Skip to content

Latest commit

 

History

History
86 lines (62 loc) · 3.17 KB

install.rst

File metadata and controls

86 lines (62 loc) · 3.17 KB
Raw

Installation

How to Install

It is recommended to use pip or conda for PyPOTS installation as shown below:

# by pip
pip install pypots            # the first time installation
pip install --upgrade pypots  # update pypots to the latest version
# by conda
conda install -c conda-forge pypots  # the first time installation
conda update -c conda-forge pypots   # update pypots to the latest version

Alternatively, you can install from the latest source code which may be not officially released yet:

pip install https://github.com/WenjieDu/PyPOTS/archive/main.zip

Required Dependencies

  • Python >=3.7, <=3.10
  • numpy
  • scipy
  • scikit-learn
  • pandas <2.0.0
  • torch >=1.10.0
  • tensorboard
  • h5py
  • tsdb
  • pycorruptor

Optional Dependencies

  • torch-geometric (optional, required for GNN models like Raindrop)
  • torch-scatter (optional, required for GNN models like Raindrop)
  • torch-sparse (optional, required for GNN models like Raindrop)

Reasons of Version Limitations on Dependencies

  • Why we need python >=3.7?

PyG (torch-geometric) is available starting from python v3.7, please refer to https://pytorch-geometric.readthedocs.io/en/latest/install/installation.html#installation-via-anaconda. Although torch-geometric is an optional dependency, we hope things go smoothly when our users opt to install it.

  • Why we need pandas <2.0.0?

Because v2 may cause ModuleNotFoundError: No module named 'pandas.core.indexes.numeric', see https://stackoverflow.com/questions/75953279/modulenotfounderror-no-module-named-pandas-core-indexes-numeric-using-metaflo.

  • Why we need PyTorch >1.10?

Because of pytorch_sparse, please refer to rusty1s/pytorch_sparse#207 (comment).

Acceleration

GPU Acceleration

Neural-network models in PyPOTS are implemented in PyTorch. So far we only support CUDA-enabled GPUs for GPU acceleration. If you have a CUDA device, you can install PyTorch with GPU support to accelerate the training and inference of neural-network models. After that, you can set the device argument to "cuda" when initializing the model to enable GPU acceleration. If you don't specify device, PyPOTS will automatically detect and use the first CUDA device (i.e. cuda:0) if multiple CUDA devices are available.

CPU Acceleration

If you're using a Mac device with Apple Silicon in you can install the accelerate data-science packages to obtain faster processing speed, because they get optimized for Apple Silicon. conda install numpy scipy scikit-learn numexpr "libblas=*=*accelerate"

If you're using devices with Intel chips in, you should install the distribution of MKL, which is optimized for multi-core Intel CPUs, conda install numpy scipy scikit-learn numexpr "libblas=*=*mkl"

If you're using devices with AMD chips in, you can install with the distribution of OpenBLAS, conda install -c conda-forge numpy scipy scikit-learn numexpr "libblas=*=*openblas"