Skip to content

LLM plugin for running models using Bedrock Meta Llama2 model

License

Notifications You must be signed in to change notification settings

flabat/llm-bedrock-meta

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

18 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

llm-bedrock-meta

PyPI Changelog Tests License

Plugin for LLM adding support for Meta LLama 2 and 3 models in Amazon Bedrock

Installation

Install this plugin in the same environment as LLM. From the current directory

llm install llm-bedrock-meta

Configuration

You will need to specify AWS Configuration with the normal boto3 and environment variables.

For example, to use the region us-west-2 and AWS credentials under the personal profile, set the environment variables

export AWS_DEFAULT_REGION=us-west-2
export AWS_PROFILE=personal

Usage

This plugin adds models called bedrock-llama2-13b, bedrock-llama2-70b, bedrock-llama3-8b-instruct, and bedrock-llama3-70b-instruct. You can also use it with the aliases like bl2 or bl2-70.

You can query them like this:

llm -m bl3-70-i "Ten great names for a new space station"
llm -m bl3-70-i "Ten great names for a new space station"

You can also chat with the model:

llm chat -m bl3-70-i

Options

  • -o max_gen_len 1024, default 2_048: The maximum number of tokens to generate before stopping.
  • -o verbose 1, default 0: Output more verbose logging.
  • -o temperature 0.8, default 0.6: Use a lower value to decrease randomness in the response.
  • top_p, default 0.9: Use a lower value to ignore less probable options. Set to 0 or 1.0 to disable.

Use like this:

llm -m bl3-70-i -o max_gen_len 20 "Sing me the alphabet"
 Ah, ah, ah! Here's the alphabet song for you:


A, B, C, D, E, F, G,
H, I, J, K, L, M, N, O,
P, Q, R, S, T, U, V, W,
X, Y, Z,
Now I know my ABCs,
Next time won't you sing with me?


Hope that brought a smile to your face!

About

LLM plugin for running models using Bedrock Meta Llama2 model

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages