Installation

Software Environment

  • Python 3.7 or 3.8.

  • Tensorflow >=2.0.0, (Deep learning models require Tensorflow2).

Install with pip

pip install hyperts

Install with conda

conda install -c conda-forge hyperts

Notes

1.1 Prophet is required by HyperTS (version < 0.2.0). When installing hyperts using pip, it is recommended that you first install Prophet using conda:

conda install -c conda-forge prophet==1.0.1
pip install hyperts

1.2 Since HyperTS version 0.2.0, hyperts relaxes prophet’s version (compatible with prophet==1.1.1), so it is possible to install hyperts without first installing prophet using conda:

pip install hyperts

  1. Tensorflow is an optional dependency for HyperTS. You can install tensorFlow manually when deep learning and neural architecture search modes, for example:

conda install -c conda-forge prophet==1.0.1
pip install hyperts tensorflow

or

conda install -c conda-forge hyperts
pip install tensorflow

  1. If your device supports GPU, you can manually install the tensorFlow-gpu version to speed up the deep learning model.

  2. If you meet the error shown below when using the DL mode, please check and change the pyparsing version not more than 2.4.7.

Frappe installation error "AttributeError: module 'pyparsing' has no attribute 'downcaseTokens'".

  1. If you meet other problems when using tensorflow, please first check the compatibility of numpy and tensorflow. Sometimes, select lower numpy versions (i.e, 1.19.5) would help to solve the problems.

NotImplementedError: Cannot convert a symbolic Tensor (gru_1/strided_slice:0) to a numpy array.
This error may indicate that you're trying to pass a Tensor to a NumPy call, which is not supported.

  1. When using the STATS model for univariate forecasting, you may encounter the following problems:

ValueError: In models with integration (d > 0) or seasonal integration (D > 0)...

Please check the statsmodels version and control that it is not greater than 0.12.1.