JSONL is a multi-line data in JSON format. Most LLM use this format for data fine-tuning.
Gasby provides the JSON Generator tool to convert your PDF, DOCX or Text to JSONL format for data fine-tuning (by splitting file content into multiple chunk of data). At the moment, you can only generate JSONL format for GPT models
To convert your content file:
- Drag and drop or click "Upload" and choose your content file
- (optional) Set
System message by click on the Edit icon - (optional) Set
User message by click on the Edit icon - Click
Download to convert and download your JSONL data file
It'll start download a ft-data.jsonl file, which can be used for fine-tuning
Note that adding System, and User messages is optional. Though it will give more information for the model, and might help produce better result in some cases.