silvic_neuron/lib.rs
1//! AWS Neuron training and explicit float32 serving.
2//!
3//! The existing Rust learner drives [`NeuronBackend`]. [`NeuronPredictor`] scores
4//! LightGBM models with float32 inputs, thresholds, leaves, ordered sums and output;
5//! it does not promise exact float64 predictions. Both require a Neuron runtime
6//! and device; device failures never select another backend.
7
8mod compiler;
9mod runtime;
10mod serving;
11mod training;
12
13pub use serving::NeuronPredictor;
14pub use training::NeuronBackend;
15
16/// Whether the native runtime exposes NeuronCores, without initializing it or reserving cores.
17pub fn is_available() -> bool {
18 runtime::is_available()
19}