Skip to content

datakami/pydata-llama-index-tutorial

Repository files navigation

PyData Amsterdam 2023 workshop materials

For workshop participants

You have three options for running the software for this workshop. They are listed from easiest to hardest:

Quick Start (using colab)

You can open the exercises in Google's Colab with the button below:

Open in Colab

Quick Start (using docker)

This assumes you have Docker and git installed. The image takes about 1.7GB of disk space.

$ git clone https://github.com/datakami/pydata-llama-index-tutorial.git
$ cd ./pydata-llama-index-tutorial
$ docker run -it --rm -v $PWD:/repo -p 127.0.0.1:8888:8888 ghcr.io/datakami/llamaindex-workshop:latest

In the output, find the line saying "Jupyter Server 2.7.3 is running at:", and click the link starting with http://127.0.0.1:8888. Jupyterlab should launch in your browser. Open the Exercises-1.ipynb file to get started.

Quick Start (no docker)

This assumes you have git and Python 3.8 - 3.11 installed.

$ git clone https://github.com/datakami/pydata-llama-index-tutorial.git
$ cd ./pydata-llama-index-tutorial
$ python -m venv ./venv
$ source ./venv/bin/activate
$ pip install -r requirements.txt
$ python check_installation.py
$ jupyter-notebook

Jupyterlab should launch in your browser. Open the Exercises-1.ipynb file to get started.

Explanation: Repository structure

There are a few different files:

  • Exercises-*.ipynb: These files contain exercises for workshop participants. Participants are meant to use these.
  • Colab-*.ipynb: merged notebooks that can be used to run this workshop in colab.
  • Solutions-*.ipynb: also contain answers. Used to generate the exercises notebooks.
  • check_installation.py: can be used to check that all dependencies are set up correctly and download the necessary models.
  • data/, indices/: workshop data and vector indices, used in the exercises.
  • Build-*.ipynb: notebooks to prepare the workshop data and vector indices.
  • build/: code to generate the exercises and colab notebooks.

For contributors

When committing

$ nb-clean clean -o [yournotebook.ipynb]

Building Exercises from Solutions

Solution filenames should start with "Solutions-". Code cells that start with a "#" are converted to questions with empty answers.

$ python -m build

About

PyData Amsterdam 2023 llama_index tutorial materials

Resources

Stars

Watchers

Forks

Releases

No releases published

Languages