INSTALL.md

Installation

We have two different configs for DDRNetOC (since it requires older wersion of PyTorch) and for all other models.

Environment Setup

Base config

• Ubuntu 20.04.3 LTS
• Python 3.8
• PyTorch v1.8.0 (CUDA 11.1 build)
• TensorFlow v2.2.0

Create a conda environment.

conda create -n daps_base python=3.8
conda activate daps_base

Install PyTorch, TensorFlow and other dependencies.

pip install torch==1.8.0+cu111 torchvision==0.9.0+cu111 torchaudio==0.8.0 -f https://download.pytorch.org/whl/torch_stable.html

pip install -r requirements_base.txt

DDRNetOC config

• Ubuntu 20.04.3 LTS
• Python 3.7
• PyTorch v1.1.0 (CUDA 10.0 build)
• TensorFlow v1.13.1

Create a conda environment.

conda create -n daps_oc python=3.7
conda activate daps_oc

Install PyTorch, TensorFlow and other dependencies.

conda install pytorch==1.1.0 torchvision==0.3.0 cudatoolkit=10.0 -c pytorch
conda install -c conda-forge cudatoolkit-dev

pip install -r requirements_oc.txt

WandB Setup

Use WandB in order to track your runs.

Create a login a file in the project directory DAPS3D/wandb.login:

Experiments_daps3d    <----    <project name>
subake                <----    <account name>
key                   <----    <wandb login key>