Module forward_model 

Solver-agnostic forward model trait.

Unlike FitModel, which is LM-specific (evaluates residuals + Jacobian in chi-squared space), ForwardModel exposes the raw model prediction and parameter Jacobian. Each solver wraps this to compute its own objective and gradient:

• LM: residuals = (data - predict) / σ, J_lm = -jacobian / σ
• KL: gradient = Σ (1 - data/predict) · jacobian

This makes new model extensions (background, temperature, etc.) work with both solvers automatically — implement ForwardModel once, both solvers benefit.

Traits

ForwardModel

Solver-agnostic forward model.