Language Model Analyst

The Language Model Analyst is a Python package and Streamlit app that enables natural language generation and analysis using HuggingFace-based language models (LLMs) and OpenAI GPT-3 with LangChain. This package and app are designed to simplify and streamline interactions with these powerful language models.

Usage

Package: LLMAnalyst

The LLMAnalyst class allows you to interact with HuggingFace-based language models. Follow these steps to use the package:

1. Install the package:

   pip install git+https://github.com/eersnington/LLMAnalyst.git
   
   # For using GPTQ models
   pip install auto-gptq --extra-index-url https://huggingface.github.io/autogptq-index/whl/cu118/  

2. Import and create an instance of LLMAnalyst:

   from LLMAnalyst import LLMAnalyst
   llm_analyst = HuggingfaceAnalyst("TheBloke/CodeLlama-13B-Instruct-GPTQ")

3. query = "How many number of rows are there?"
   df = pd.read_csv("data.csv")
   result = llm_analyst.conversational_chat(query, df)

The OpenAIGPTAnalyst class enables interaction with the OpenAI GPT-3 model. Follow these steps to use the package:

1. Import and create an instance of OpenAIGPTAnalyst:

   from LLMAnalyst import OpenAIGPTAnalyst
   
   openai_analyst = OpenAIGPTAnalyst(
       api_key='YOUR_OPENAI_API_KEY'
   )

2. Communicate with GPT-3:

   query = "Calculate the mean of monthly sold data."
   df = pd.read_csv("data.csv")
   result = openai_analyst.conversational_chat(query, your_dataframe)