Fine-Tune Your Chat-GPT: Train 3.5 on YOUR Docs
The latest version of chatGPT is out and they amped up the release to allow for robust fine-tuning of your own documents on version 3.5. This will enable you (through the API) to take a set of documents and train the system automatically on that data.
To date, when you use the chat interface of chatGPT it can only remember so far into the past. It's kind of like a fine sourdough starter but eventually everything is new again. Now, you can create models based on your data that will be persistent and then call on those models just like you would in a chatGPT interface. Here, I show you how to do that step-by-step.
For example, I write fundraising text messages for political and nonprofit groups.
I take the best-performing texts and create a data set from them.
I would train the model on that data set.
I can run requests against it.
This is basically what I do on ChatGPT in the basic interface but now it will have a pre-trained model so I don't have to keep refreshing it when it forgets the data.
First step is to prep your data. This is perhaps hardest step as it entails creating robust chat conversations as examples in a JSON format. Luckily for you, I asked ChatGPT to create for me an Excel file that will handle the creation of the JSON pretty well. Or at least get me most of the way there.
Here’s the basic conversation I had:
Let’s go!
Keep reading with a 7-day free trial
Subscribe to Hello, AI to keep reading this post and get 7 days of free access to the full post archives.