The project involves forecasting pool volumes using time-series data. It begins with loading data from Polar into a DataFrame, extracting relevant features such as day of the week, month, and year, and performing one-hot encoding for categorical variables. After standardizing numerical features and preparing the input sequences by sliding a window across the time-series data, a WaveNet model is built and trained using TensorFlow/Keras. The trained model is then utilized to predict pool volumes on unseen data. The predictions are further processed using ONNX, EZKL, with the verifiable=True setting converting them to Cairo format, optimizing them for deployment or downstream tasks.
- Giza Actions SDK
- Giza cli
- Giza Virtual Environment
- Giza Dataset
- WaveNet
- Jupyter Notebook
- Tensorflow
- Poetry
- Cairo
- EZKL
- ONNX
Welcome to the Giza Actions SDK template! The Giza Actions SDK is tailored to assist you in designing your ZKML workflows efficiently. This project provides pre-configured actions ready for deployment on the Giza platform. The purpose of this template is to demonstrate how to construct your ZKML workflows using the Giza Actions SDK.
Note: This template is based on the MNIST tutorial. Please be aware that certain steps, such as transpiling the model and deploying the generated model on Giza Plateform, are required between action executions. For a more comprehensive understanding, refer to the tutorial.
- Python 3.11
- Poetry
$ poetry shell
$ poetry installWithin the yearn directory, you'll discover multiple generated files:
train_action.py: Contains actions for training your model.predict_onnx_action.py: Includes actions for making predictions with an ONNX model.predict_cairo_action.py: Includes actions for making verifiable predictions with the Orion Cairo model.
To use this project, follow these steps:
- Install the required dependencies.
- Execute any of the provided action scripts using the command
python yearn/{action_file}.py, for example,python yearn/{train_action}.py.
Explore more about the Giza Actions SDK here.
This template was generated using cookiecutter.