- Run `create_jsonl.py` to create jsonl files for dataset.txt
- dataset.txt is the training-pending dataset
- I regularly run a new fine-tuning job on the dataset.txt. After the job is done, I will move the dataset to `datasets/` and remove it from `dataset.txt`.