-
Notifications
You must be signed in to change notification settings - Fork 20
Installation PyTorch from the source
Doing this with your normal user account. Do not use root.
refer to here
$ sudo apt install -y make build-essential libssl-dev zlib1g-dev libbz2-dev \
libreadline-dev libsqlite3-dev wget curl llvm libncurses5-dev libncursesw5-dev \
xz-utils tk-dev libffi-dev liblzma-dev python-openssl
install the pyenv using installation script
$ curl https://pyenv.run | bash
add env setting to your ~/.bashrc
export PATH="/home/jinserk/.pyenv/bin:$PATH"
eval "$(pyenv init -)"
eval "$(pyenv virtualenv-init -)"
refresh the env
$ source ~/.bashrc
install your preferred python binary. The installation path will be `$HOME/.pyenv/
$ pyenv install 3.7.2
update pip and install basic python env
$ pyenv local 3.7.2
$ pip install --upgrade pip
$ pip install --upgrade numpy scipy sklearn matplotlib pyyaml wheel
download deb(network) file from here
install deb file
$ sudo dpkg -i cuda-repo-ubuntu18sudo apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/7fa2af80.pub
04_10.0.130-1_amd64.deb
add GPG key for the repository
$ sudo apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/7fa2af80.pub
update repository info
$ sudo apt update
install cuda
$ sudo apt install cuda
Download deb files from here. You will need to register the NVIDIA developer membership.
Install deb files
$ sudo dpkg -i libcudnn7_7.4.2.24-1+cuda10.0_amd64.deb
$ sudo dpkg -i libcudnn7-dev_7.4.2.24-1+cuda10.0_amd64.deb
$ sudo dpkg -i libcudnn7-doc_7.4.2.24-1+cuda10.0_amd64.deb
If you have two or more GPUs on your machine, or you will use multi-GPUs cluster, you will need to install NCCL.
download Network Installers (x86) from here
Install deb file
$ sudo dpkg -i nvidia-machine-learning-repo-ubuntu1804_1.0.0-1_amd64.deb
update repository info
$ sudo apt update
install libnccl2 package
$ sudo apt install libnccl2
install latest pytorch
$ git clone https://github.com/pytorch/pytorch.git --recursive
$ cd pytorch
$ pyenv local 3.7.2
$ python setup.py bdist_wheel
$ pip install --upgrade dist/<built-whl>
install torchaudio
$ cd ..
$ git clone https://github.com/pytorch/audio.git
$ cd audio
$ python setup.py bdist_wheel
$ pip install --upgrade dist/<built-whl>
install torchnet
$ cd ..
$ git clone https://github.com/pytorch/tnt.git
$ cd tnt
$ python setup.py bdist_wheel
$ pip install --upgrade dist/<built-whl>