- Before you start, please get your
<YOUR-LAMINI-API-KEY>at https://app.lamini.ai/account. Create~/.lamini/configure.yamland write the file as follows, this will persist your Lamini credentials:
production:
key: <YOUR-LAMINI-API-KEY>- Install python requirements
pip3 install -r ./lamini-examples/requirements.txtIn this repo, we include tutorials for achieving high-quality results with Language Models (LLMs), like Llama3, using Lamini. With Lamini, you own the LLM you create -- you can deploy it or release it open source.
These examples show effective tools for building LLMs. We strongly encourage following the examples in order as the concepts build on each other and are sorted by difficulty.
- Llama3 - Generate text with Llama3, a powerful LLM.
- Evaluation - Evaluate the quality of your LLM.
- Prompt Tuning - Improve the quality of your LLM by tuning the prompts you use.
- RAG Tuning - Retrieval Augmented Generation (RAG), improve the quality of your LLM by tuning the retrieval component.
- Data Pipeline - Prepare your data for training an LLM.
- Memory Tuning - Memory tune an LLM on your data.
Extra Examples:
- Meta Llama Memory Tuning 🔗 - Embed facts into an LLM and improve factual accuracy on your data.
- JSON Output - Extract structured output from an LLM, following a guaranteed JSON schema.
- Slackbot - Learn how to create a Slack bot that calls an LLM.
The goal of this repo is to teach and provide examples of important tools for building LLMs; the examples emphasize simplicitly and readibility, not heavy optimization.
Once you have mastered a module from this repo, consider forking it and adapting it to your own application.
All of the code in this repository is licensed Apache 2. You are free to use it for any purpose including commercial applications.
The source code for this repo can be found on GitHub at lamini-ai/lamini-examples. Feel free to explore and contribute!
Lamini is the LLM platform for developers to specialize LLMs on their own data and infrastructure: easier, faster, and better than any LLM for their use case.
Our mission is to build customizable superintelligence that anyone can build and own.
