Models & Tensors
Direct model execution, tensor allocation, and the explicit memory model behind the core primitives of React Native ExecuTorch.
Schema Validation
Declare the shape and dtype contract a model must satisfy, and validate any .pte against it before allocating tensors or running inference.
Operations & Utilities
The building blocks for custom pipelines — native tensor operations (math, cv, speech) plus higher-level TypeScript utilities for tokenization and language models (nlp, llm).
Error Handling
Every error thrown by the library includes a machine-readable code. Use isRnExecuTorchError to safely narrow errors across worklet and JSI boundaries.
Worklets & Threading
Execute native model loading and inference off the React Native JavaScript thread using worklet runtimes, wrapAsync, and thread-safe native tensors.
Exporting Custom Models
Bring your own .pte model and make it fit a pipeline — inspecting schemas, matching the contract, and declaring dynamic shapes with get_model_schema.
Native Libraries
How React Native ExecuTorch downloads, ships, and links native binaries on demand.