Usage

To generate datasets using this library, follow these steps:

1. Set up your environment: Ensure you have Python installed on your system. This library requires Python 3.6 or later.

2. Install the dependencies:

 pip install -r requirements.txt

3. Prepare your API key: The library uses OpenAI's GPT models for generating datasets. You need to have an OpenAI API key. Set your API key as an environment variable:

   export OPENAI_API_KEY='your_api_key_here'
   

4. Generate datasets: Use the create_dataset.py script to generate datasets. This script supports generating datasets from PDF documents or directly from SEC filings using a ticker symbol and year.

   Available Variables:

   • --model_name: The model name to use for generating the dataset. Default is 'gpt-3.5-turbo'.
   • --pdf_path: The path to the PDF document you want to generate the dataset from.
   • --ticker: The stock ticker symbol for generating a dataset from a 10-K filing.
   • --year: The year of the 10-K filing to generate from.
   • --max_questions: The maximum number of questions to generate from the document or filing.

   Example command to generate a dataset from a PDF document:

   python create_dataset.py --model_name 'gpt-3.5-turbo' --pdf_path '/path/to/your/document.pdf' --max_questions 100
   

   Example command to generate a dataset from a 10-K filing:

   python create_dataset.py --model_name 'gpt-3.5-turbo' --ticker 'AAPL' --year 2021 --max_questions 100
   

   The script will generate a dataset and save it as a JSON file in the data directory.

Note: Replace 'gpt-3.5-turbo' with the model of your choice supported by OpenAI. Adjust the --max_questions parameter based on how many questions you want to generate from the document or filing.

Based on: