Libraries for large language models: llm-lib

ChatGPT interactive use

Login at the following address

https://chat.openai.com/auth/login

Introduction

https://platform.openai.com/docs/introduction/overview

• Content generation
• Summarization
• Classification, categorization, and sentiment analysis
• Data extraction
• Translation

Run openai-quickstart-python

See openai-quickstart-python

Run openai-examples

1. Setup API key

source ./do_not_checkin/openai_key.sh

2. Run openai-example in a poetry virtual environment

make open-ai-example

Run openai-cookbook

1. Clone the project

git clone https://github.com/openai/openai-cookbook

2. Change to the project root

Datasets

Fine food reviews

There are two files

./openai-cookbook/examples/data/fine_food_reviews_1k.csv
./openai-cookbook/examples/data/fine_food_reviews_with_embeddings_1k.csv

The following notebooks use the above two data files.

• Obtain_dataset.ipynb - read the first file and create the second file with embeddings
• Visualizing_embeddings_in_2D.ipynb - visualize embeddings using tSNE
• Visualizing_embeddings_with_Atlas.ipynb - requires Nomic login https://atlas.nomic.ai/
• Clustering.ipynb - kMeans clustering and tSNE
• Visualizing_embeddings_in_W&B.ipynb - requires Weights and Biases
• Zero-shot_classification_with_embeddings.ipynb - uses cosine distance of embeddings to classify positive/negative
• Classification_using_embeddings.ipynb - uses RandomForest classifier with embeddings to learn ratings
• Regression_using_embeddings.ipynb - uses RandomForest regressor to learn ratings
• Semantic_text_search_using_embeddings.ipynb

Ollama

Ollama is used to run large language models locally.

Install Ollama using the command

curl -fsSL https://ollama.com/install.sh | sh

Llama index

https://github.com/run-llama/llama_index

Semantic kernel

https://github.com/microsoft/semantic-kernel

Jupyterlab in a different virtual environment

https://samedwardes.com/2022/10/23/best-jupyter-lab-install/

1. Install jupyterlab using pipx

pipx install jupyterlab --include-deps

2. Check pipx binary directories are added to PATH

pipx ensurepath

3. Check for the jupyterlab environment

pipx list --short

4. List available kernels

jupyter kernelspec list

5. Switch to the virtual environment

poetry shell

6. Install ipykernel in the environment

poetry add ipykernel

7. Register the kernel

python -m ipykernel install --user --display-name ${PWD} --name ${PWD##*/}

8. List available kernels. A new one is added

jupyter kernelspec list

Links

Prompt engineering guide

Chat GPT login

https://github.com/langchain-ai/langchain

Phidata is a Python library that helps build AI assistants with memory, knowledge and tools.

Building applications with Large Language Models (LLMs)

[Gorilla][1030] is a Python library that connects large language models (LLMs) to application programming interfaces (APIs).

Getting started with LLMs using LangChain

https://www.pinecone.io/learn/langchain-intro/

Videos

https://www.youtube.com/@DataIndependent/videos

https://www.youtube.com/@echohive/videos

https://www.youtube.com/watch?v=aywZrzNaKjs

Open-source RAG tool to chat with documents

https://github.com/Cinnamon/kotaemon

Other

How to build a semantic search

Question answering chatbot

Byte pair encoding

Github libraries

https://github.com/hwchase17/langchain - 61k stars

https://github.com/nomic-ai/gpt4all - 52k stars

https://github.com/microsoft/guidance - 13k stars

https://github.com/stanfordnlp/dspy - 2.4k stars

https://github.com/jerryjliu/llama_index - 21k stars

https://github.com/langflow-ai/langflow - 53.9k stars

Comparison of libraries

https://medium.com/badal-io/exploring-langchain-and-llamaindex-to-achieve-standardization-and-interoperability-in-large-2b5f3fabc360