Configuration

103 public types in this namespace, organized by kind.

Models & Types (18)

Type Summary
AiModelDataPipeline<T, TInput, TOutput> Default implementation of IAiModelDataPipeline.
AiModelTrainingCore<T, TInput, TOutput> Default implementation of IAiModelTrainingCore.
AutoMLSearchSpace Expert control over the AutoML search space — which model types AutoML tries and the hyperparameter ranges it explores for them.
EvolutionDescriptorYamlConverter Reads and writes an EvolutionDescriptorDefinition in a YAML configuration file.
FusedOptimizerPathEvent Emitted when TrainWithTape enters or skips the fused-compiled fast path.
GradientNormEvent Emitted per parameter tensor after tape.ComputeGradients returns, before the optimizer step runs.
ProgramInterpreterSpecification Names the executable and argument template used to run one language inside the program sandbox.
RLEpisodeMetrics<T> Metrics for a completed RL episode.
RLStepMetrics<T> Metrics for a single RL training step.
RLTrainingSummary<T> Summary of completed RL training.
TrainingLossEvent Emitted once per call to TrainWithTape with the scalar loss value returned by the loss function for that batch.
TrainingMessageEvent Free-form text event for callers that just want to emit a typed log line at a specific level.
YamlLicenseSection YAML configuration section for license key settings.
YamlOptimizerSection YAML section for selecting an optimizer type by name.
YamlPipelineSection YAML section for pipeline-style configurations (preprocessing, postprocessing, etc.).
YamlRagSection Composite YAML section for the full retrieval-augmented generation (RAG) pipeline.
YamlTimeSeriesModelSection YAML section for selecting a time series model type by name.
YamlTypeSection Generic YAML section for any type-based configuration.

Base Classes (1)

Type Summary
TrainingDiagnosticEvent Base type for structured training-pipeline diagnostic events.

Interfaces (2)

Type Summary
IAiModelDataPipeline<T, TInput, TOutput> Component that owns the data-pipeline configuration for an AI model build: preprocessing, postprocessing, data loading, data preparation, and augmentation.
IAiModelTrainingCore<T, TInput, TOutput> Component that owns the core training configuration for an AI model build: the model itself, optimizer, regularization, fitness calculator, fit detector, training pipeline, training monitor, checkpoint manager, and memory management.

Enums (11)

Type Summary
AttentionMaskingMode Controls how attention masking is applied for optimized attention implementations.
CacheEvictionPolicy Cache eviction policies for KV cache management.
DifficultyEstimatorType Types of difficulty estimators for curriculum learning.
ExplorationDecayType Type of exploration decay schedule.
GpuDiagnosticLevel Verbosity level for GPU backend diagnostic output.
InferenceQuantizationMode Specifies the weight quantization mode for inference optimization.
KVCachePrecisionMode Controls the numeric precision of KV-cache storage.
KVCacheQuantizationMode Controls optional KV-cache quantization for inference.
SpeculationPolicy Policies for enabling/disabling speculative decoding at runtime.
SpeculativeMethod Selects the speculative decoding method.
TrainingDiagnosticLevel Verbosity level for training-pipeline diagnostic output (gradient norms, optimizer step traces, tape replay events, etc.).

Delegates (2)

Type Summary
GpuDiagnosticSink Delegate that receives GPU backend diagnostic messages in lieu of WriteLine.
TrainingDiagnosticSink Delegate that receives training-pipeline diagnostic events in lieu of the default WriteLine path.

Options & Configuration (63)

Type Summary
AutoMLBudgetOptions Configuration options that control AutoML compute budgets.
AutoMLEnsembleOptions Configuration options for AutoML ensembling.
AutoMLMultiFidelityOptions Configuration options for multi-fidelity/ASHA AutoML search.
AutoMLOptions<T, TInput, TOutput> Configuration options for running AutoML through the AiDotNet facade.
BenchmarkingOptions Configuration options for running benchmarks through the AiDotNet facade.
CausalLmBenchmarkOptions Configuration options for causal language-model dataset suites (LAMBADA and Perplexity).
ChatClientOptions Everything a chat client needs around it: defaults, retry policy, timeouts, filters, and recording.
CifarFederatedBenchmarkOptions Configuration options for running CIFAR-based federated benchmark suites.
CompetenceBasedOptions Options specific to competence-based curriculum learning.
CurriculumEarlyStoppingOptions Early stopping options for curriculum learning.
CurriculumLearningOptions<T, TInput, TOutput> Configuration options for Curriculum Learning through the AiDotNet facade.
DenseNetConfiguration Configuration options for DenseNet neural network architectures.
EfficientNetConfiguration Configuration options for EfficientNet neural network architectures.
EmbeddingNoveltyOptions Immutable, constructor-validated thresholds for the structural, embedding, and judge novelty stages.
EvolutionOptions The user-facing control panel for an evolution run: budgets, islands, the archive grid, and persistence.
EvolutionSeedOptions The starting points a program-evolution run begins from, expressed as plain source text.
ExplorationScheduleConfig<T> Configuration for exploration schedule (epsilon decay for epsilon-greedy).
FederatedTabularBenchmarkOptions Configuration options for federated tabular benchmark suites.
FederatedTextBenchmarkOptions Configuration options for federated text benchmark suites.
FederatedVisionBenchmarkOptions Configuration options for federated vision benchmark suites.
GpuDiagnosticsConfig Process-global control for GPU backend diagnostic output visibility.
GpuDiagnosticsOptions Options for controlling GPU backend diagnostic output visibility.
InferenceOptimizationConfig Configuration for inference-time optimizations to maximize prediction throughput and efficiency.
JitCompilationConfig Configuration for JIT (Just-In-Time) compilation of model forward/backward passes.
LeafFederatedBenchmarkOptions Configuration options for running LEAF-backed federated benchmark suites.
LlmFeedbackOptions Configures how a language model’s opinion of a program is blended into that program’s fitness.
LlmProgramVariationOptions Configures how a language model is prompted to propose the next candidate program.
ManualChatClientOptions Settings for the chat client that asks a person instead of a model.
MapElitesAutoMLOptions Configures quality-diversity search when AutoML uses the MAP-Elites strategy.
MobileNetV2Configuration Configuration options for MobileNetV2 neural network architectures.
MobileNetV3Configuration Configuration options for MobileNetV3 neural network architectures.
PrioritizedReplayConfig<T> Configuration for prioritized experience replay (PER).
ProcessChatClientOptions Settings for a chat client that runs a command-line program instead of calling an HTTP endpoint.
ProgramArtifactStoreOptions Limits and retention rules applied to stored evaluation artifacts.
ProgramDiffOptions Configures how SEARCH/REPLACE edit blocks are parsed from a model response and applied to a program.
ProgramEvolutionOptions The single configuration surface for evolving source code: seeds, bounds, prompting, sandbox, and search.
ProgramEvolutionPromptOptions Controls what a program-evolution prompt says, what it shows, and how large it is allowed to get.
ProgramMetricAggregationOptions Declares how a metrics dictionary is collapsed into the single quality an archive ranks on.
ProgramRunOutputOptions Names the directories and files a run writes its best program into.
ProgramSandboxLimitOptions Resource limits applied to every sandboxed execution of an untrusted candidate program.
ProgramSandboxOptions Configures the boundary, the limits, and the per-language commands used to execute untrusted programs.
ProposalProvenanceOptions Configures how much of each language-model request and answer is kept as provenance.
RLAutoMLOptions<T> Configuration options for running AutoML over reinforcement learning agents and hyperparameters.
RLCheckpointConfig Configuration for checkpointing during RL training.
RLEarlyStoppingConfig<T> Configuration for early stopping during RL training.
RLEvaluationConfig Configuration for evaluation during training.
RLTrainingOptions<T> Configuration options for reinforcement learning training loops via AiModelBuilder.
ReasoningModelOptions Configures how requests are adapted before they are sent to a reasoning model.
RedditFederatedBenchmarkOptions Configuration options for running the Reddit federated benchmark suite.
ResNetConfiguration Configuration options for ResNet (Residual Network) neural network architectures.
RewardClippingConfig<T> Configuration for reward clipping.
ScriptProgramEvaluationOptions Configures how a caller-supplied evaluator script scores candidate programs inside the sandbox.
SelfPacedOptions Options specific to self-paced curriculum learning.
Sent140FederatedBenchmarkOptions Configuration options for running the Sent140 LEAF federated benchmark suite.
ShakespeareFederatedBenchmarkOptions Configuration options for running the Shakespeare LEAF federated benchmark suite.
SpeculativeDecodingOptions Tuning knobs for speculative decoding — the technique that speeds up autoregressive (GPT-style) text generation by letting a small, fast “draft” model guess the next few tokens and having the main model verify all of them in a single pass.
StackOverflowFederatedBenchmarkOptions Configuration options for running the StackOverflow federated benchmark suite.
SyntheticTabularFederatedBenchmarkOptions Configuration options for the synthetic federated tabular benchmark suite.
TargetNetworkConfig<T> Configuration for target network updates in DQN-family algorithms.
TrainingDiagnosticsConfig Process-global control for training-pipeline diagnostic output (gradient norms, optimizer step traces, tape events).
VGGConfiguration Configuration options for VGG neural network architectures.
WeightedEnsembleChatClientOptions Controls how a weighted ensemble picks a model, what it does when one fails, and what it reports as.
YamlModelConfig Root POCO that YAML/JSON config files deserialize into.

Helpers & Utilities (5)

Type Summary
GpuDiagnosticsLoggerExtensions Extension methods that adapt ILogger into a GpuDiagnosticSink.
YamlConfigLoader Loads and deserializes YAML configuration files into strongly-typed configuration objects.
YamlDocsGenerator Generates markdown reference documentation for the AiDotNet YAML configuration system.
YamlJsonSchema Generates a JSON Schema for the AiDotNet YAML configuration system.
YamlVariableResolver Substitutes ${NAME} references in a YAML document before it is deserialized.

Attributes (1)

Type Summary
YamlConfigurableAttribute Marks an interface or abstract base class as discoverable by the YAML configuration system.