AiModelResultTrainingRecipeExtensions
Helpers & Utilities · AiDotNet.Models.Results
Reports what a trained model’s research paper specifies for training, what was actually used, and every difference between the two.
How It Works
Reached through the facade, alongside the family-specific inference extensions from #1836, so a
caller who holds only an AiModelResult can still ask the
question. Model stays internal.
The gap this closes: every other library leaves a paper’s training recipe in prose — a model card, a README, a training script — detached from the model, not applied, and silently replaced by a generic default. Nothing tells you the model you just trained is an order of magnitude away from its published learning rate. That is exactly what #1928 found across 685 construction sites, and it went unnoticed for the life of the library.
Methods
| Method | Summary |
|---|---|
IReadOnlyList<TrainingRecipeReport> GetTrainingRecipe(AiModelResult<T, TInput, TOutput> result) |
What the paper specifies, what was applied, and why they differ — one report per component. |
IReadOnlyList<string> DescribeTrainingRecipe(AiModelResult<T, TInput, TOutput> result) |
A one-line summary per component, suitable for logging or a diagnostics panel. |
bool TrainsExactlyAsPublished(AiModelResult<T, TInput, TOutput> result) |
True when every component reproduces its paper exactly, with no adaptations and nothing unhonoured. |