ezkl is a library and command-line tool for doing inference for deep learning models and other computational graphs in a zk-snark (ZKML) using halo2 as a backend. Official documentation.
ezkl represents a pioneering framework in zero-knowledge machine learning (ZKML) trusted by Giza. It is designed with ease of use in mind and efficiency to make easy zero-knowledge inference.
EZKL Framework Features
The transpilation process refers to performing the setup for the model where an ONNX model and a set of input data are used to generate the circuit settings, proving key and verification key files that are used to perform the inference. In Giza we aim to provide an easy way to perform this step providing compute resources and a simple command line interface. More on this in the transpile documentation for ezkl.
Creating Verifiable Proofs
Create a proof using the generated outputs of the setup process, we handle this step for you by loading the necessary files and performing the proof generation for you. More on this in the prove documentation for ezkl.
Verify the proof generated by an ezkl version, here we manage the execution of the verification process and the compute as well for you. More on this in the verify documentation for ezkl