Deployment

1. Clone the repository

   git clone -b simple-client https://github.com/pw-02/ezkl-exploration.git

2. Install Python libraries:

   pip install -r requirements.txt
   

3. Make 'to_run.sh' executable:

    chmod +x to_run.sh

4. Execute 'to_run.sh':

    ./to_run.sh

   The 'to_run.sh' contains the commands to generate proofs based on the specified configuration (see config section below for information)

Configurations :

Parameter	Description
model	The model for which to generate a proof. This should match the name of a .yaml file in the conf/model folder. The YAML file includes default values for the following: - onnx_file: Path to the ONNX model used for proving. - input_file: Path to the input data for inference. - split_group_size: Number of sub-models per group when splitting the model. For example, a model with 100 nodes and split_group_size = 5 will be divided into 20 sub-models. To prove the model without splitting, set this value to null or 0. These values can be overridden via the command line. For example, to set split_group_size, use: model.split_group_size={your_value}
cache_setup_files	A Boolean (True/False) value indicating whether to save setup files between runs. If enabled, the files will be stored in a folder named Cache at the root directory. Default is False.

Reporting

All reports are saved in a /reports/model/{name}/{timestamp} directory, where {name} corresponds to the model's name and {timestamp} represents the time of execution. The reports folder is located at the root of the project. There are several different reports:

Report Name	Description
models_to_prove_summary.csv	A summary of the models to be proven as part of the job. This report is generated before proof generation begins and contains information from the EZKL settings API.
msms_summary.csv	A summary of MSM operations across all proven sub-models.
ntts_summary.csv	A summary of NTT operations across all proven sub-models.
halo2_prover_summary.csv	Performance metrics collected directly from Halo2 for each proven sub-model.
ezkl_perf_summary.csv	Performance metrics of EZKL APIs. Note: EZKL includes some additional overhead compared to Halo2.
halo2_circuit_summary.csv	A summary of the Halo2 circuit for each sub-model.