This is a TensorFlow Lite native module for Veracruz.
It is meant to be executed by the native module sandboxer in a sandbox environment everytime a WebAssembly program invokes it.
Just like any native module, it is an entry point to a more complex library and only exposes preselected high-level features to the programs invoking it.
This native module takes an execution configuration file specifying:
- the input tensor's path (only one supported yet)
- the model's path
- the output tensor's path (only one supported yet)
- the number of CPU threads to use
It then performs inference on the model with the given input tensor and outputs a tensor. Training is not supported yet.