llmkit

Repo for all work using large language models (LLMs) for applications

List of open LLMs

https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard

Pre-reqs(poetry installation): Optional

Install the poetry

NOTE : if you don't want to use poetry you can manually install packages listed in pyproject.toml

Installation

1. clone and cd into the repo

2. Run

   poetry install
   

   to install packages

3. Get your own OpenAPI key

4. Run

   poetry shell
   

   to to activate the environment

5. Run

   poetry run jupyter notebook [--no-browser]
   

   to open juypter notebook

6. Use any of the following examples from the example folder

Check usage of OpenAPI credits

OpenAPI credits Usage

Machine Manual testing with llama2-13b-chat-german

check the examples folder for different tests performed on machine manual using llama2 machine_manual

Streamlit + OpenAI agent retrieval (English)

https://colab.research.google.com/drive/12NgFnqVsit1rtb9lyNCASaMZbSg4FR_E#scrollTo=qeMBGrAYy0OQ

Falcon40b-Instruct - Sustainability Report (English)

https://colab.research.google.com/drive/1hQZpJHK4w1N_lNNZoPqLgFZcX37whbac#scrollTo=SBNL9U6y3mx9

Falcon40b (Raw, not tuned for chat) - Multilingual(German) - Injection Moulding Manual

https://colab.research.google.com/drive/177CVK3nuHTP8mwEDe0oy50UxUKSSpRRv#scrollTo=8ow4ttMszJpo

Inference time for above models is high (in some minutes) even on A100 GPU with 40GB vRAM (need to research about quantization of models and its effects on model inference speed)

WizardLM - No OpenAI (under construction)

https://colab.research.google.com/drive/1JwzfVRZMEUSRhZZldDZrz5s-MyjkGn-i?usp=sharing

WizardLM with chat history and custom prompts

https://colab.research.google.com/drive/12o313CJW69mKnGBUQL5KtrdP2Rty5LoQ?usp=sharing

WizardLM - Injection Moulding Manual (Don't use it yet. too expensive and still under progress)

https://colab.research.google.com/drive/1bARZZnxF8lYBxtPGYFtMD7Rfc9-ckbzz?usp=sharing

LangChain + GPT4ALL

https://colab.research.google.com/drive/1Je6znV4sbr-ys4IRQFU43yMoCXzVkyM2#scrollTo=S6n0Ml7IvjC5

LangChain + OpenAI

https://colab.research.google.com/drive/1k7BM_bnDXqOMOx1ZFDhabWZN2qW-2Wjw#scrollTo=iJZIckCVPtg8

LangChain Summerization

https://colab.research.google.com/drive/1-cFy1gZBJ6AehZ77WZcGyBI_TOz5L9y1#scrollTo=qBwYQOIHgp17

LangChain Agent based interactions (not working yet)

https://colab.research.google.com/drive/1XrapNN72dgkUG2lyT7Ky7swYYaoCmNZ_#scrollTo=7qe3yyVSz9jt

LangChain + Aleph Alpha

https://colab.research.google.com/drive/1bixmqgFXArp86hOqDa5o2cWBvPFEk3GU?usp=sharing