Skip to content

Installation

Required

1. Install PyTorch (prerequisite)

Sapsan can be run on both cpu and gpu. Below are the requirements for each version

Device
CPU torch>=1.9.0 torchvision>=0.10.0
GPU torch>=1.9.0+cu111 torchvision>=0.10.0+cu111

Please follow the instructions on PyTorch to install either version. CUDA>=11.1 can be installed directly with PyTorch as well.

Install CUDA-Toolkit

If you are planning to train on GPU, then you need to check that you have an nvidia-cuda-toolkit installed: 

1
nvcc --version
if it is missing, you can install it via: 
1
sudo apt install nvidia-cuda-toolkit

Device

If GPU is available, Sapsan will always try to run on a GPU by default (that includes the tests). However, you can specify the device='cpu' in model config.

1
pip install sapsan

2b. Clone from github (alternative)

1
2
3
git clone https://github.com/pikarpov-LANL/Sapsan.git
cd Sapsan/
python setup.py install

If you experience any issues, you can try installing packages individually with: 

1
pip install -r requirements.txt

Version

Make sure you are using the latest release version!

Optional

Install Graphviz

Info

pip version of graphviz that installs with Sapsan is not enough!

In order to create model graphs, Sapsan is using graphviz. If you would like to utilize this functionality, pip graphviz wrapper is not enough. To install the core package for graphviz: 

1
conda install graphviz
or 
1
sudo apt-get install graphviz

Install Docker

In order to run Sapsan through Docker or build your own container to share, you will need to install it 

1
pip install docker

Next, you can build a docker setup with the following:

1
make build-container

this will create a container named sapsan-docker.

If you want to run the container, type:

1
make run-container
a Jupyter notebook will be launched at localhost:7654

Troubleshooting

libGL.so error

If you get the following error: 

1
ImportError: libGL.so.1: cannot open shared object file: No such file or directory
your opencv-python package has some dependency issues. To resolve, try the following: 
1
2
apt-get update
apt-get install ffmpeg libsm6 libxext6  -y

My Kernel is Dying!

If when attempting to train the model (executing Train.run()) the kernel is dying without any errors, check if you have a GPU on your machine while not having a nvidia-cuda-toolkit installed. To resolve, you can either install it: 

1
sudo apt install nvidia-cuda-toolkit
or specify device='cpu' in your model config. By defaul Sapsan always tries to train the models on a GPU if one is available.