Creating a JSONL Dataset
Based on what we've learned so far, let's carry out a practical exercise to create and implement a dataset.
Introduction to the Practice Environment
We will now create a training dataset to fine-tune ChatGPT as a sassy informal AI.
As introduced in the [previous lesson], a JSONL dataset for fine-tuning must include at least 10 pairs of conversations.
Creating an Informal Dataset
-
Modal Window: Click the
Select Databutton to open a modal window. (Modal Window: This lists the dataset files.) -
Create New File: Click the
Create New Filebutton at the top. -
Mode Selection:
Simple Modeis a basic mode for creating a dataset.Editorallows you to directly write a JSONL file. -
Write Dataset: Enter the file name and click the
Createbutton to create the file. -
Create File: Enter the file name and click the
Createbutton to create the file. -
Apply: Click the
Applybutton to select the created file.
Lessons in this chapter · Preparing Learning Data
- 1. Data Formats Used for Training AI
- 2. Deep Dive into JSON Data Format
- 3. Fuel for Fine-Tuning (JSONL)
- 4. JSONL Format for Fine-Tuning
- 5. Preparing to Train AI (Dataset Essentials)
- 6. Types of Datasets Used in AI Training
- 7. How to Split a Dataset
- 8. Preparing Data for Consumption (Preprocessing)
- 9. What Data Formats Do Different AI Models Use?
- 10. Creating a JSONL Dataset
- 11. Comparing Responses of General vs Fine-tuned Models
- 12. Multiple-choice quiz
What is the most appropriate word to fill in the blank?
Lecture
AI Tutor
Design
Upload
Notes
Favorites
Help