Skip to main content

Module inference

Module inference 

Source
Expand description

Inference-only model format for WASM deployment

This module provides a pure Rust implementation of neural network inference that stays off the training-side Burn tensor stack so it can compile to WebAssembly with a minimal bundle size. The training side (PPO/DQN) runs on Burn; weights are exported as JSON and consumed here for browser inference. See crate::inference module docs and docs/WASM_ROADMAP.md.

Structs§

InferenceModel
A serializable MLP model for inference
SnakeCNNInference
A serializable CNN model for Snake inference
TrainingMetadata
Training metadata for the model

Enums§

InferenceActivation
Activation function type for inference