mflux-model-porting - Port ML Models into mflux/MLX
Ports machine learning models into mflux/MLX, validates parity against a reference implementation, and refactors the result toward mflux conventions.
Tags
Updated: 2026-09-30Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Define parity targets
- Inspect reference artifacts
- Implement model packages
- Map model weights
- Validate latent and image parity
- Run deterministic tests
- Refactor shared components
- Wire standard model surfaces
Inputs
- Reference implementation
- Model configurations
- Model checkpoints
- Hugging Face cache tensors
- Prompts and generation settings
- Reference latents
- mflux source tree
Outputs
- Ported mflux model package
- Weight definitions and mappings
- Generated images and latent tensors
- Deterministic validation tests
- CLI and API integration changes
- Model documentation
Requirements
- MLX runtime
- mflux development environment
- Reference model environment
- uv test runner
- Required model access permissions