Overview
On the dataset page, you can control who has access to your dataset using privacy settings. Right now, there are only two options:
Private – Only you can see and use the dataset.
Team – Everyone in your organization can see and use the dataset.
You can also see the number of examples in your dataset, which helps you gauge whether it’s large enough for effective training.
How to Use
Choose privacy settings
Go to the Datasets tab and select your dataset.
Under Privacy, choose either Private or Team depending on how you want to share access.
Check dataset size
The dataset page displays the number of examples you’ve uploaded.
While there’s no strict minimum, here are rough guidelines:
Good dataset: ~10,000 examples
Acceptable dataset: ~3,000 examples
The right number depends on the quality of your data and the complexity of the task.
Use in training
Any dataset you own or have access to (private or team) can be selected for training your models.
Tips & Best Practices
Start with quality: A smaller, well-structured dataset often outperforms a large but messy one.
Think about collaboration: Use Team mode to share datasets with your organization for faster iteration.
Iterate often: Upload a baseline dataset, train, then refine by adding more examples.
Troubleshooting
I can’t share my dataset with teammates. Why?
I can’t share my dataset with teammates. Why?
Make sure the dataset privacy is set to Team. By default, datasets are private. If you continue having trouble, though, please email us at [email protected].
How many examples do I really need?
How many examples do I really need?
There’s no single answer, but 3,000 examples is a reasonable minimum. For higher quality or more complex tasks, aim for around 10,000 examples.