What’s New in ModelPie (formerly MrMTgui): Features & Migration Tips

From MrMTgui to ModelPie: Migration Checklist and Best Practices

Overview

ModelPie (formerly MrMTgui) is a rebranded/improved version of the MrMTgui toolset. Migrating projects from MrMTgui to ModelPie usually requires updating file paths, configuration formats, and adapting to new features or workflow changes. The checklist below gives a practical, step-by-step migration plan plus best practices to minimize downtime and preserve reproducibility.

Migration checklist

1. Inventory projects and assets

   • List all MrMTgui projects, scripts, models, datasets, and custom plugins.
   • Note versions and any non-standard dependencies.

2. Backup everything

   • Create versioned backups of project directories, config files, and databases.
   • Export trained models and metadata (including hyperparameters and training logs).

3. Read release notes & migration docs

   • Review ModelPie changelog and official migration guide for breaking changes and new config schemas.

4. Map configuration differences

   • Compare MrMTgui config files to ModelPie’s format; create a mapping table for keys that changed.
   • Update any deprecated parameters with their ModelPie equivalents.

5. Update file paths and naming

   • Adjust scripts and CI pipelines to new executable names, module imports, and CLI options.
   • Rename assets if ModelPie expects different directory structures or filenames.

6. Migrate custom code and plugins

   • Port custom plugins to the ModelPie plugin API; replace deprecated hooks.
   • Run static analysis and unit tests; refactor code where ModelPie provides built-in features that duplicate custom logic.

7. Migrate environments and dependencies

   • Recreate virtual environments using ModelPie’s recommended dependency set.
   • Pin versions and test for conflicts.

8. Test on a staging copy

   • Run a full pipeline (data prep → training → evaluation → export) in staging.
   • Compare outputs and metrics to original MrMTgui runs to confirm parity.

9. Validate model artifacts

   • Check exported model formats, weights, and metadata for compatibility with deployment targets.
   • Run inference tests (unit and integration) against saved baselines.

10. Update CI/CD and deployment

    • Modify build/deploy scripts, container images, and orchestration configs.
    • Ensure monitoring, logging, and rollback strategies are compatible.

11. Document changes

    • Record migration steps, updated configs, and known issues for team reference.
    • Update READMEs and onboarding docs.

12. Schedule cutover and support

    • Plan a maintenance window for production switch-over.
    • Have rollback plan and staff on call during migration.

Best practices

• Start small: Migrate a low-risk project first to learn pitfalls.
• Automate transforms: Script config translations and bulk renames to avoid manual errors.
• Keep reproducibility: Preserve exact random seeds, environment specs, and dataset snapshots.
• Leverage new features: Replace fragile custom scripts with ModelPie-native utilities when stable.
• Monitor closely post-migration: Compare model performance metrics for drift or regressions.
• Communicate changes: Inform stakeholders about new commands, paths, and expected behavior.
• Maintain backward compatibility: Where possible, wrap ModelPie calls to emulate MrMTgui interfaces until all consumers update.

Quick troubleshooting

• If outputs differ: re-check preprocessing steps and RNG seeds.
• If plugins fail: confirm API version and rebuild plugin with ModelPie SDK.
• If dependency conflicts: isolate with containers (Docker) to reproduce environment.

If you want, I can generate a config-mapping template or a scripted converter for typical MrMTgui -> ModelPie config files — tell me the format you use (YAML/JSON/INI).