Skip to main content

Module config

Module config 

Source
Expand description

Inference-facing configuration types.

Carved from the main crate’s config module (Wave B, hyprstream-inference extraction): the runtime/model/generation/training configuration surface the inference engine and training pipeline own. The main crate re-exports these from crate::config so existing paths keep resolving.

AGPL-3.0-only. This crate is part of hyprstream’s combined AGPL distribution.

Structs§

GenerationConfig
Text generation parameters
GenerationResult
Generation result with metadata
HyprstreamTrainingConfig
Model-level training mode configuration (embedded in config.json under “hyprstream_training”)
ModelConfig
Model loading and identification
RuntimeConfig
Runtime execution configuration
TTTMetrics
TTT adaptation metrics (mirrors training::ttt::TTTResult)
TTTTrainingConfig
TTT-specific configuration

Enums§

FinishReason
Why generation stopped
TrainingMode
Training mode configuration

Functions§

default_lora_rank
default_target_modules