Skip to main content

Crate reflow_litert

Crate reflow_litert 

Source
Expand description

Backend boundary for LiteRT-class inference in Reflow.

The default build remains deterministic and mock-first so CI and graph authoring do not require native ML libraries. The real LiteRT adapter is optional and targets the published Offbit litert crates, wiring them in behind the same backend traits without changing graph-facing actors or taskpacks.

Structs§

InferenceInput
InferenceOutput
MockBackend
ModelInfo
TensorSpec

Traits§

InferenceBackend
InferenceSession