Create a Dataset

You need to have dataset to train LLM. If you don't have one, you can create one in dataset generator.

In dataset generator, you can describe what dataset you want. You can also upload documents as a reference.

Then, you choose number of dataset you want.

Click estimate cost to see estimated credit cost.

And click generate training example to confirm.

Once finished, examples of dataset are shown.

You can download it in jsonl format.

You also can download txt to manually check.

There is python driven dataset structural checker.

There is also AI powered dataset review section.

AI evaluate, score the dataset quality, along with suggestion of revision. You can accept or skip.

Review the flagged examples and choose to accept or skip each suggestion.

Once you click to apply change, you can download revised jsonl file.

Download the amended dataset with your accepted changes applied.