Expand description
Model execution traits for replay
These traits define how to re-execute a model during replay. Implementors connect the replay engine to actual model inference.
Structs§
- Comparison
Result - Comparison result between original and replayed outputs
- Echo
Executor - Echo executor that returns the original snapshot outputs. Useful for testing the replay pipeline without a real model.
- Execution
Config - Configuration for how to execute during replay
- Execution
Result - Result of model execution during replay
- Field
Comparison - NoOp
Executor - No-op executor that returns empty outputs. Used when no real model execution is available.
Traits§
- Model
Executor - Async trait for model re-execution during replay.
- Sync
Model Executor - Sync version of ModelExecutor for non-async contexts