Manage datasets | Dagnam.AI Docs

Manage datasets Once a dataset is uploaded, the Datasets page is where you find it, preview it, confirm its column roles, and rename or delete it. This page also covers linking a dataset to a project and the built in datasets. Find a dataset The Datasets page has a search box, filters by Type , Format and Analysis Status , and sorts by Name , Size , Samples or Created ; on wider screens you can switch between a table view and a grid view. A Storage Usage bar at the top shows how much of your quota is used; the row menu and a bulk Delete cover individual and multiple datasets at once. Preview and statistics Click a dataset's name to open its preview. Statistics shows summary numbers for the dataset; Preview renders a sample of its rows, images, text or audio depending on its type. When you open a preview from Studio's Data tab with a project open, Add to Project drops the dataset onto the canvas as an unconnected Dataset node. Column roles Once analysis completes, the preview's role editor shows the dataset's Task Type and lets you assign each column a role, for example Feature , Target , Text Input or Image Input , so Dagnam knows how to use each column when you link the dataset to a training objective. Until you review the roles and click Save , the preview shows a Roles not confirmed badge and the editor an Unconfirmed badge. Before analysis completes, the editor is replaced by "Role editing available after analysis completes". Rename, delete and download The row menu offers Preview , Rename and Delete which asks for confirmation and cannot be undone . To download a dataset, use dagnam dataset download from the CLI /docs/dag-lib/cli , which needs a personal API key. Use a dataset in a project Link a dataset to a project with a role: Training , Validation or Testing . On the dashboard, open a project's menu, choose Manage Datasets , click Add Dataset , pick a dataset, choose its Dataset Role , and click Link Dataset . In Studio, you use a dataset by dragging it from the Data tab onto the canvas as a Dataset node, which has its own Training , Validation and Testing roles and split. Adding a Dataset node does not link the dataset to the project. See Studio inputs and outputs /docs/studio/inputs-and-outputs for how the Dataset node's splits work. Built-in datasets Studio's Data tab, under System , ships these built-in datasets: Dataset Task ------------------------ -------------------- MNIST Handwritten Digits Image classification Fashion-MNIST Image classification CIFAR-10 Image classification CIFAR-100 Image classification IMDB Movie Reviews Text classification Oxford-IIIT Pet Dataset Image segmentation Speech Commands Audio classification WikiText-2 Language modeling Synthetic Video Clips Video classification Use a dataset in Python The dagnam SDK loads any of your datasets, including the built-in ones, without going through the web app. See Load datasets in Python /docs/dag-lib/datasets for load dataset and the framework-specific converters.