Author: Lijun Yu

Email: lijun@lj-y.com

Base image for ActEV SDL submission of CMU.

• Python 3.7
• CUDA 10.1
• PyTorch 1.6
• PyTorch Lightning
• OpenCV
• Open3D
• Anaconda
• Detectron2
• Pyturbo
• AVI-R with FFmpeg 4.3.1
• Dependencies of ActEV_Scorer

Github Packages

docker pull ghcr.io/cmu-inf-diva/actev_base

# Test it
docker run -it --rm --gpus all --ipc host ghcr.io/cmu-inf-diva/actev_base \
    bash -ic "python -c 'import torch; assert torch.cuda.is_available()'; nvidia-smi"

Container directory structure

• /app - Work directory
  • env - Python environment
  • core - System implementation
  • data - Mount point for data

Install miniconda, then run

./setup.sh [<name>] # name defaults to `actev`

If you want to patch PyTorch with this PR, run

conda activate <name>
./patch_pytorch.sh