This is an exploratory project into machine learning by initially training a GPT-2 model (gpt2-medium) on the Natural Questions (NQ) dataset to grasp question-answering mechanisms. Post this foundational training, the model underwent fine-tuning to pivot its capabilities towards generating job title prompts using prompt examples created by 'text-davinci-002' and further augmented with source and target noise. Used PyTorch and Hugging Face's Transformers api. The model is published at on Hugging Face: sgarbi/gpt-nq-prompt-generator.
This project is shared under the MIT License. For more information, refer to the LICENSE file. While open-source, this project is developed and maintained exclusively by its creator, serving as a personal history exploring of NER technologies.