The experiment run table currently displays only a fixed set of columns: name, tags, and status. We need the ability to configure which attributes appear as columns.
Current Behavior:
- Run table shows fixed columns: name, tags, status
- No way to add custom attributes as columns
- Important metadata (e.g., job IDs, usernames) requires clicking into each run
Desired Behavior:
- Allow users to select which attributes appear as table columns
- Support common attributes out of the box:
- Trainy job ID
- User name
- Experiment config (e.g., model type, dataset)
- Hardware info (GPU type, node count)
- Persist column configuration per user/project
- Enable sorting and filtering by custom columns