Projects and history.
Keep a book, score one applicant or a file against it, and find every run again.
Projects
A project groups the datasets and runs of one case. A run started inside a project saves its training data there as a dataset version, with its rows, columns and column types, so the same book can be reused without uploading it again.
From a project you can score:
- One applicant, filling in a form built from the dataset's columns.
- A batch, uploading a file of new rows.
Deleting a project removes its stored datasets and every run; the dialog asks you to type the project's name to confirm.
History
History lists your recent runs grouped by project, with the ad-hoc runs that belong to none. Each row shows when it ran, its status and mode, the target, the context and test sizes, accuracy, ROC-AUC, PR-AUC, GPU time and cost. A run opens to its details and, where they are stored, its predictions to download.
Statuses
| Status | Meaning |
|---|---|
| queued | Waiting for a worker. |
| running | Being scored. |
| done | Finished; results are stored. |
| error | Stopped with an error. |
| cancelled | Stopped by you before it finished. |