Providing context for fine-tuning LLMs and walking through why I chose to fine tune an instance of GPT-3.5-Turbo, my specific use case exploration, and all of the steps I took to get there!
The Code: github.com/ALucek/ft-openai-v...
Output Comparison: docs.google.com/spreadsheets/...
GoEmotions Dataset: huggingface.co/datasets/go_em...
Fine Tuning Graphic: / 1*jsjbbnslbe9s5i77rz9r...
OpenAI Fine Tuning Documentation: platform.openai.com/docs/guid...
OpenAI Cookbook - Fine Tuning Data Prep : cookbook.openai.com/examples/...
Chapters:
00:00 - Intro
00:21 - What is Fine Tuning?
01:40 - OpenAI Fine Tuning Documentation
03:51 - Why I Fine Tune Models
05:04 - Dataset of Interest
06:44 - Training Data Format
07:18 - Loading and Examining the Dataset
10:24 - Generating a GPT-4-T Baseline
13:00 - Formatting Dataset into JSONL
15:47 - Validating Training Data & Price Estimation
19:13 - Starting the Fine Tuning Job
19:47 - Fine Tuned Model Metrics Overview
20:31 - Talking About Training Loss
23:29 - Running Inference With the Fine Tuned Model
24:02 - Output Comparisons & Discussion
27:01 - Outro