financy-PM-1

The Story: After successfully generating a few synthetic financial conversation datasets, I wanted to see if I would be able to successfully use them for the fine-tuning of my own open source model.

The Result: With the help of Gemini and Claude in the Cursor IDE, I was able to make a fine-tuning notebook (private) for this use case. I was able to successfully fine tune an open source Llama 3 8B model from meta, sourced from huggingface.com. In a notebook, I am able to see that the model has been fine-tuned to providing financial advice based on a dataset of 250 conversations. I would like to use larger datasets in the future.

As of (7/31/2025), I have encountered a bug on huggingface.com that is preventing me from properly setting up a demo for this model.

For More: The repo and hub for this project is currently private.