Nutlope / finetuning

Python scripts to finetune GPT 3.5

GPT-3.5 Fine tuning script

This repo contains script to clean up data and These are two files you can run to do finetuning.

Steps

1. Get your data into a CSV, ideally two columns. One will contain the prompt and one will contain the expected response.
2. Specify your system prompt in the SYSTEM_PROMPT variable in preparing-data.py.
3. Run preparing-data.py to transform data into JSONL format with a system prompt of your choice.
4. Make sure you've exported the env variable OPENAI_API_KEY with your API key.
5. Specify the name of the model then run finetuning-data.csv to actually trigger a fine-tuning job through OpenAI.

To add

• Validation script and step