Skip to main content

Framework and Dependency Installation Guide

aifare platform supports installing AI frameworks, Python dependencies, and system dependencies via pip, conda, apt, and other methods. Below are common installation methods.

Install PyTorch​

  1. It is recommended to use pip for installation. For version selection, refer to the PyTorch official website
pip install torch==2.4.1 torchvision==0.19.1 torchaudio==2.4.1
  1. For users in China, it is recommended to remove the -f/--index-url parameter to use domestic pip mirrors for faster speed.

Install TensorFlow​

  1. It is recommended to use pip to install the official whl package. For version selection, refer to the TensorFlow official website
pip install https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-2.6.0-cp38-cp38-manylinux2010_x86_64.whl

Install PaddlePaddle​

pip install paddlepaddle-gpu==2.6.2

Install Python Dependencies​

  • Use pip to install
pip install opencv-python scipy numpy Pillow
  • Use conda to install
conda install numpy
  • Search for package name/version
pip search xxxx
conda search xxxx
  • View available pip versions
pip install xxx==9.9  # This will throw an error and list all available versions

Install System Dependencies​

  • For example, to install zip
apt-get update    # Only needed once
apt-get install -y zip
  • Search for system package name
apt-cache search xxxxx

Dependency Environment Reuse and Migration​

  • Dependencies will not be lost after the aifare platform instance is shut down; no need to reinstall after restart
  • To reuse the dependency environment across hosts, you can save it as an image or migrate the instance. See related documentation for details

For more dependency installation tips, please refer to the aifare platform documentation or community resources.