EZKL Test

This is a test example of ezkl. The test model is a simple convolutional neural network containing two convolutional layers and two fully connected layers, using ReLU as the activation function. It can be used to process batches of data with input shapes of (32, 1, 28, 28) and ultimately outputs predicted scores for 10 categories corresponding to each sample. Proofs are generated and then tested for validation on the chain

Dependencies

• Python>=3.7
• ezkl=11.2.2
• onnx=1.16.1

Usage

Run a local block chain

# install anvil if you haven't already
cargo install --git https://github.com/foundry-rs/foundry --profile local --locked anvil

# spin up a local EVM through anvil in a separate terminal 
anvil -p 3030

Install

EZKL

Install as ezkl installing document

Python Environment

python3 install.py

Run

python3 main.py