Skip to content

Model Comparison Report for Training Runs #1551

Description

@coderleeon

Is your feature request related to a problem? Please describe.

When experimenting with different training configurations, it can be difficult to compare model quality across runs.

Users often rely on subjective visual inspection, making it hard to determine whether changes in settings, datasets, or training duration actually improved results.

Describe the solution you'd like

Add a training comparison report that allows users to compare multiple model runs side-by-side.

Potential metrics:

  • Training duration
  • Iteration count
  • Loss trends
  • Preview snapshots at selected checkpoints
  • Model configuration differences

The report could help users understand which changes led to meaningful improvements and make training experiments more reproducible.

Describe alternatives you've considered

Users can manually save previews and training logs, but comparing runs becomes tedious as the number of experiments grows.

Additional context

This would be particularly useful for users tuning models, testing new training settings, or documenting reproducible workflows.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions