- Cuda 11.4.4 (filename
cuda_11.4.4_472.50_windows.exe
, fourth update) - CUDNN 8.2.1 (filename
cudnn-11.3-windows-x64-v8.2.1.32.zip
) - TensorRT 8.2.3.0 (filename
TensorRT-8.2.3.0.Windows10.x86_64.cuda-11.4.cudnn8.2.zip
, second update)
The docker image is available at andrew/ecub_env:latest
, or you can build it with:
docker build -t ecub .
You can launch the bash of the docker's image with access to the GPUs and also mounting the actual path with the following command
docker run -it --rm --gpus=all -v "%cd%":/home/ecub ecub:latest /bin/bash
The engines are already contained inside the repository thanks to Git-LFS. You can find the engines also inside the release, or you can rebuild them on your machine. From the root directory (not inside docker):
python scripts/manager.py
python scripts/source.py
python scripts/sink.py
From the root directory (inside the docker):
python scripts/grasping_pipeline.py
From the ISBFSAR directory (inside the docker (you may have to set the PYTHONPATH)):
PYTHONPATH="/home/ecub/ISBFSAR"; python ISBFSAR/main.py