
Set up a profile for the workshop which looks like this. You do this
by opening a text editor and adding this text to the ~/.aws/credentials
file.
[workshop]
aws_access_key_id = [access key goes here]
aws_secret_access_key = [secret access key goes here]
region = us-west-2
You can put that at the end of the file. The bit of text [workshop]
gives this a profile name of workshop, we'll set this later.
Next: Prepare to deploy